1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3
4 #include <linux/fs.h>
5 #include <linux/kernel.h>
6 #include <linux/sched/signal.h>
7 #include <linux/slab.h>
8 #include <linux/vmalloc.h>
9 #include <linux/wait.h>
10 #include <linux/writeback.h>
11 #include <linux/iversion.h>
12
13 #include "super.h"
14 #include "mds_client.h"
15 #include "cache.h"
16 #include <linux/ceph/decode.h>
17 #include <linux/ceph/messenger.h>
18
19 /*
20 * Capability management
21 *
22 * The Ceph metadata servers control client access to inode metadata
23 * and file data by issuing capabilities, granting clients permission
24 * to read and/or write both inode field and file data to OSDs
25 * (storage nodes). Each capability consists of a set of bits
26 * indicating which operations are allowed.
27 *
28 * If the client holds a *_SHARED cap, the client has a coherent value
29 * that can be safely read from the cached inode.
30 *
31 * In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the
32 * client is allowed to change inode attributes (e.g., file size,
33 * mtime), note its dirty state in the ceph_cap, and asynchronously
34 * flush that metadata change to the MDS.
35 *
36 * In the event of a conflicting operation (perhaps by another
37 * client), the MDS will revoke the conflicting client capabilities.
38 *
39 * In order for a client to cache an inode, it must hold a capability
40 * with at least one MDS server. When inodes are released, release
41 * notifications are batched and periodically sent en masse to the MDS
42 * cluster to release server state.
43 */
44
45 static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc);
46 static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
47 struct ceph_mds_session *session,
48 struct ceph_inode_info *ci,
49 u64 oldest_flush_tid);
50
51 /*
52 * Generate readable cap strings for debugging output.
53 */
54 #define MAX_CAP_STR 20
55 static char cap_str[MAX_CAP_STR][40];
56 static DEFINE_SPINLOCK(cap_str_lock);
57 static int last_cap_str;
58
gcap_string(char * s,int c)59 static char *gcap_string(char *s, int c)
60 {
61 if (c & CEPH_CAP_GSHARED)
62 *s++ = 's';
63 if (c & CEPH_CAP_GEXCL)
64 *s++ = 'x';
65 if (c & CEPH_CAP_GCACHE)
66 *s++ = 'c';
67 if (c & CEPH_CAP_GRD)
68 *s++ = 'r';
69 if (c & CEPH_CAP_GWR)
70 *s++ = 'w';
71 if (c & CEPH_CAP_GBUFFER)
72 *s++ = 'b';
73 if (c & CEPH_CAP_GWREXTEND)
74 *s++ = 'a';
75 if (c & CEPH_CAP_GLAZYIO)
76 *s++ = 'l';
77 return s;
78 }
79
ceph_cap_string(int caps)80 const char *ceph_cap_string(int caps)
81 {
82 int i;
83 char *s;
84 int c;
85
86 spin_lock(&cap_str_lock);
87 i = last_cap_str++;
88 if (last_cap_str == MAX_CAP_STR)
89 last_cap_str = 0;
90 spin_unlock(&cap_str_lock);
91
92 s = cap_str[i];
93
94 if (caps & CEPH_CAP_PIN)
95 *s++ = 'p';
96
97 c = (caps >> CEPH_CAP_SAUTH) & 3;
98 if (c) {
99 *s++ = 'A';
100 s = gcap_string(s, c);
101 }
102
103 c = (caps >> CEPH_CAP_SLINK) & 3;
104 if (c) {
105 *s++ = 'L';
106 s = gcap_string(s, c);
107 }
108
109 c = (caps >> CEPH_CAP_SXATTR) & 3;
110 if (c) {
111 *s++ = 'X';
112 s = gcap_string(s, c);
113 }
114
115 c = caps >> CEPH_CAP_SFILE;
116 if (c) {
117 *s++ = 'F';
118 s = gcap_string(s, c);
119 }
120
121 if (s == cap_str[i])
122 *s++ = '-';
123 *s = 0;
124 return cap_str[i];
125 }
126
ceph_caps_init(struct ceph_mds_client * mdsc)127 void ceph_caps_init(struct ceph_mds_client *mdsc)
128 {
129 INIT_LIST_HEAD(&mdsc->caps_list);
130 spin_lock_init(&mdsc->caps_list_lock);
131 }
132
ceph_caps_finalize(struct ceph_mds_client * mdsc)133 void ceph_caps_finalize(struct ceph_mds_client *mdsc)
134 {
135 struct ceph_cap *cap;
136
137 spin_lock(&mdsc->caps_list_lock);
138 while (!list_empty(&mdsc->caps_list)) {
139 cap = list_first_entry(&mdsc->caps_list,
140 struct ceph_cap, caps_item);
141 list_del(&cap->caps_item);
142 kmem_cache_free(ceph_cap_cachep, cap);
143 }
144 mdsc->caps_total_count = 0;
145 mdsc->caps_avail_count = 0;
146 mdsc->caps_use_count = 0;
147 mdsc->caps_reserve_count = 0;
148 mdsc->caps_min_count = 0;
149 spin_unlock(&mdsc->caps_list_lock);
150 }
151
ceph_adjust_caps_max_min(struct ceph_mds_client * mdsc,struct ceph_mount_options * fsopt)152 void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
153 struct ceph_mount_options *fsopt)
154 {
155 spin_lock(&mdsc->caps_list_lock);
156 mdsc->caps_min_count = fsopt->max_readdir;
157 if (mdsc->caps_min_count < 1024)
158 mdsc->caps_min_count = 1024;
159 mdsc->caps_use_max = fsopt->caps_max;
160 if (mdsc->caps_use_max > 0 &&
161 mdsc->caps_use_max < mdsc->caps_min_count)
162 mdsc->caps_use_max = mdsc->caps_min_count;
163 spin_unlock(&mdsc->caps_list_lock);
164 }
165
__ceph_unreserve_caps(struct ceph_mds_client * mdsc,int nr_caps)166 static void __ceph_unreserve_caps(struct ceph_mds_client *mdsc, int nr_caps)
167 {
168 struct ceph_cap *cap;
169 int i;
170
171 if (nr_caps) {
172 BUG_ON(mdsc->caps_reserve_count < nr_caps);
173 mdsc->caps_reserve_count -= nr_caps;
174 if (mdsc->caps_avail_count >=
175 mdsc->caps_reserve_count + mdsc->caps_min_count) {
176 mdsc->caps_total_count -= nr_caps;
177 for (i = 0; i < nr_caps; i++) {
178 cap = list_first_entry(&mdsc->caps_list,
179 struct ceph_cap, caps_item);
180 list_del(&cap->caps_item);
181 kmem_cache_free(ceph_cap_cachep, cap);
182 }
183 } else {
184 mdsc->caps_avail_count += nr_caps;
185 }
186
187 dout("%s: caps %d = %d used + %d resv + %d avail\n",
188 __func__,
189 mdsc->caps_total_count, mdsc->caps_use_count,
190 mdsc->caps_reserve_count, mdsc->caps_avail_count);
191 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
192 mdsc->caps_reserve_count +
193 mdsc->caps_avail_count);
194 }
195 }
196
197 /*
198 * Called under mdsc->mutex.
199 */
ceph_reserve_caps(struct ceph_mds_client * mdsc,struct ceph_cap_reservation * ctx,int need)200 int ceph_reserve_caps(struct ceph_mds_client *mdsc,
201 struct ceph_cap_reservation *ctx, int need)
202 {
203 int i, j;
204 struct ceph_cap *cap;
205 int have;
206 int alloc = 0;
207 int max_caps;
208 int err = 0;
209 bool trimmed = false;
210 struct ceph_mds_session *s;
211 LIST_HEAD(newcaps);
212
213 dout("reserve caps ctx=%p need=%d\n", ctx, need);
214
215 /* first reserve any caps that are already allocated */
216 spin_lock(&mdsc->caps_list_lock);
217 if (mdsc->caps_avail_count >= need)
218 have = need;
219 else
220 have = mdsc->caps_avail_count;
221 mdsc->caps_avail_count -= have;
222 mdsc->caps_reserve_count += have;
223 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
224 mdsc->caps_reserve_count +
225 mdsc->caps_avail_count);
226 spin_unlock(&mdsc->caps_list_lock);
227
228 for (i = have; i < need; ) {
229 cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
230 if (cap) {
231 list_add(&cap->caps_item, &newcaps);
232 alloc++;
233 i++;
234 continue;
235 }
236
237 if (!trimmed) {
238 for (j = 0; j < mdsc->max_sessions; j++) {
239 s = __ceph_lookup_mds_session(mdsc, j);
240 if (!s)
241 continue;
242 mutex_unlock(&mdsc->mutex);
243
244 mutex_lock(&s->s_mutex);
245 max_caps = s->s_nr_caps - (need - i);
246 ceph_trim_caps(mdsc, s, max_caps);
247 mutex_unlock(&s->s_mutex);
248
249 ceph_put_mds_session(s);
250 mutex_lock(&mdsc->mutex);
251 }
252 trimmed = true;
253
254 spin_lock(&mdsc->caps_list_lock);
255 if (mdsc->caps_avail_count) {
256 int more_have;
257 if (mdsc->caps_avail_count >= need - i)
258 more_have = need - i;
259 else
260 more_have = mdsc->caps_avail_count;
261
262 i += more_have;
263 have += more_have;
264 mdsc->caps_avail_count -= more_have;
265 mdsc->caps_reserve_count += more_have;
266
267 }
268 spin_unlock(&mdsc->caps_list_lock);
269
270 continue;
271 }
272
273 pr_warn("reserve caps ctx=%p ENOMEM need=%d got=%d\n",
274 ctx, need, have + alloc);
275 err = -ENOMEM;
276 break;
277 }
278
279 if (!err) {
280 BUG_ON(have + alloc != need);
281 ctx->count = need;
282 ctx->used = 0;
283 }
284
285 spin_lock(&mdsc->caps_list_lock);
286 mdsc->caps_total_count += alloc;
287 mdsc->caps_reserve_count += alloc;
288 list_splice(&newcaps, &mdsc->caps_list);
289
290 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
291 mdsc->caps_reserve_count +
292 mdsc->caps_avail_count);
293
294 if (err)
295 __ceph_unreserve_caps(mdsc, have + alloc);
296
297 spin_unlock(&mdsc->caps_list_lock);
298
299 dout("reserve caps ctx=%p %d = %d used + %d resv + %d avail\n",
300 ctx, mdsc->caps_total_count, mdsc->caps_use_count,
301 mdsc->caps_reserve_count, mdsc->caps_avail_count);
302 return err;
303 }
304
ceph_unreserve_caps(struct ceph_mds_client * mdsc,struct ceph_cap_reservation * ctx)305 void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
306 struct ceph_cap_reservation *ctx)
307 {
308 bool reclaim = false;
309 if (!ctx->count)
310 return;
311
312 dout("unreserve caps ctx=%p count=%d\n", ctx, ctx->count);
313 spin_lock(&mdsc->caps_list_lock);
314 __ceph_unreserve_caps(mdsc, ctx->count);
315 ctx->count = 0;
316
317 if (mdsc->caps_use_max > 0 &&
318 mdsc->caps_use_count > mdsc->caps_use_max)
319 reclaim = true;
320 spin_unlock(&mdsc->caps_list_lock);
321
322 if (reclaim)
323 ceph_reclaim_caps_nr(mdsc, ctx->used);
324 }
325
ceph_get_cap(struct ceph_mds_client * mdsc,struct ceph_cap_reservation * ctx)326 struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
327 struct ceph_cap_reservation *ctx)
328 {
329 struct ceph_cap *cap = NULL;
330
331 /* temporary, until we do something about cap import/export */
332 if (!ctx) {
333 cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
334 if (cap) {
335 spin_lock(&mdsc->caps_list_lock);
336 mdsc->caps_use_count++;
337 mdsc->caps_total_count++;
338 spin_unlock(&mdsc->caps_list_lock);
339 } else {
340 spin_lock(&mdsc->caps_list_lock);
341 if (mdsc->caps_avail_count) {
342 BUG_ON(list_empty(&mdsc->caps_list));
343
344 mdsc->caps_avail_count--;
345 mdsc->caps_use_count++;
346 cap = list_first_entry(&mdsc->caps_list,
347 struct ceph_cap, caps_item);
348 list_del(&cap->caps_item);
349
350 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
351 mdsc->caps_reserve_count + mdsc->caps_avail_count);
352 }
353 spin_unlock(&mdsc->caps_list_lock);
354 }
355
356 return cap;
357 }
358
359 spin_lock(&mdsc->caps_list_lock);
360 dout("get_cap ctx=%p (%d) %d = %d used + %d resv + %d avail\n",
361 ctx, ctx->count, mdsc->caps_total_count, mdsc->caps_use_count,
362 mdsc->caps_reserve_count, mdsc->caps_avail_count);
363 BUG_ON(!ctx->count);
364 BUG_ON(ctx->count > mdsc->caps_reserve_count);
365 BUG_ON(list_empty(&mdsc->caps_list));
366
367 ctx->count--;
368 ctx->used++;
369 mdsc->caps_reserve_count--;
370 mdsc->caps_use_count++;
371
372 cap = list_first_entry(&mdsc->caps_list, struct ceph_cap, caps_item);
373 list_del(&cap->caps_item);
374
375 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
376 mdsc->caps_reserve_count + mdsc->caps_avail_count);
377 spin_unlock(&mdsc->caps_list_lock);
378 return cap;
379 }
380
ceph_put_cap(struct ceph_mds_client * mdsc,struct ceph_cap * cap)381 void ceph_put_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap)
382 {
383 spin_lock(&mdsc->caps_list_lock);
384 dout("put_cap %p %d = %d used + %d resv + %d avail\n",
385 cap, mdsc->caps_total_count, mdsc->caps_use_count,
386 mdsc->caps_reserve_count, mdsc->caps_avail_count);
387 mdsc->caps_use_count--;
388 /*
389 * Keep some preallocated caps around (ceph_min_count), to
390 * avoid lots of free/alloc churn.
391 */
392 if (mdsc->caps_avail_count >= mdsc->caps_reserve_count +
393 mdsc->caps_min_count) {
394 mdsc->caps_total_count--;
395 kmem_cache_free(ceph_cap_cachep, cap);
396 } else {
397 mdsc->caps_avail_count++;
398 list_add(&cap->caps_item, &mdsc->caps_list);
399 }
400
401 BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
402 mdsc->caps_reserve_count + mdsc->caps_avail_count);
403 spin_unlock(&mdsc->caps_list_lock);
404 }
405
ceph_reservation_status(struct ceph_fs_client * fsc,int * total,int * avail,int * used,int * reserved,int * min)406 void ceph_reservation_status(struct ceph_fs_client *fsc,
407 int *total, int *avail, int *used, int *reserved,
408 int *min)
409 {
410 struct ceph_mds_client *mdsc = fsc->mdsc;
411
412 spin_lock(&mdsc->caps_list_lock);
413
414 if (total)
415 *total = mdsc->caps_total_count;
416 if (avail)
417 *avail = mdsc->caps_avail_count;
418 if (used)
419 *used = mdsc->caps_use_count;
420 if (reserved)
421 *reserved = mdsc->caps_reserve_count;
422 if (min)
423 *min = mdsc->caps_min_count;
424
425 spin_unlock(&mdsc->caps_list_lock);
426 }
427
428 /*
429 * Find ceph_cap for given mds, if any.
430 *
431 * Called with i_ceph_lock held.
432 */
__get_cap_for_mds(struct ceph_inode_info * ci,int mds)433 static struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, int mds)
434 {
435 struct ceph_cap *cap;
436 struct rb_node *n = ci->i_caps.rb_node;
437
438 while (n) {
439 cap = rb_entry(n, struct ceph_cap, ci_node);
440 if (mds < cap->mds)
441 n = n->rb_left;
442 else if (mds > cap->mds)
443 n = n->rb_right;
444 else
445 return cap;
446 }
447 return NULL;
448 }
449
ceph_get_cap_for_mds(struct ceph_inode_info * ci,int mds)450 struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, int mds)
451 {
452 struct ceph_cap *cap;
453
454 spin_lock(&ci->i_ceph_lock);
455 cap = __get_cap_for_mds(ci, mds);
456 spin_unlock(&ci->i_ceph_lock);
457 return cap;
458 }
459
460 /*
461 * Called under i_ceph_lock.
462 */
__insert_cap_node(struct ceph_inode_info * ci,struct ceph_cap * new)463 static void __insert_cap_node(struct ceph_inode_info *ci,
464 struct ceph_cap *new)
465 {
466 struct rb_node **p = &ci->i_caps.rb_node;
467 struct rb_node *parent = NULL;
468 struct ceph_cap *cap = NULL;
469
470 while (*p) {
471 parent = *p;
472 cap = rb_entry(parent, struct ceph_cap, ci_node);
473 if (new->mds < cap->mds)
474 p = &(*p)->rb_left;
475 else if (new->mds > cap->mds)
476 p = &(*p)->rb_right;
477 else
478 BUG();
479 }
480
481 rb_link_node(&new->ci_node, parent, p);
482 rb_insert_color(&new->ci_node, &ci->i_caps);
483 }
484
485 /*
486 * (re)set cap hold timeouts, which control the delayed release
487 * of unused caps back to the MDS. Should be called on cap use.
488 */
__cap_set_timeouts(struct ceph_mds_client * mdsc,struct ceph_inode_info * ci)489 static void __cap_set_timeouts(struct ceph_mds_client *mdsc,
490 struct ceph_inode_info *ci)
491 {
492 struct ceph_mount_options *opt = mdsc->fsc->mount_options;
493 ci->i_hold_caps_max = round_jiffies(jiffies +
494 opt->caps_wanted_delay_max * HZ);
495 dout("__cap_set_timeouts %p %lu\n", &ci->netfs.inode,
496 ci->i_hold_caps_max - jiffies);
497 }
498
499 /*
500 * (Re)queue cap at the end of the delayed cap release list.
501 *
502 * If I_FLUSH is set, leave the inode at the front of the list.
503 *
504 * Caller holds i_ceph_lock
505 * -> we take mdsc->cap_delay_lock
506 */
__cap_delay_requeue(struct ceph_mds_client * mdsc,struct ceph_inode_info * ci)507 static void __cap_delay_requeue(struct ceph_mds_client *mdsc,
508 struct ceph_inode_info *ci)
509 {
510 dout("__cap_delay_requeue %p flags 0x%lx at %lu\n", &ci->netfs.inode,
511 ci->i_ceph_flags, ci->i_hold_caps_max);
512 if (!mdsc->stopping) {
513 spin_lock(&mdsc->cap_delay_lock);
514 if (!list_empty(&ci->i_cap_delay_list)) {
515 if (ci->i_ceph_flags & CEPH_I_FLUSH)
516 goto no_change;
517 list_del_init(&ci->i_cap_delay_list);
518 }
519 __cap_set_timeouts(mdsc, ci);
520 list_add_tail(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
521 no_change:
522 spin_unlock(&mdsc->cap_delay_lock);
523 }
524 }
525
526 /*
527 * Queue an inode for immediate writeback. Mark inode with I_FLUSH,
528 * indicating we should send a cap message to flush dirty metadata
529 * asap, and move to the front of the delayed cap list.
530 */
__cap_delay_requeue_front(struct ceph_mds_client * mdsc,struct ceph_inode_info * ci)531 static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc,
532 struct ceph_inode_info *ci)
533 {
534 dout("__cap_delay_requeue_front %p\n", &ci->netfs.inode);
535 spin_lock(&mdsc->cap_delay_lock);
536 ci->i_ceph_flags |= CEPH_I_FLUSH;
537 if (!list_empty(&ci->i_cap_delay_list))
538 list_del_init(&ci->i_cap_delay_list);
539 list_add(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
540 spin_unlock(&mdsc->cap_delay_lock);
541 }
542
543 /*
544 * Cancel delayed work on cap.
545 *
546 * Caller must hold i_ceph_lock.
547 */
__cap_delay_cancel(struct ceph_mds_client * mdsc,struct ceph_inode_info * ci)548 static void __cap_delay_cancel(struct ceph_mds_client *mdsc,
549 struct ceph_inode_info *ci)
550 {
551 dout("__cap_delay_cancel %p\n", &ci->netfs.inode);
552 if (list_empty(&ci->i_cap_delay_list))
553 return;
554 spin_lock(&mdsc->cap_delay_lock);
555 list_del_init(&ci->i_cap_delay_list);
556 spin_unlock(&mdsc->cap_delay_lock);
557 }
558
559 /* Common issue checks for add_cap, handle_cap_grant. */
__check_cap_issue(struct ceph_inode_info * ci,struct ceph_cap * cap,unsigned issued)560 static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap,
561 unsigned issued)
562 {
563 unsigned had = __ceph_caps_issued(ci, NULL);
564
565 lockdep_assert_held(&ci->i_ceph_lock);
566
567 /*
568 * Each time we receive FILE_CACHE anew, we increment
569 * i_rdcache_gen.
570 */
571 if (S_ISREG(ci->netfs.inode.i_mode) &&
572 (issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
573 (had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) {
574 ci->i_rdcache_gen++;
575 }
576
577 /*
578 * If FILE_SHARED is newly issued, mark dir not complete. We don't
579 * know what happened to this directory while we didn't have the cap.
580 * If FILE_SHARED is being revoked, also mark dir not complete. It
581 * stops on-going cached readdir.
582 */
583 if ((issued & CEPH_CAP_FILE_SHARED) != (had & CEPH_CAP_FILE_SHARED)) {
584 if (issued & CEPH_CAP_FILE_SHARED)
585 atomic_inc(&ci->i_shared_gen);
586 if (S_ISDIR(ci->netfs.inode.i_mode)) {
587 dout(" marking %p NOT complete\n", &ci->netfs.inode);
588 __ceph_dir_clear_complete(ci);
589 }
590 }
591
592 /* Wipe saved layout if we're losing DIR_CREATE caps */
593 if (S_ISDIR(ci->netfs.inode.i_mode) && (had & CEPH_CAP_DIR_CREATE) &&
594 !(issued & CEPH_CAP_DIR_CREATE)) {
595 ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
596 memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
597 }
598 }
599
600 /**
601 * change_auth_cap_ses - move inode to appropriate lists when auth caps change
602 * @ci: inode to be moved
603 * @session: new auth caps session
604 */
change_auth_cap_ses(struct ceph_inode_info * ci,struct ceph_mds_session * session)605 static void change_auth_cap_ses(struct ceph_inode_info *ci,
606 struct ceph_mds_session *session)
607 {
608 lockdep_assert_held(&ci->i_ceph_lock);
609
610 if (list_empty(&ci->i_dirty_item) && list_empty(&ci->i_flushing_item))
611 return;
612
613 spin_lock(&session->s_mdsc->cap_dirty_lock);
614 if (!list_empty(&ci->i_dirty_item))
615 list_move(&ci->i_dirty_item, &session->s_cap_dirty);
616 if (!list_empty(&ci->i_flushing_item))
617 list_move_tail(&ci->i_flushing_item, &session->s_cap_flushing);
618 spin_unlock(&session->s_mdsc->cap_dirty_lock);
619 }
620
621 /*
622 * Add a capability under the given MDS session.
623 *
624 * Caller should hold session snap_rwsem (read) and ci->i_ceph_lock
625 *
626 * @fmode is the open file mode, if we are opening a file, otherwise
627 * it is < 0. (This is so we can atomically add the cap and add an
628 * open file reference to it.)
629 */
ceph_add_cap(struct inode * inode,struct ceph_mds_session * session,u64 cap_id,unsigned issued,unsigned wanted,unsigned seq,unsigned mseq,u64 realmino,int flags,struct ceph_cap ** new_cap)630 void ceph_add_cap(struct inode *inode,
631 struct ceph_mds_session *session, u64 cap_id,
632 unsigned issued, unsigned wanted,
633 unsigned seq, unsigned mseq, u64 realmino, int flags,
634 struct ceph_cap **new_cap)
635 {
636 struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
637 struct ceph_inode_info *ci = ceph_inode(inode);
638 struct ceph_cap *cap;
639 int mds = session->s_mds;
640 int actual_wanted;
641 u32 gen;
642
643 lockdep_assert_held(&ci->i_ceph_lock);
644
645 dout("add_cap %p mds%d cap %llx %s seq %d\n", inode,
646 session->s_mds, cap_id, ceph_cap_string(issued), seq);
647
648 gen = atomic_read(&session->s_cap_gen);
649
650 cap = __get_cap_for_mds(ci, mds);
651 if (!cap) {
652 cap = *new_cap;
653 *new_cap = NULL;
654
655 cap->issued = 0;
656 cap->implemented = 0;
657 cap->mds = mds;
658 cap->mds_wanted = 0;
659 cap->mseq = 0;
660
661 cap->ci = ci;
662 __insert_cap_node(ci, cap);
663
664 /* add to session cap list */
665 cap->session = session;
666 spin_lock(&session->s_cap_lock);
667 list_add_tail(&cap->session_caps, &session->s_caps);
668 session->s_nr_caps++;
669 atomic64_inc(&mdsc->metric.total_caps);
670 spin_unlock(&session->s_cap_lock);
671 } else {
672 spin_lock(&session->s_cap_lock);
673 list_move_tail(&cap->session_caps, &session->s_caps);
674 spin_unlock(&session->s_cap_lock);
675
676 if (cap->cap_gen < gen)
677 cap->issued = cap->implemented = CEPH_CAP_PIN;
678
679 /*
680 * auth mds of the inode changed. we received the cap export
681 * message, but still haven't received the cap import message.
682 * handle_cap_export() updated the new auth MDS' cap.
683 *
684 * "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing
685 * a message that was send before the cap import message. So
686 * don't remove caps.
687 */
688 if (ceph_seq_cmp(seq, cap->seq) <= 0) {
689 WARN_ON(cap != ci->i_auth_cap);
690 WARN_ON(cap->cap_id != cap_id);
691 seq = cap->seq;
692 mseq = cap->mseq;
693 issued |= cap->issued;
694 flags |= CEPH_CAP_FLAG_AUTH;
695 }
696 }
697
698 if (!ci->i_snap_realm ||
699 ((flags & CEPH_CAP_FLAG_AUTH) &&
700 realmino != (u64)-1 && ci->i_snap_realm->ino != realmino)) {
701 /*
702 * add this inode to the appropriate snap realm
703 */
704 struct ceph_snap_realm *realm = ceph_lookup_snap_realm(mdsc,
705 realmino);
706 if (realm)
707 ceph_change_snap_realm(inode, realm);
708 else
709 WARN(1, "%s: couldn't find snap realm 0x%llx (ino 0x%llx oldrealm 0x%llx)\n",
710 __func__, realmino, ci->i_vino.ino,
711 ci->i_snap_realm ? ci->i_snap_realm->ino : 0);
712 }
713
714 __check_cap_issue(ci, cap, issued);
715
716 /*
717 * If we are issued caps we don't want, or the mds' wanted
718 * value appears to be off, queue a check so we'll release
719 * later and/or update the mds wanted value.
720 */
721 actual_wanted = __ceph_caps_wanted(ci);
722 if ((wanted & ~actual_wanted) ||
723 (issued & ~actual_wanted & CEPH_CAP_ANY_WR)) {
724 dout(" issued %s, mds wanted %s, actual %s, queueing\n",
725 ceph_cap_string(issued), ceph_cap_string(wanted),
726 ceph_cap_string(actual_wanted));
727 __cap_delay_requeue(mdsc, ci);
728 }
729
730 if (flags & CEPH_CAP_FLAG_AUTH) {
731 if (!ci->i_auth_cap ||
732 ceph_seq_cmp(ci->i_auth_cap->mseq, mseq) < 0) {
733 if (ci->i_auth_cap &&
734 ci->i_auth_cap->session != cap->session)
735 change_auth_cap_ses(ci, cap->session);
736 ci->i_auth_cap = cap;
737 cap->mds_wanted = wanted;
738 }
739 } else {
740 WARN_ON(ci->i_auth_cap == cap);
741 }
742
743 dout("add_cap inode %p (%llx.%llx) cap %p %s now %s seq %d mds%d\n",
744 inode, ceph_vinop(inode), cap, ceph_cap_string(issued),
745 ceph_cap_string(issued|cap->issued), seq, mds);
746 cap->cap_id = cap_id;
747 cap->issued = issued;
748 cap->implemented |= issued;
749 if (ceph_seq_cmp(mseq, cap->mseq) > 0)
750 cap->mds_wanted = wanted;
751 else
752 cap->mds_wanted |= wanted;
753 cap->seq = seq;
754 cap->issue_seq = seq;
755 cap->mseq = mseq;
756 cap->cap_gen = gen;
757 }
758
759 /*
760 * Return true if cap has not timed out and belongs to the current
761 * generation of the MDS session (i.e. has not gone 'stale' due to
762 * us losing touch with the mds).
763 */
__cap_is_valid(struct ceph_cap * cap)764 static int __cap_is_valid(struct ceph_cap *cap)
765 {
766 unsigned long ttl;
767 u32 gen;
768
769 gen = atomic_read(&cap->session->s_cap_gen);
770 ttl = cap->session->s_cap_ttl;
771
772 if (cap->cap_gen < gen || time_after_eq(jiffies, ttl)) {
773 dout("__cap_is_valid %p cap %p issued %s "
774 "but STALE (gen %u vs %u)\n", &cap->ci->netfs.inode,
775 cap, ceph_cap_string(cap->issued), cap->cap_gen, gen);
776 return 0;
777 }
778
779 return 1;
780 }
781
782 /*
783 * Return set of valid cap bits issued to us. Note that caps time
784 * out, and may be invalidated in bulk if the client session times out
785 * and session->s_cap_gen is bumped.
786 */
__ceph_caps_issued(struct ceph_inode_info * ci,int * implemented)787 int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented)
788 {
789 int have = ci->i_snap_caps;
790 struct ceph_cap *cap;
791 struct rb_node *p;
792
793 if (implemented)
794 *implemented = 0;
795 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
796 cap = rb_entry(p, struct ceph_cap, ci_node);
797 if (!__cap_is_valid(cap))
798 continue;
799 dout("__ceph_caps_issued %p cap %p issued %s\n",
800 &ci->netfs.inode, cap, ceph_cap_string(cap->issued));
801 have |= cap->issued;
802 if (implemented)
803 *implemented |= cap->implemented;
804 }
805 /*
806 * exclude caps issued by non-auth MDS, but are been revoking
807 * by the auth MDS. The non-auth MDS should be revoking/exporting
808 * these caps, but the message is delayed.
809 */
810 if (ci->i_auth_cap) {
811 cap = ci->i_auth_cap;
812 have &= ~cap->implemented | cap->issued;
813 }
814 return have;
815 }
816
817 /*
818 * Get cap bits issued by caps other than @ocap
819 */
__ceph_caps_issued_other(struct ceph_inode_info * ci,struct ceph_cap * ocap)820 int __ceph_caps_issued_other(struct ceph_inode_info *ci, struct ceph_cap *ocap)
821 {
822 int have = ci->i_snap_caps;
823 struct ceph_cap *cap;
824 struct rb_node *p;
825
826 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
827 cap = rb_entry(p, struct ceph_cap, ci_node);
828 if (cap == ocap)
829 continue;
830 if (!__cap_is_valid(cap))
831 continue;
832 have |= cap->issued;
833 }
834 return have;
835 }
836
837 /*
838 * Move a cap to the end of the LRU (oldest caps at list head, newest
839 * at list tail).
840 */
__touch_cap(struct ceph_cap * cap)841 static void __touch_cap(struct ceph_cap *cap)
842 {
843 struct ceph_mds_session *s = cap->session;
844
845 spin_lock(&s->s_cap_lock);
846 if (!s->s_cap_iterator) {
847 dout("__touch_cap %p cap %p mds%d\n", &cap->ci->netfs.inode, cap,
848 s->s_mds);
849 list_move_tail(&cap->session_caps, &s->s_caps);
850 } else {
851 dout("__touch_cap %p cap %p mds%d NOP, iterating over caps\n",
852 &cap->ci->netfs.inode, cap, s->s_mds);
853 }
854 spin_unlock(&s->s_cap_lock);
855 }
856
857 /*
858 * Check if we hold the given mask. If so, move the cap(s) to the
859 * front of their respective LRUs. (This is the preferred way for
860 * callers to check for caps they want.)
861 */
__ceph_caps_issued_mask(struct ceph_inode_info * ci,int mask,int touch)862 int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
863 {
864 struct ceph_cap *cap;
865 struct rb_node *p;
866 int have = ci->i_snap_caps;
867
868 if ((have & mask) == mask) {
869 dout("__ceph_caps_issued_mask ino 0x%llx snap issued %s"
870 " (mask %s)\n", ceph_ino(&ci->netfs.inode),
871 ceph_cap_string(have),
872 ceph_cap_string(mask));
873 return 1;
874 }
875
876 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
877 cap = rb_entry(p, struct ceph_cap, ci_node);
878 if (!__cap_is_valid(cap))
879 continue;
880 if ((cap->issued & mask) == mask) {
881 dout("__ceph_caps_issued_mask ino 0x%llx cap %p issued %s"
882 " (mask %s)\n", ceph_ino(&ci->netfs.inode), cap,
883 ceph_cap_string(cap->issued),
884 ceph_cap_string(mask));
885 if (touch)
886 __touch_cap(cap);
887 return 1;
888 }
889
890 /* does a combination of caps satisfy mask? */
891 have |= cap->issued;
892 if ((have & mask) == mask) {
893 dout("__ceph_caps_issued_mask ino 0x%llx combo issued %s"
894 " (mask %s)\n", ceph_ino(&ci->netfs.inode),
895 ceph_cap_string(cap->issued),
896 ceph_cap_string(mask));
897 if (touch) {
898 struct rb_node *q;
899
900 /* touch this + preceding caps */
901 __touch_cap(cap);
902 for (q = rb_first(&ci->i_caps); q != p;
903 q = rb_next(q)) {
904 cap = rb_entry(q, struct ceph_cap,
905 ci_node);
906 if (!__cap_is_valid(cap))
907 continue;
908 if (cap->issued & mask)
909 __touch_cap(cap);
910 }
911 }
912 return 1;
913 }
914 }
915
916 return 0;
917 }
918
__ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)919 int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
920 int touch)
921 {
922 struct ceph_fs_client *fsc = ceph_sb_to_client(ci->netfs.inode.i_sb);
923 int r;
924
925 r = __ceph_caps_issued_mask(ci, mask, touch);
926 if (r)
927 ceph_update_cap_hit(&fsc->mdsc->metric);
928 else
929 ceph_update_cap_mis(&fsc->mdsc->metric);
930 return r;
931 }
932
933 /*
934 * Return true if mask caps are currently being revoked by an MDS.
935 */
__ceph_caps_revoking_other(struct ceph_inode_info * ci,struct ceph_cap * ocap,int mask)936 int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
937 struct ceph_cap *ocap, int mask)
938 {
939 struct ceph_cap *cap;
940 struct rb_node *p;
941
942 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
943 cap = rb_entry(p, struct ceph_cap, ci_node);
944 if (cap != ocap &&
945 (cap->implemented & ~cap->issued & mask))
946 return 1;
947 }
948 return 0;
949 }
950
ceph_caps_revoking(struct ceph_inode_info * ci,int mask)951 int ceph_caps_revoking(struct ceph_inode_info *ci, int mask)
952 {
953 struct inode *inode = &ci->netfs.inode;
954 int ret;
955
956 spin_lock(&ci->i_ceph_lock);
957 ret = __ceph_caps_revoking_other(ci, NULL, mask);
958 spin_unlock(&ci->i_ceph_lock);
959 dout("ceph_caps_revoking %p %s = %d\n", inode,
960 ceph_cap_string(mask), ret);
961 return ret;
962 }
963
__ceph_caps_used(struct ceph_inode_info * ci)964 int __ceph_caps_used(struct ceph_inode_info *ci)
965 {
966 int used = 0;
967 if (ci->i_pin_ref)
968 used |= CEPH_CAP_PIN;
969 if (ci->i_rd_ref)
970 used |= CEPH_CAP_FILE_RD;
971 if (ci->i_rdcache_ref ||
972 (S_ISREG(ci->netfs.inode.i_mode) &&
973 ci->netfs.inode.i_data.nrpages))
974 used |= CEPH_CAP_FILE_CACHE;
975 if (ci->i_wr_ref)
976 used |= CEPH_CAP_FILE_WR;
977 if (ci->i_wb_ref || ci->i_wrbuffer_ref)
978 used |= CEPH_CAP_FILE_BUFFER;
979 if (ci->i_fx_ref)
980 used |= CEPH_CAP_FILE_EXCL;
981 return used;
982 }
983
984 #define FMODE_WAIT_BIAS 1000
985
986 /*
987 * wanted, by virtue of open file modes
988 */
__ceph_caps_file_wanted(struct ceph_inode_info * ci)989 int __ceph_caps_file_wanted(struct ceph_inode_info *ci)
990 {
991 const int PIN_SHIFT = ffs(CEPH_FILE_MODE_PIN);
992 const int RD_SHIFT = ffs(CEPH_FILE_MODE_RD);
993 const int WR_SHIFT = ffs(CEPH_FILE_MODE_WR);
994 const int LAZY_SHIFT = ffs(CEPH_FILE_MODE_LAZY);
995 struct ceph_mount_options *opt =
996 ceph_inode_to_client(&ci->netfs.inode)->mount_options;
997 unsigned long used_cutoff = jiffies - opt->caps_wanted_delay_max * HZ;
998 unsigned long idle_cutoff = jiffies - opt->caps_wanted_delay_min * HZ;
999
1000 if (S_ISDIR(ci->netfs.inode.i_mode)) {
1001 int want = 0;
1002
1003 /* use used_cutoff here, to keep dir's wanted caps longer */
1004 if (ci->i_nr_by_mode[RD_SHIFT] > 0 ||
1005 time_after(ci->i_last_rd, used_cutoff))
1006 want |= CEPH_CAP_ANY_SHARED;
1007
1008 if (ci->i_nr_by_mode[WR_SHIFT] > 0 ||
1009 time_after(ci->i_last_wr, used_cutoff)) {
1010 want |= CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_EXCL;
1011 if (opt->flags & CEPH_MOUNT_OPT_ASYNC_DIROPS)
1012 want |= CEPH_CAP_ANY_DIR_OPS;
1013 }
1014
1015 if (want || ci->i_nr_by_mode[PIN_SHIFT] > 0)
1016 want |= CEPH_CAP_PIN;
1017
1018 return want;
1019 } else {
1020 int bits = 0;
1021
1022 if (ci->i_nr_by_mode[RD_SHIFT] > 0) {
1023 if (ci->i_nr_by_mode[RD_SHIFT] >= FMODE_WAIT_BIAS ||
1024 time_after(ci->i_last_rd, used_cutoff))
1025 bits |= 1 << RD_SHIFT;
1026 } else if (time_after(ci->i_last_rd, idle_cutoff)) {
1027 bits |= 1 << RD_SHIFT;
1028 }
1029
1030 if (ci->i_nr_by_mode[WR_SHIFT] > 0) {
1031 if (ci->i_nr_by_mode[WR_SHIFT] >= FMODE_WAIT_BIAS ||
1032 time_after(ci->i_last_wr, used_cutoff))
1033 bits |= 1 << WR_SHIFT;
1034 } else if (time_after(ci->i_last_wr, idle_cutoff)) {
1035 bits |= 1 << WR_SHIFT;
1036 }
1037
1038 /* check lazyio only when read/write is wanted */
1039 if ((bits & (CEPH_FILE_MODE_RDWR << 1)) &&
1040 ci->i_nr_by_mode[LAZY_SHIFT] > 0)
1041 bits |= 1 << LAZY_SHIFT;
1042
1043 return bits ? ceph_caps_for_mode(bits >> 1) : 0;
1044 }
1045 }
1046
1047 /*
1048 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
1049 */
__ceph_caps_wanted(struct ceph_inode_info * ci)1050 int __ceph_caps_wanted(struct ceph_inode_info *ci)
1051 {
1052 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
1053 if (S_ISDIR(ci->netfs.inode.i_mode)) {
1054 /* we want EXCL if holding caps of dir ops */
1055 if (w & CEPH_CAP_ANY_DIR_OPS)
1056 w |= CEPH_CAP_FILE_EXCL;
1057 } else {
1058 /* we want EXCL if dirty data */
1059 if (w & CEPH_CAP_FILE_BUFFER)
1060 w |= CEPH_CAP_FILE_EXCL;
1061 }
1062 return w;
1063 }
1064
1065 /*
1066 * Return caps we have registered with the MDS(s) as 'wanted'.
1067 */
__ceph_caps_mds_wanted(struct ceph_inode_info * ci,bool check)1068 int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check)
1069 {
1070 struct ceph_cap *cap;
1071 struct rb_node *p;
1072 int mds_wanted = 0;
1073
1074 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
1075 cap = rb_entry(p, struct ceph_cap, ci_node);
1076 if (check && !__cap_is_valid(cap))
1077 continue;
1078 if (cap == ci->i_auth_cap)
1079 mds_wanted |= cap->mds_wanted;
1080 else
1081 mds_wanted |= (cap->mds_wanted & ~CEPH_CAP_ANY_FILE_WR);
1082 }
1083 return mds_wanted;
1084 }
1085
ceph_is_any_caps(struct inode * inode)1086 int ceph_is_any_caps(struct inode *inode)
1087 {
1088 struct ceph_inode_info *ci = ceph_inode(inode);
1089 int ret;
1090
1091 spin_lock(&ci->i_ceph_lock);
1092 ret = __ceph_is_any_real_caps(ci);
1093 spin_unlock(&ci->i_ceph_lock);
1094
1095 return ret;
1096 }
1097
1098 /*
1099 * Remove a cap. Take steps to deal with a racing iterate_session_caps.
1100 *
1101 * caller should hold i_ceph_lock.
1102 * caller will not hold session s_mutex if called from destroy_inode.
1103 */
__ceph_remove_cap(struct ceph_cap * cap,bool queue_release)1104 void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
1105 {
1106 struct ceph_mds_session *session = cap->session;
1107 struct ceph_inode_info *ci = cap->ci;
1108 struct ceph_mds_client *mdsc;
1109 int removed = 0;
1110
1111 /* 'ci' being NULL means the remove have already occurred */
1112 if (!ci) {
1113 dout("%s: cap inode is NULL\n", __func__);
1114 return;
1115 }
1116
1117 lockdep_assert_held(&ci->i_ceph_lock);
1118
1119 dout("__ceph_remove_cap %p from %p\n", cap, &ci->netfs.inode);
1120
1121 mdsc = ceph_inode_to_client(&ci->netfs.inode)->mdsc;
1122
1123 /* remove from inode's cap rbtree, and clear auth cap */
1124 rb_erase(&cap->ci_node, &ci->i_caps);
1125 if (ci->i_auth_cap == cap)
1126 ci->i_auth_cap = NULL;
1127
1128 /* remove from session list */
1129 spin_lock(&session->s_cap_lock);
1130 if (session->s_cap_iterator == cap) {
1131 /* not yet, we are iterating over this very cap */
1132 dout("__ceph_remove_cap delaying %p removal from session %p\n",
1133 cap, cap->session);
1134 } else {
1135 list_del_init(&cap->session_caps);
1136 session->s_nr_caps--;
1137 atomic64_dec(&mdsc->metric.total_caps);
1138 cap->session = NULL;
1139 removed = 1;
1140 }
1141 /* protect backpointer with s_cap_lock: see iterate_session_caps */
1142 cap->ci = NULL;
1143
1144 /*
1145 * s_cap_reconnect is protected by s_cap_lock. no one changes
1146 * s_cap_gen while session is in the reconnect state.
1147 */
1148 if (queue_release &&
1149 (!session->s_cap_reconnect ||
1150 cap->cap_gen == atomic_read(&session->s_cap_gen))) {
1151 cap->queue_release = 1;
1152 if (removed) {
1153 __ceph_queue_cap_release(session, cap);
1154 removed = 0;
1155 }
1156 } else {
1157 cap->queue_release = 0;
1158 }
1159 cap->cap_ino = ci->i_vino.ino;
1160
1161 spin_unlock(&session->s_cap_lock);
1162
1163 if (removed)
1164 ceph_put_cap(mdsc, cap);
1165
1166 if (!__ceph_is_any_real_caps(ci)) {
1167 /* when reconnect denied, we remove session caps forcibly,
1168 * i_wr_ref can be non-zero. If there are ongoing write,
1169 * keep i_snap_realm.
1170 */
1171 if (ci->i_wr_ref == 0 && ci->i_snap_realm)
1172 ceph_change_snap_realm(&ci->netfs.inode, NULL);
1173
1174 __cap_delay_cancel(mdsc, ci);
1175 }
1176 }
1177
ceph_remove_cap(struct ceph_cap * cap,bool queue_release)1178 void ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
1179 {
1180 struct ceph_inode_info *ci = cap->ci;
1181 struct ceph_fs_client *fsc;
1182
1183 /* 'ci' being NULL means the remove have already occurred */
1184 if (!ci) {
1185 dout("%s: cap inode is NULL\n", __func__);
1186 return;
1187 }
1188
1189 lockdep_assert_held(&ci->i_ceph_lock);
1190
1191 fsc = ceph_inode_to_client(&ci->netfs.inode);
1192 WARN_ON_ONCE(ci->i_auth_cap == cap &&
1193 !list_empty(&ci->i_dirty_item) &&
1194 !fsc->blocklisted &&
1195 !ceph_inode_is_shutdown(&ci->netfs.inode));
1196
1197 __ceph_remove_cap(cap, queue_release);
1198 }
1199
1200 struct cap_msg_args {
1201 struct ceph_mds_session *session;
1202 u64 ino, cid, follows;
1203 u64 flush_tid, oldest_flush_tid, size, max_size;
1204 u64 xattr_version;
1205 u64 change_attr;
1206 struct ceph_buffer *xattr_buf;
1207 struct ceph_buffer *old_xattr_buf;
1208 struct timespec64 atime, mtime, ctime, btime;
1209 int op, caps, wanted, dirty;
1210 u32 seq, issue_seq, mseq, time_warp_seq;
1211 u32 flags;
1212 kuid_t uid;
1213 kgid_t gid;
1214 umode_t mode;
1215 bool inline_data;
1216 bool wake;
1217 };
1218
1219 /*
1220 * cap struct size + flock buffer size + inline version + inline data size +
1221 * osd_epoch_barrier + oldest_flush_tid
1222 */
1223 #define CAP_MSG_SIZE (sizeof(struct ceph_mds_caps) + \
1224 4 + 8 + 4 + 4 + 8 + 4 + 4 + 4 + 8 + 8 + 4)
1225
1226 /* Marshal up the cap msg to the MDS */
encode_cap_msg(struct ceph_msg * msg,struct cap_msg_args * arg)1227 static void encode_cap_msg(struct ceph_msg *msg, struct cap_msg_args *arg)
1228 {
1229 struct ceph_mds_caps *fc;
1230 void *p;
1231 struct ceph_osd_client *osdc = &arg->session->s_mdsc->fsc->client->osdc;
1232
1233 dout("%s %s %llx %llx caps %s wanted %s dirty %s seq %u/%u tid %llu/%llu mseq %u follows %lld size %llu/%llu xattr_ver %llu xattr_len %d\n",
1234 __func__, ceph_cap_op_name(arg->op), arg->cid, arg->ino,
1235 ceph_cap_string(arg->caps), ceph_cap_string(arg->wanted),
1236 ceph_cap_string(arg->dirty), arg->seq, arg->issue_seq,
1237 arg->flush_tid, arg->oldest_flush_tid, arg->mseq, arg->follows,
1238 arg->size, arg->max_size, arg->xattr_version,
1239 arg->xattr_buf ? (int)arg->xattr_buf->vec.iov_len : 0);
1240
1241 msg->hdr.version = cpu_to_le16(10);
1242 msg->hdr.tid = cpu_to_le64(arg->flush_tid);
1243
1244 fc = msg->front.iov_base;
1245 memset(fc, 0, sizeof(*fc));
1246
1247 fc->cap_id = cpu_to_le64(arg->cid);
1248 fc->op = cpu_to_le32(arg->op);
1249 fc->seq = cpu_to_le32(arg->seq);
1250 fc->issue_seq = cpu_to_le32(arg->issue_seq);
1251 fc->migrate_seq = cpu_to_le32(arg->mseq);
1252 fc->caps = cpu_to_le32(arg->caps);
1253 fc->wanted = cpu_to_le32(arg->wanted);
1254 fc->dirty = cpu_to_le32(arg->dirty);
1255 fc->ino = cpu_to_le64(arg->ino);
1256 fc->snap_follows = cpu_to_le64(arg->follows);
1257
1258 fc->size = cpu_to_le64(arg->size);
1259 fc->max_size = cpu_to_le64(arg->max_size);
1260 ceph_encode_timespec64(&fc->mtime, &arg->mtime);
1261 ceph_encode_timespec64(&fc->atime, &arg->atime);
1262 ceph_encode_timespec64(&fc->ctime, &arg->ctime);
1263 fc->time_warp_seq = cpu_to_le32(arg->time_warp_seq);
1264
1265 fc->uid = cpu_to_le32(from_kuid(&init_user_ns, arg->uid));
1266 fc->gid = cpu_to_le32(from_kgid(&init_user_ns, arg->gid));
1267 fc->mode = cpu_to_le32(arg->mode);
1268
1269 fc->xattr_version = cpu_to_le64(arg->xattr_version);
1270 if (arg->xattr_buf) {
1271 msg->middle = ceph_buffer_get(arg->xattr_buf);
1272 fc->xattr_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
1273 msg->hdr.middle_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
1274 }
1275
1276 p = fc + 1;
1277 /* flock buffer size (version 2) */
1278 ceph_encode_32(&p, 0);
1279 /* inline version (version 4) */
1280 ceph_encode_64(&p, arg->inline_data ? 0 : CEPH_INLINE_NONE);
1281 /* inline data size */
1282 ceph_encode_32(&p, 0);
1283 /*
1284 * osd_epoch_barrier (version 5)
1285 * The epoch_barrier is protected osdc->lock, so READ_ONCE here in
1286 * case it was recently changed
1287 */
1288 ceph_encode_32(&p, READ_ONCE(osdc->epoch_barrier));
1289 /* oldest_flush_tid (version 6) */
1290 ceph_encode_64(&p, arg->oldest_flush_tid);
1291
1292 /*
1293 * caller_uid/caller_gid (version 7)
1294 *
1295 * Currently, we don't properly track which caller dirtied the caps
1296 * last, and force a flush of them when there is a conflict. For now,
1297 * just set this to 0:0, to emulate how the MDS has worked up to now.
1298 */
1299 ceph_encode_32(&p, 0);
1300 ceph_encode_32(&p, 0);
1301
1302 /* pool namespace (version 8) (mds always ignores this) */
1303 ceph_encode_32(&p, 0);
1304
1305 /* btime and change_attr (version 9) */
1306 ceph_encode_timespec64(p, &arg->btime);
1307 p += sizeof(struct ceph_timespec);
1308 ceph_encode_64(&p, arg->change_attr);
1309
1310 /* Advisory flags (version 10) */
1311 ceph_encode_32(&p, arg->flags);
1312 }
1313
1314 /*
1315 * Queue cap releases when an inode is dropped from our cache.
1316 */
__ceph_remove_caps(struct ceph_inode_info * ci)1317 void __ceph_remove_caps(struct ceph_inode_info *ci)
1318 {
1319 struct rb_node *p;
1320
1321 /* lock i_ceph_lock, because ceph_d_revalidate(..., LOOKUP_RCU)
1322 * may call __ceph_caps_issued_mask() on a freeing inode. */
1323 spin_lock(&ci->i_ceph_lock);
1324 p = rb_first(&ci->i_caps);
1325 while (p) {
1326 struct ceph_cap *cap = rb_entry(p, struct ceph_cap, ci_node);
1327 p = rb_next(p);
1328 ceph_remove_cap(cap, true);
1329 }
1330 spin_unlock(&ci->i_ceph_lock);
1331 }
1332
1333 /*
1334 * Prepare to send a cap message to an MDS. Update the cap state, and populate
1335 * the arg struct with the parameters that will need to be sent. This should
1336 * be done under the i_ceph_lock to guard against changes to cap state.
1337 *
1338 * Make note of max_size reported/requested from mds, revoked caps
1339 * that have now been implemented.
1340 */
__prep_cap(struct cap_msg_args * arg,struct ceph_cap * cap,int op,int flags,int used,int want,int retain,int flushing,u64 flush_tid,u64 oldest_flush_tid)1341 static void __prep_cap(struct cap_msg_args *arg, struct ceph_cap *cap,
1342 int op, int flags, int used, int want, int retain,
1343 int flushing, u64 flush_tid, u64 oldest_flush_tid)
1344 {
1345 struct ceph_inode_info *ci = cap->ci;
1346 struct inode *inode = &ci->netfs.inode;
1347 int held, revoking;
1348
1349 lockdep_assert_held(&ci->i_ceph_lock);
1350
1351 held = cap->issued | cap->implemented;
1352 revoking = cap->implemented & ~cap->issued;
1353 retain &= ~revoking;
1354
1355 dout("%s %p cap %p session %p %s -> %s (revoking %s)\n",
1356 __func__, inode, cap, cap->session,
1357 ceph_cap_string(held), ceph_cap_string(held & retain),
1358 ceph_cap_string(revoking));
1359 BUG_ON((retain & CEPH_CAP_PIN) == 0);
1360
1361 ci->i_ceph_flags &= ~CEPH_I_FLUSH;
1362
1363 cap->issued &= retain; /* drop bits we don't want */
1364 /*
1365 * Wake up any waiters on wanted -> needed transition. This is due to
1366 * the weird transition from buffered to sync IO... we need to flush
1367 * dirty pages _before_ allowing sync writes to avoid reordering.
1368 */
1369 arg->wake = cap->implemented & ~cap->issued;
1370 cap->implemented &= cap->issued | used;
1371 cap->mds_wanted = want;
1372
1373 arg->session = cap->session;
1374 arg->ino = ceph_vino(inode).ino;
1375 arg->cid = cap->cap_id;
1376 arg->follows = flushing ? ci->i_head_snapc->seq : 0;
1377 arg->flush_tid = flush_tid;
1378 arg->oldest_flush_tid = oldest_flush_tid;
1379
1380 arg->size = i_size_read(inode);
1381 ci->i_reported_size = arg->size;
1382 arg->max_size = ci->i_wanted_max_size;
1383 if (cap == ci->i_auth_cap) {
1384 if (want & CEPH_CAP_ANY_FILE_WR)
1385 ci->i_requested_max_size = arg->max_size;
1386 else
1387 ci->i_requested_max_size = 0;
1388 }
1389
1390 if (flushing & CEPH_CAP_XATTR_EXCL) {
1391 arg->old_xattr_buf = __ceph_build_xattrs_blob(ci);
1392 arg->xattr_version = ci->i_xattrs.version;
1393 arg->xattr_buf = ci->i_xattrs.blob;
1394 } else {
1395 arg->xattr_buf = NULL;
1396 arg->old_xattr_buf = NULL;
1397 }
1398
1399 arg->mtime = inode->i_mtime;
1400 arg->atime = inode->i_atime;
1401 arg->ctime = inode->i_ctime;
1402 arg->btime = ci->i_btime;
1403 arg->change_attr = inode_peek_iversion_raw(inode);
1404
1405 arg->op = op;
1406 arg->caps = cap->implemented;
1407 arg->wanted = want;
1408 arg->dirty = flushing;
1409
1410 arg->seq = cap->seq;
1411 arg->issue_seq = cap->issue_seq;
1412 arg->mseq = cap->mseq;
1413 arg->time_warp_seq = ci->i_time_warp_seq;
1414
1415 arg->uid = inode->i_uid;
1416 arg->gid = inode->i_gid;
1417 arg->mode = inode->i_mode;
1418
1419 arg->inline_data = ci->i_inline_version != CEPH_INLINE_NONE;
1420 if (!(flags & CEPH_CLIENT_CAPS_PENDING_CAPSNAP) &&
1421 !list_empty(&ci->i_cap_snaps)) {
1422 struct ceph_cap_snap *capsnap;
1423 list_for_each_entry_reverse(capsnap, &ci->i_cap_snaps, ci_item) {
1424 if (capsnap->cap_flush.tid)
1425 break;
1426 if (capsnap->need_flush) {
1427 flags |= CEPH_CLIENT_CAPS_PENDING_CAPSNAP;
1428 break;
1429 }
1430 }
1431 }
1432 arg->flags = flags;
1433 }
1434
1435 /*
1436 * Send a cap msg on the given inode.
1437 *
1438 * Caller should hold snap_rwsem (read), s_mutex.
1439 */
__send_cap(struct cap_msg_args * arg,struct ceph_inode_info * ci)1440 static void __send_cap(struct cap_msg_args *arg, struct ceph_inode_info *ci)
1441 {
1442 struct ceph_msg *msg;
1443 struct inode *inode = &ci->netfs.inode;
1444
1445 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, CAP_MSG_SIZE, GFP_NOFS, false);
1446 if (!msg) {
1447 pr_err("error allocating cap msg: ino (%llx.%llx) flushing %s tid %llu, requeuing cap.\n",
1448 ceph_vinop(inode), ceph_cap_string(arg->dirty),
1449 arg->flush_tid);
1450 spin_lock(&ci->i_ceph_lock);
1451 __cap_delay_requeue(arg->session->s_mdsc, ci);
1452 spin_unlock(&ci->i_ceph_lock);
1453 return;
1454 }
1455
1456 encode_cap_msg(msg, arg);
1457 ceph_con_send(&arg->session->s_con, msg);
1458 ceph_buffer_put(arg->old_xattr_buf);
1459 if (arg->wake)
1460 wake_up_all(&ci->i_cap_wq);
1461 }
1462
__send_flush_snap(struct inode * inode,struct ceph_mds_session * session,struct ceph_cap_snap * capsnap,u32 mseq,u64 oldest_flush_tid)1463 static inline int __send_flush_snap(struct inode *inode,
1464 struct ceph_mds_session *session,
1465 struct ceph_cap_snap *capsnap,
1466 u32 mseq, u64 oldest_flush_tid)
1467 {
1468 struct cap_msg_args arg;
1469 struct ceph_msg *msg;
1470
1471 msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, CAP_MSG_SIZE, GFP_NOFS, false);
1472 if (!msg)
1473 return -ENOMEM;
1474
1475 arg.session = session;
1476 arg.ino = ceph_vino(inode).ino;
1477 arg.cid = 0;
1478 arg.follows = capsnap->follows;
1479 arg.flush_tid = capsnap->cap_flush.tid;
1480 arg.oldest_flush_tid = oldest_flush_tid;
1481
1482 arg.size = capsnap->size;
1483 arg.max_size = 0;
1484 arg.xattr_version = capsnap->xattr_version;
1485 arg.xattr_buf = capsnap->xattr_blob;
1486 arg.old_xattr_buf = NULL;
1487
1488 arg.atime = capsnap->atime;
1489 arg.mtime = capsnap->mtime;
1490 arg.ctime = capsnap->ctime;
1491 arg.btime = capsnap->btime;
1492 arg.change_attr = capsnap->change_attr;
1493
1494 arg.op = CEPH_CAP_OP_FLUSHSNAP;
1495 arg.caps = capsnap->issued;
1496 arg.wanted = 0;
1497 arg.dirty = capsnap->dirty;
1498
1499 arg.seq = 0;
1500 arg.issue_seq = 0;
1501 arg.mseq = mseq;
1502 arg.time_warp_seq = capsnap->time_warp_seq;
1503
1504 arg.uid = capsnap->uid;
1505 arg.gid = capsnap->gid;
1506 arg.mode = capsnap->mode;
1507
1508 arg.inline_data = capsnap->inline_data;
1509 arg.flags = 0;
1510 arg.wake = false;
1511
1512 encode_cap_msg(msg, &arg);
1513 ceph_con_send(&arg.session->s_con, msg);
1514 return 0;
1515 }
1516
1517 /*
1518 * When a snapshot is taken, clients accumulate dirty metadata on
1519 * inodes with capabilities in ceph_cap_snaps to describe the file
1520 * state at the time the snapshot was taken. This must be flushed
1521 * asynchronously back to the MDS once sync writes complete and dirty
1522 * data is written out.
1523 *
1524 * Called under i_ceph_lock.
1525 */
__ceph_flush_snaps(struct ceph_inode_info * ci,struct ceph_mds_session * session)1526 static void __ceph_flush_snaps(struct ceph_inode_info *ci,
1527 struct ceph_mds_session *session)
1528 __releases(ci->i_ceph_lock)
1529 __acquires(ci->i_ceph_lock)
1530 {
1531 struct inode *inode = &ci->netfs.inode;
1532 struct ceph_mds_client *mdsc = session->s_mdsc;
1533 struct ceph_cap_snap *capsnap;
1534 u64 oldest_flush_tid = 0;
1535 u64 first_tid = 1, last_tid = 0;
1536
1537 dout("__flush_snaps %p session %p\n", inode, session);
1538
1539 list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
1540 /*
1541 * we need to wait for sync writes to complete and for dirty
1542 * pages to be written out.
1543 */
1544 if (capsnap->dirty_pages || capsnap->writing)
1545 break;
1546
1547 /* should be removed by ceph_try_drop_cap_snap() */
1548 BUG_ON(!capsnap->need_flush);
1549
1550 /* only flush each capsnap once */
1551 if (capsnap->cap_flush.tid > 0) {
1552 dout(" already flushed %p, skipping\n", capsnap);
1553 continue;
1554 }
1555
1556 spin_lock(&mdsc->cap_dirty_lock);
1557 capsnap->cap_flush.tid = ++mdsc->last_cap_flush_tid;
1558 list_add_tail(&capsnap->cap_flush.g_list,
1559 &mdsc->cap_flush_list);
1560 if (oldest_flush_tid == 0)
1561 oldest_flush_tid = __get_oldest_flush_tid(mdsc);
1562 if (list_empty(&ci->i_flushing_item)) {
1563 list_add_tail(&ci->i_flushing_item,
1564 &session->s_cap_flushing);
1565 }
1566 spin_unlock(&mdsc->cap_dirty_lock);
1567
1568 list_add_tail(&capsnap->cap_flush.i_list,
1569 &ci->i_cap_flush_list);
1570
1571 if (first_tid == 1)
1572 first_tid = capsnap->cap_flush.tid;
1573 last_tid = capsnap->cap_flush.tid;
1574 }
1575
1576 ci->i_ceph_flags &= ~CEPH_I_FLUSH_SNAPS;
1577
1578 while (first_tid <= last_tid) {
1579 struct ceph_cap *cap = ci->i_auth_cap;
1580 struct ceph_cap_flush *cf = NULL, *iter;
1581 int ret;
1582
1583 if (!(cap && cap->session == session)) {
1584 dout("__flush_snaps %p auth cap %p not mds%d, "
1585 "stop\n", inode, cap, session->s_mds);
1586 break;
1587 }
1588
1589 ret = -ENOENT;
1590 list_for_each_entry(iter, &ci->i_cap_flush_list, i_list) {
1591 if (iter->tid >= first_tid) {
1592 cf = iter;
1593 ret = 0;
1594 break;
1595 }
1596 }
1597 if (ret < 0)
1598 break;
1599
1600 first_tid = cf->tid + 1;
1601
1602 capsnap = container_of(cf, struct ceph_cap_snap, cap_flush);
1603 refcount_inc(&capsnap->nref);
1604 spin_unlock(&ci->i_ceph_lock);
1605
1606 dout("__flush_snaps %p capsnap %p tid %llu %s\n",
1607 inode, capsnap, cf->tid, ceph_cap_string(capsnap->dirty));
1608
1609 ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
1610 oldest_flush_tid);
1611 if (ret < 0) {
1612 pr_err("__flush_snaps: error sending cap flushsnap, "
1613 "ino (%llx.%llx) tid %llu follows %llu\n",
1614 ceph_vinop(inode), cf->tid, capsnap->follows);
1615 }
1616
1617 ceph_put_cap_snap(capsnap);
1618 spin_lock(&ci->i_ceph_lock);
1619 }
1620 }
1621
ceph_flush_snaps(struct ceph_inode_info * ci,struct ceph_mds_session ** psession)1622 void ceph_flush_snaps(struct ceph_inode_info *ci,
1623 struct ceph_mds_session **psession)
1624 {
1625 struct inode *inode = &ci->netfs.inode;
1626 struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
1627 struct ceph_mds_session *session = NULL;
1628 int mds;
1629
1630 dout("ceph_flush_snaps %p\n", inode);
1631 if (psession)
1632 session = *psession;
1633 retry:
1634 spin_lock(&ci->i_ceph_lock);
1635 if (!(ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)) {
1636 dout(" no capsnap needs flush, doing nothing\n");
1637 goto out;
1638 }
1639 if (!ci->i_auth_cap) {
1640 dout(" no auth cap (migrating?), doing nothing\n");
1641 goto out;
1642 }
1643
1644 mds = ci->i_auth_cap->session->s_mds;
1645 if (session && session->s_mds != mds) {
1646 dout(" oops, wrong session %p mutex\n", session);
1647 ceph_put_mds_session(session);
1648 session = NULL;
1649 }
1650 if (!session) {
1651 spin_unlock(&ci->i_ceph_lock);
1652 mutex_lock(&mdsc->mutex);
1653 session = __ceph_lookup_mds_session(mdsc, mds);
1654 mutex_unlock(&mdsc->mutex);
1655 goto retry;
1656 }
1657
1658 // make sure flushsnap messages are sent in proper order.
1659 if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
1660 __kick_flushing_caps(mdsc, session, ci, 0);
1661
1662 __ceph_flush_snaps(ci, session);
1663 out:
1664 spin_unlock(&ci->i_ceph_lock);
1665
1666 if (psession)
1667 *psession = session;
1668 else
1669 ceph_put_mds_session(session);
1670 /* we flushed them all; remove this inode from the queue */
1671 spin_lock(&mdsc->snap_flush_lock);
1672 list_del_init(&ci->i_snap_flush_item);
1673 spin_unlock(&mdsc->snap_flush_lock);
1674 }
1675
1676 /*
1677 * Mark caps dirty. If inode is newly dirty, return the dirty flags.
1678 * Caller is then responsible for calling __mark_inode_dirty with the
1679 * returned flags value.
1680 */
__ceph_mark_dirty_caps(struct ceph_inode_info * ci,int mask,struct ceph_cap_flush ** pcf)1681 int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
1682 struct ceph_cap_flush **pcf)
1683 {
1684 struct ceph_mds_client *mdsc =
1685 ceph_sb_to_client(ci->netfs.inode.i_sb)->mdsc;
1686 struct inode *inode = &ci->netfs.inode;
1687 int was = ci->i_dirty_caps;
1688 int dirty = 0;
1689
1690 lockdep_assert_held(&ci->i_ceph_lock);
1691
1692 if (!ci->i_auth_cap) {
1693 pr_warn("__mark_dirty_caps %p %llx mask %s, "
1694 "but no auth cap (session was closed?)\n",
1695 inode, ceph_ino(inode), ceph_cap_string(mask));
1696 return 0;
1697 }
1698
1699 dout("__mark_dirty_caps %p %s dirty %s -> %s\n", &ci->netfs.inode,
1700 ceph_cap_string(mask), ceph_cap_string(was),
1701 ceph_cap_string(was | mask));
1702 ci->i_dirty_caps |= mask;
1703 if (was == 0) {
1704 struct ceph_mds_session *session = ci->i_auth_cap->session;
1705
1706 WARN_ON_ONCE(ci->i_prealloc_cap_flush);
1707 swap(ci->i_prealloc_cap_flush, *pcf);
1708
1709 if (!ci->i_head_snapc) {
1710 WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem));
1711 ci->i_head_snapc = ceph_get_snap_context(
1712 ci->i_snap_realm->cached_context);
1713 }
1714 dout(" inode %p now dirty snapc %p auth cap %p\n",
1715 &ci->netfs.inode, ci->i_head_snapc, ci->i_auth_cap);
1716 BUG_ON(!list_empty(&ci->i_dirty_item));
1717 spin_lock(&mdsc->cap_dirty_lock);
1718 list_add(&ci->i_dirty_item, &session->s_cap_dirty);
1719 spin_unlock(&mdsc->cap_dirty_lock);
1720 if (ci->i_flushing_caps == 0) {
1721 ihold(inode);
1722 dirty |= I_DIRTY_SYNC;
1723 }
1724 } else {
1725 WARN_ON_ONCE(!ci->i_prealloc_cap_flush);
1726 }
1727 BUG_ON(list_empty(&ci->i_dirty_item));
1728 if (((was | ci->i_flushing_caps) & CEPH_CAP_FILE_BUFFER) &&
1729 (mask & CEPH_CAP_FILE_BUFFER))
1730 dirty |= I_DIRTY_DATASYNC;
1731 __cap_delay_requeue(mdsc, ci);
1732 return dirty;
1733 }
1734
ceph_alloc_cap_flush(void)1735 struct ceph_cap_flush *ceph_alloc_cap_flush(void)
1736 {
1737 struct ceph_cap_flush *cf;
1738
1739 cf = kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
1740 if (!cf)
1741 return NULL;
1742
1743 cf->is_capsnap = false;
1744 return cf;
1745 }
1746
ceph_free_cap_flush(struct ceph_cap_flush * cf)1747 void ceph_free_cap_flush(struct ceph_cap_flush *cf)
1748 {
1749 if (cf)
1750 kmem_cache_free(ceph_cap_flush_cachep, cf);
1751 }
1752
__get_oldest_flush_tid(struct ceph_mds_client * mdsc)1753 static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc)
1754 {
1755 if (!list_empty(&mdsc->cap_flush_list)) {
1756 struct ceph_cap_flush *cf =
1757 list_first_entry(&mdsc->cap_flush_list,
1758 struct ceph_cap_flush, g_list);
1759 return cf->tid;
1760 }
1761 return 0;
1762 }
1763
1764 /*
1765 * Remove cap_flush from the mdsc's or inode's flushing cap list.
1766 * Return true if caller needs to wake up flush waiters.
1767 */
__detach_cap_flush_from_mdsc(struct ceph_mds_client * mdsc,struct ceph_cap_flush * cf)1768 static bool __detach_cap_flush_from_mdsc(struct ceph_mds_client *mdsc,
1769 struct ceph_cap_flush *cf)
1770 {
1771 struct ceph_cap_flush *prev;
1772 bool wake = cf->wake;
1773
1774 if (wake && cf->g_list.prev != &mdsc->cap_flush_list) {
1775 prev = list_prev_entry(cf, g_list);
1776 prev->wake = true;
1777 wake = false;
1778 }
1779 list_del_init(&cf->g_list);
1780 return wake;
1781 }
1782
__detach_cap_flush_from_ci(struct ceph_inode_info * ci,struct ceph_cap_flush * cf)1783 static bool __detach_cap_flush_from_ci(struct ceph_inode_info *ci,
1784 struct ceph_cap_flush *cf)
1785 {
1786 struct ceph_cap_flush *prev;
1787 bool wake = cf->wake;
1788
1789 if (wake && cf->i_list.prev != &ci->i_cap_flush_list) {
1790 prev = list_prev_entry(cf, i_list);
1791 prev->wake = true;
1792 wake = false;
1793 }
1794 list_del_init(&cf->i_list);
1795 return wake;
1796 }
1797
1798 /*
1799 * Add dirty inode to the flushing list. Assigned a seq number so we
1800 * can wait for caps to flush without starving.
1801 *
1802 * Called under i_ceph_lock. Returns the flush tid.
1803 */
__mark_caps_flushing(struct inode * inode,struct ceph_mds_session * session,bool wake,u64 * oldest_flush_tid)1804 static u64 __mark_caps_flushing(struct inode *inode,
1805 struct ceph_mds_session *session, bool wake,
1806 u64 *oldest_flush_tid)
1807 {
1808 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
1809 struct ceph_inode_info *ci = ceph_inode(inode);
1810 struct ceph_cap_flush *cf = NULL;
1811 int flushing;
1812
1813 lockdep_assert_held(&ci->i_ceph_lock);
1814 BUG_ON(ci->i_dirty_caps == 0);
1815 BUG_ON(list_empty(&ci->i_dirty_item));
1816 BUG_ON(!ci->i_prealloc_cap_flush);
1817
1818 flushing = ci->i_dirty_caps;
1819 dout("__mark_caps_flushing flushing %s, flushing_caps %s -> %s\n",
1820 ceph_cap_string(flushing),
1821 ceph_cap_string(ci->i_flushing_caps),
1822 ceph_cap_string(ci->i_flushing_caps | flushing));
1823 ci->i_flushing_caps |= flushing;
1824 ci->i_dirty_caps = 0;
1825 dout(" inode %p now !dirty\n", inode);
1826
1827 swap(cf, ci->i_prealloc_cap_flush);
1828 cf->caps = flushing;
1829 cf->wake = wake;
1830
1831 spin_lock(&mdsc->cap_dirty_lock);
1832 list_del_init(&ci->i_dirty_item);
1833
1834 cf->tid = ++mdsc->last_cap_flush_tid;
1835 list_add_tail(&cf->g_list, &mdsc->cap_flush_list);
1836 *oldest_flush_tid = __get_oldest_flush_tid(mdsc);
1837
1838 if (list_empty(&ci->i_flushing_item)) {
1839 list_add_tail(&ci->i_flushing_item, &session->s_cap_flushing);
1840 mdsc->num_cap_flushing++;
1841 }
1842 spin_unlock(&mdsc->cap_dirty_lock);
1843
1844 list_add_tail(&cf->i_list, &ci->i_cap_flush_list);
1845
1846 return cf->tid;
1847 }
1848
1849 /*
1850 * try to invalidate mapping pages without blocking.
1851 */
try_nonblocking_invalidate(struct inode * inode)1852 static int try_nonblocking_invalidate(struct inode *inode)
1853 __releases(ci->i_ceph_lock)
1854 __acquires(ci->i_ceph_lock)
1855 {
1856 struct ceph_inode_info *ci = ceph_inode(inode);
1857 u32 invalidating_gen = ci->i_rdcache_gen;
1858
1859 spin_unlock(&ci->i_ceph_lock);
1860 ceph_fscache_invalidate(inode, false);
1861 invalidate_mapping_pages(&inode->i_data, 0, -1);
1862 spin_lock(&ci->i_ceph_lock);
1863
1864 if (inode->i_data.nrpages == 0 &&
1865 invalidating_gen == ci->i_rdcache_gen) {
1866 /* success. */
1867 dout("try_nonblocking_invalidate %p success\n", inode);
1868 /* save any racing async invalidate some trouble */
1869 ci->i_rdcache_revoking = ci->i_rdcache_gen - 1;
1870 return 0;
1871 }
1872 dout("try_nonblocking_invalidate %p failed\n", inode);
1873 return -1;
1874 }
1875
__ceph_should_report_size(struct ceph_inode_info * ci)1876 bool __ceph_should_report_size(struct ceph_inode_info *ci)
1877 {
1878 loff_t size = i_size_read(&ci->netfs.inode);
1879 /* mds will adjust max size according to the reported size */
1880 if (ci->i_flushing_caps & CEPH_CAP_FILE_WR)
1881 return false;
1882 if (size >= ci->i_max_size)
1883 return true;
1884 /* half of previous max_size increment has been used */
1885 if (ci->i_max_size > ci->i_reported_size &&
1886 (size << 1) >= ci->i_max_size + ci->i_reported_size)
1887 return true;
1888 return false;
1889 }
1890
1891 /*
1892 * Swiss army knife function to examine currently used and wanted
1893 * versus held caps. Release, flush, ack revoked caps to mds as
1894 * appropriate.
1895 *
1896 * CHECK_CAPS_AUTHONLY - we should only check the auth cap
1897 * CHECK_CAPS_FLUSH - we should flush any dirty caps immediately, without
1898 * further delay.
1899 */
ceph_check_caps(struct ceph_inode_info * ci,int flags,struct ceph_mds_session * session)1900 void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1901 struct ceph_mds_session *session)
1902 {
1903 struct inode *inode = &ci->netfs.inode;
1904 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
1905 struct ceph_cap *cap;
1906 u64 flush_tid, oldest_flush_tid;
1907 int file_wanted, used, cap_used;
1908 int issued, implemented, want, retain, revoking, flushing = 0;
1909 int mds = -1; /* keep track of how far we've gone through i_caps list
1910 to avoid an infinite loop on retry */
1911 struct rb_node *p;
1912 bool queue_invalidate = false;
1913 bool tried_invalidate = false;
1914 bool queue_writeback = false;
1915
1916 if (session)
1917 ceph_get_mds_session(session);
1918
1919 spin_lock(&ci->i_ceph_lock);
1920 if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE) {
1921 /* Don't send messages until we get async create reply */
1922 spin_unlock(&ci->i_ceph_lock);
1923 ceph_put_mds_session(session);
1924 return;
1925 }
1926
1927 if (ci->i_ceph_flags & CEPH_I_FLUSH)
1928 flags |= CHECK_CAPS_FLUSH;
1929 retry:
1930 /* Caps wanted by virtue of active open files. */
1931 file_wanted = __ceph_caps_file_wanted(ci);
1932
1933 /* Caps which have active references against them */
1934 used = __ceph_caps_used(ci);
1935
1936 /*
1937 * "issued" represents the current caps that the MDS wants us to have.
1938 * "implemented" is the set that we have been granted, and includes the
1939 * ones that have not yet been returned to the MDS (the "revoking" set,
1940 * usually because they have outstanding references).
1941 */
1942 issued = __ceph_caps_issued(ci, &implemented);
1943 revoking = implemented & ~issued;
1944
1945 want = file_wanted;
1946
1947 /* The ones we currently want to retain (may be adjusted below) */
1948 retain = file_wanted | used | CEPH_CAP_PIN;
1949 if (!mdsc->stopping && inode->i_nlink > 0) {
1950 if (file_wanted) {
1951 retain |= CEPH_CAP_ANY; /* be greedy */
1952 } else if (S_ISDIR(inode->i_mode) &&
1953 (issued & CEPH_CAP_FILE_SHARED) &&
1954 __ceph_dir_is_complete(ci)) {
1955 /*
1956 * If a directory is complete, we want to keep
1957 * the exclusive cap. So that MDS does not end up
1958 * revoking the shared cap on every create/unlink
1959 * operation.
1960 */
1961 if (IS_RDONLY(inode)) {
1962 want = CEPH_CAP_ANY_SHARED;
1963 } else {
1964 want |= CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_EXCL;
1965 }
1966 retain |= want;
1967 } else {
1968
1969 retain |= CEPH_CAP_ANY_SHARED;
1970 /*
1971 * keep RD only if we didn't have the file open RW,
1972 * because then the mds would revoke it anyway to
1973 * journal max_size=0.
1974 */
1975 if (ci->i_max_size == 0)
1976 retain |= CEPH_CAP_ANY_RD;
1977 }
1978 }
1979
1980 dout("check_caps %llx.%llx file_want %s used %s dirty %s flushing %s"
1981 " issued %s revoking %s retain %s %s%s\n", ceph_vinop(inode),
1982 ceph_cap_string(file_wanted),
1983 ceph_cap_string(used), ceph_cap_string(ci->i_dirty_caps),
1984 ceph_cap_string(ci->i_flushing_caps),
1985 ceph_cap_string(issued), ceph_cap_string(revoking),
1986 ceph_cap_string(retain),
1987 (flags & CHECK_CAPS_AUTHONLY) ? " AUTHONLY" : "",
1988 (flags & CHECK_CAPS_FLUSH) ? " FLUSH" : "");
1989
1990 /*
1991 * If we no longer need to hold onto old our caps, and we may
1992 * have cached pages, but don't want them, then try to invalidate.
1993 * If we fail, it's because pages are locked.... try again later.
1994 */
1995 if ((!(flags & CHECK_CAPS_NOINVAL) || mdsc->stopping) &&
1996 S_ISREG(inode->i_mode) &&
1997 !(ci->i_wb_ref || ci->i_wrbuffer_ref) && /* no dirty pages... */
1998 inode->i_data.nrpages && /* have cached pages */
1999 (revoking & (CEPH_CAP_FILE_CACHE|
2000 CEPH_CAP_FILE_LAZYIO)) && /* or revoking cache */
2001 !tried_invalidate) {
2002 dout("check_caps trying to invalidate on %llx.%llx\n",
2003 ceph_vinop(inode));
2004 if (try_nonblocking_invalidate(inode) < 0) {
2005 dout("check_caps queuing invalidate\n");
2006 queue_invalidate = true;
2007 ci->i_rdcache_revoking = ci->i_rdcache_gen;
2008 }
2009 tried_invalidate = true;
2010 goto retry;
2011 }
2012
2013 for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
2014 int mflags = 0;
2015 struct cap_msg_args arg;
2016
2017 cap = rb_entry(p, struct ceph_cap, ci_node);
2018
2019 /* avoid looping forever */
2020 if (mds >= cap->mds ||
2021 ((flags & CHECK_CAPS_AUTHONLY) && cap != ci->i_auth_cap))
2022 continue;
2023
2024 /*
2025 * If we have an auth cap, we don't need to consider any
2026 * overlapping caps as used.
2027 */
2028 cap_used = used;
2029 if (ci->i_auth_cap && cap != ci->i_auth_cap)
2030 cap_used &= ~ci->i_auth_cap->issued;
2031
2032 revoking = cap->implemented & ~cap->issued;
2033 dout(" mds%d cap %p used %s issued %s implemented %s revoking %s\n",
2034 cap->mds, cap, ceph_cap_string(cap_used),
2035 ceph_cap_string(cap->issued),
2036 ceph_cap_string(cap->implemented),
2037 ceph_cap_string(revoking));
2038
2039 if (cap == ci->i_auth_cap &&
2040 (cap->issued & CEPH_CAP_FILE_WR)) {
2041 /* request larger max_size from MDS? */
2042 if (ci->i_wanted_max_size > ci->i_max_size &&
2043 ci->i_wanted_max_size > ci->i_requested_max_size) {
2044 dout("requesting new max_size\n");
2045 goto ack;
2046 }
2047
2048 /* approaching file_max? */
2049 if (__ceph_should_report_size(ci)) {
2050 dout("i_size approaching max_size\n");
2051 goto ack;
2052 }
2053 }
2054 /* flush anything dirty? */
2055 if (cap == ci->i_auth_cap) {
2056 if ((flags & CHECK_CAPS_FLUSH) && ci->i_dirty_caps) {
2057 dout("flushing dirty caps\n");
2058 goto ack;
2059 }
2060 if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS) {
2061 dout("flushing snap caps\n");
2062 goto ack;
2063 }
2064 }
2065
2066 /* completed revocation? going down and there are no caps? */
2067 if (revoking) {
2068 if ((revoking & cap_used) == 0) {
2069 dout("completed revocation of %s\n",
2070 ceph_cap_string(cap->implemented & ~cap->issued));
2071 goto ack;
2072 }
2073
2074 /*
2075 * If the "i_wrbuffer_ref" was increased by mmap or generic
2076 * cache write just before the ceph_check_caps() is called,
2077 * the Fb capability revoking will fail this time. Then we
2078 * must wait for the BDI's delayed work to flush the dirty
2079 * pages and to release the "i_wrbuffer_ref", which will cost
2080 * at most 5 seconds. That means the MDS needs to wait at
2081 * most 5 seconds to finished the Fb capability's revocation.
2082 *
2083 * Let's queue a writeback for it.
2084 */
2085 if (S_ISREG(inode->i_mode) && ci->i_wrbuffer_ref &&
2086 (revoking & CEPH_CAP_FILE_BUFFER))
2087 queue_writeback = true;
2088 }
2089
2090 /* want more caps from mds? */
2091 if (want & ~cap->mds_wanted) {
2092 if (want & ~(cap->mds_wanted | cap->issued))
2093 goto ack;
2094 if (!__cap_is_valid(cap))
2095 goto ack;
2096 }
2097
2098 /* things we might delay */
2099 if ((cap->issued & ~retain) == 0)
2100 continue; /* nope, all good */
2101
2102 ack:
2103 ceph_put_mds_session(session);
2104 session = ceph_get_mds_session(cap->session);
2105
2106 /* kick flushing and flush snaps before sending normal
2107 * cap message */
2108 if (cap == ci->i_auth_cap &&
2109 (ci->i_ceph_flags &
2110 (CEPH_I_KICK_FLUSH | CEPH_I_FLUSH_SNAPS))) {
2111 if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
2112 __kick_flushing_caps(mdsc, session, ci, 0);
2113 if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
2114 __ceph_flush_snaps(ci, session);
2115
2116 goto retry;
2117 }
2118
2119 if (cap == ci->i_auth_cap && ci->i_dirty_caps) {
2120 flushing = ci->i_dirty_caps;
2121 flush_tid = __mark_caps_flushing(inode, session, false,
2122 &oldest_flush_tid);
2123 if (flags & CHECK_CAPS_FLUSH &&
2124 list_empty(&session->s_cap_dirty))
2125 mflags |= CEPH_CLIENT_CAPS_SYNC;
2126 } else {
2127 flushing = 0;
2128 flush_tid = 0;
2129 spin_lock(&mdsc->cap_dirty_lock);
2130 oldest_flush_tid = __get_oldest_flush_tid(mdsc);
2131 spin_unlock(&mdsc->cap_dirty_lock);
2132 }
2133
2134 mds = cap->mds; /* remember mds, so we don't repeat */
2135
2136 __prep_cap(&arg, cap, CEPH_CAP_OP_UPDATE, mflags, cap_used,
2137 want, retain, flushing, flush_tid, oldest_flush_tid);
2138
2139 spin_unlock(&ci->i_ceph_lock);
2140 __send_cap(&arg, ci);
2141 spin_lock(&ci->i_ceph_lock);
2142
2143 goto retry; /* retake i_ceph_lock and restart our cap scan. */
2144 }
2145
2146 /* periodically re-calculate caps wanted by open files */
2147 if (__ceph_is_any_real_caps(ci) &&
2148 list_empty(&ci->i_cap_delay_list) &&
2149 (file_wanted & ~CEPH_CAP_PIN) &&
2150 !(used & (CEPH_CAP_FILE_RD | CEPH_CAP_ANY_FILE_WR))) {
2151 __cap_delay_requeue(mdsc, ci);
2152 }
2153
2154 spin_unlock(&ci->i_ceph_lock);
2155
2156 ceph_put_mds_session(session);
2157 if (queue_writeback)
2158 ceph_queue_writeback(inode);
2159 if (queue_invalidate)
2160 ceph_queue_invalidate(inode);
2161 }
2162
2163 /*
2164 * Try to flush dirty caps back to the auth mds.
2165 */
try_flush_caps(struct inode * inode,u64 * ptid)2166 static int try_flush_caps(struct inode *inode, u64 *ptid)
2167 {
2168 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
2169 struct ceph_inode_info *ci = ceph_inode(inode);
2170 int flushing = 0;
2171 u64 flush_tid = 0, oldest_flush_tid = 0;
2172
2173 spin_lock(&ci->i_ceph_lock);
2174 retry_locked:
2175 if (ci->i_dirty_caps && ci->i_auth_cap) {
2176 struct ceph_cap *cap = ci->i_auth_cap;
2177 struct cap_msg_args arg;
2178 struct ceph_mds_session *session = cap->session;
2179
2180 if (session->s_state < CEPH_MDS_SESSION_OPEN) {
2181 spin_unlock(&ci->i_ceph_lock);
2182 goto out;
2183 }
2184
2185 if (ci->i_ceph_flags &
2186 (CEPH_I_KICK_FLUSH | CEPH_I_FLUSH_SNAPS)) {
2187 if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
2188 __kick_flushing_caps(mdsc, session, ci, 0);
2189 if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
2190 __ceph_flush_snaps(ci, session);
2191 goto retry_locked;
2192 }
2193
2194 flushing = ci->i_dirty_caps;
2195 flush_tid = __mark_caps_flushing(inode, session, true,
2196 &oldest_flush_tid);
2197
2198 __prep_cap(&arg, cap, CEPH_CAP_OP_FLUSH, CEPH_CLIENT_CAPS_SYNC,
2199 __ceph_caps_used(ci), __ceph_caps_wanted(ci),
2200 (cap->issued | cap->implemented),
2201 flushing, flush_tid, oldest_flush_tid);
2202 spin_unlock(&ci->i_ceph_lock);
2203
2204 __send_cap(&arg, ci);
2205 } else {
2206 if (!list_empty(&ci->i_cap_flush_list)) {
2207 struct ceph_cap_flush *cf =
2208 list_last_entry(&ci->i_cap_flush_list,
2209 struct ceph_cap_flush, i_list);
2210 cf->wake = true;
2211 flush_tid = cf->tid;
2212 }
2213 flushing = ci->i_flushing_caps;
2214 spin_unlock(&ci->i_ceph_lock);
2215 }
2216 out:
2217 *ptid = flush_tid;
2218 return flushing;
2219 }
2220
2221 /*
2222 * Return true if we've flushed caps through the given flush_tid.
2223 */
caps_are_flushed(struct inode * inode,u64 flush_tid)2224 static int caps_are_flushed(struct inode *inode, u64 flush_tid)
2225 {
2226 struct ceph_inode_info *ci = ceph_inode(inode);
2227 int ret = 1;
2228
2229 spin_lock(&ci->i_ceph_lock);
2230 if (!list_empty(&ci->i_cap_flush_list)) {
2231 struct ceph_cap_flush * cf =
2232 list_first_entry(&ci->i_cap_flush_list,
2233 struct ceph_cap_flush, i_list);
2234 if (cf->tid <= flush_tid)
2235 ret = 0;
2236 }
2237 spin_unlock(&ci->i_ceph_lock);
2238 return ret;
2239 }
2240
2241 /*
2242 * flush the mdlog and wait for any unsafe requests to complete.
2243 */
flush_mdlog_and_wait_inode_unsafe_requests(struct inode * inode)2244 static int flush_mdlog_and_wait_inode_unsafe_requests(struct inode *inode)
2245 {
2246 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
2247 struct ceph_inode_info *ci = ceph_inode(inode);
2248 struct ceph_mds_request *req1 = NULL, *req2 = NULL;
2249 unsigned int max_sessions;
2250 int ret, err = 0;
2251
2252 spin_lock(&ci->i_unsafe_lock);
2253 if (S_ISDIR(inode->i_mode) && !list_empty(&ci->i_unsafe_dirops)) {
2254 req1 = list_last_entry(&ci->i_unsafe_dirops,
2255 struct ceph_mds_request,
2256 r_unsafe_dir_item);
2257 ceph_mdsc_get_request(req1);
2258 }
2259 if (!list_empty(&ci->i_unsafe_iops)) {
2260 req2 = list_last_entry(&ci->i_unsafe_iops,
2261 struct ceph_mds_request,
2262 r_unsafe_target_item);
2263 ceph_mdsc_get_request(req2);
2264 }
2265 spin_unlock(&ci->i_unsafe_lock);
2266
2267 /*
2268 * The mdsc->max_sessions is unlikely to be changed
2269 * mostly, here we will retry it by reallocating the
2270 * sessions array memory to get rid of the mdsc->mutex
2271 * lock.
2272 */
2273 retry:
2274 max_sessions = mdsc->max_sessions;
2275
2276 /*
2277 * Trigger to flush the journal logs in all the relevant MDSes
2278 * manually, or in the worst case we must wait at most 5 seconds
2279 * to wait the journal logs to be flushed by the MDSes periodically.
2280 */
2281 if ((req1 || req2) && likely(max_sessions)) {
2282 struct ceph_mds_session **sessions = NULL;
2283 struct ceph_mds_session *s;
2284 struct ceph_mds_request *req;
2285 int i;
2286
2287 sessions = kzalloc(max_sessions * sizeof(s), GFP_KERNEL);
2288 if (!sessions) {
2289 err = -ENOMEM;
2290 goto out;
2291 }
2292
2293 spin_lock(&ci->i_unsafe_lock);
2294 if (req1) {
2295 list_for_each_entry(req, &ci->i_unsafe_dirops,
2296 r_unsafe_dir_item) {
2297 s = req->r_session;
2298 if (!s)
2299 continue;
2300 if (unlikely(s->s_mds >= max_sessions)) {
2301 spin_unlock(&ci->i_unsafe_lock);
2302 for (i = 0; i < max_sessions; i++) {
2303 s = sessions[i];
2304 if (s)
2305 ceph_put_mds_session(s);
2306 }
2307 kfree(sessions);
2308 goto retry;
2309 }
2310 if (!sessions[s->s_mds]) {
2311 s = ceph_get_mds_session(s);
2312 sessions[s->s_mds] = s;
2313 }
2314 }
2315 }
2316 if (req2) {
2317 list_for_each_entry(req, &ci->i_unsafe_iops,
2318 r_unsafe_target_item) {
2319 s = req->r_session;
2320 if (!s)
2321 continue;
2322 if (unlikely(s->s_mds >= max_sessions)) {
2323 spin_unlock(&ci->i_unsafe_lock);
2324 for (i = 0; i < max_sessions; i++) {
2325 s = sessions[i];
2326 if (s)
2327 ceph_put_mds_session(s);
2328 }
2329 kfree(sessions);
2330 goto retry;
2331 }
2332 if (!sessions[s->s_mds]) {
2333 s = ceph_get_mds_session(s);
2334 sessions[s->s_mds] = s;
2335 }
2336 }
2337 }
2338 spin_unlock(&ci->i_unsafe_lock);
2339
2340 /* the auth MDS */
2341 spin_lock(&ci->i_ceph_lock);
2342 if (ci->i_auth_cap) {
2343 s = ci->i_auth_cap->session;
2344 if (!sessions[s->s_mds])
2345 sessions[s->s_mds] = ceph_get_mds_session(s);
2346 }
2347 spin_unlock(&ci->i_ceph_lock);
2348
2349 /* send flush mdlog request to MDSes */
2350 for (i = 0; i < max_sessions; i++) {
2351 s = sessions[i];
2352 if (s) {
2353 send_flush_mdlog(s);
2354 ceph_put_mds_session(s);
2355 }
2356 }
2357 kfree(sessions);
2358 }
2359
2360 dout("%s %p wait on tid %llu %llu\n", __func__,
2361 inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
2362 if (req1) {
2363 ret = !wait_for_completion_timeout(&req1->r_safe_completion,
2364 ceph_timeout_jiffies(req1->r_timeout));
2365 if (ret)
2366 err = -EIO;
2367 }
2368 if (req2) {
2369 ret = !wait_for_completion_timeout(&req2->r_safe_completion,
2370 ceph_timeout_jiffies(req2->r_timeout));
2371 if (ret)
2372 err = -EIO;
2373 }
2374
2375 out:
2376 if (req1)
2377 ceph_mdsc_put_request(req1);
2378 if (req2)
2379 ceph_mdsc_put_request(req2);
2380 return err;
2381 }
2382
ceph_fsync(struct file * file,loff_t start,loff_t end,int datasync)2383 int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync)
2384 {
2385 struct inode *inode = file->f_mapping->host;
2386 struct ceph_inode_info *ci = ceph_inode(inode);
2387 u64 flush_tid;
2388 int ret, err;
2389 int dirty;
2390
2391 dout("fsync %p%s\n", inode, datasync ? " datasync" : "");
2392
2393 ret = file_write_and_wait_range(file, start, end);
2394 if (datasync)
2395 goto out;
2396
2397 ret = ceph_wait_on_async_create(inode);
2398 if (ret)
2399 goto out;
2400
2401 dirty = try_flush_caps(inode, &flush_tid);
2402 dout("fsync dirty caps are %s\n", ceph_cap_string(dirty));
2403
2404 err = flush_mdlog_and_wait_inode_unsafe_requests(inode);
2405
2406 /*
2407 * only wait on non-file metadata writeback (the mds
2408 * can recover size and mtime, so we don't need to
2409 * wait for that)
2410 */
2411 if (!err && (dirty & ~CEPH_CAP_ANY_FILE_WR)) {
2412 err = wait_event_interruptible(ci->i_cap_wq,
2413 caps_are_flushed(inode, flush_tid));
2414 }
2415
2416 if (err < 0)
2417 ret = err;
2418
2419 err = file_check_and_advance_wb_err(file);
2420 if (err < 0)
2421 ret = err;
2422 out:
2423 dout("fsync %p%s result=%d\n", inode, datasync ? " datasync" : "", ret);
2424 return ret;
2425 }
2426
2427 /*
2428 * Flush any dirty caps back to the mds. If we aren't asked to wait,
2429 * queue inode for flush but don't do so immediately, because we can
2430 * get by with fewer MDS messages if we wait for data writeback to
2431 * complete first.
2432 */
ceph_write_inode(struct inode * inode,struct writeback_control * wbc)2433 int ceph_write_inode(struct inode *inode, struct writeback_control *wbc)
2434 {
2435 struct ceph_inode_info *ci = ceph_inode(inode);
2436 u64 flush_tid;
2437 int err = 0;
2438 int dirty;
2439 int wait = (wbc->sync_mode == WB_SYNC_ALL && !wbc->for_sync);
2440
2441 dout("write_inode %p wait=%d\n", inode, wait);
2442 ceph_fscache_unpin_writeback(inode, wbc);
2443 if (wait) {
2444 err = ceph_wait_on_async_create(inode);
2445 if (err)
2446 return err;
2447 dirty = try_flush_caps(inode, &flush_tid);
2448 if (dirty)
2449 err = wait_event_interruptible(ci->i_cap_wq,
2450 caps_are_flushed(inode, flush_tid));
2451 } else {
2452 struct ceph_mds_client *mdsc =
2453 ceph_sb_to_client(inode->i_sb)->mdsc;
2454
2455 spin_lock(&ci->i_ceph_lock);
2456 if (__ceph_caps_dirty(ci))
2457 __cap_delay_requeue_front(mdsc, ci);
2458 spin_unlock(&ci->i_ceph_lock);
2459 }
2460 return err;
2461 }
2462
__kick_flushing_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session,struct ceph_inode_info * ci,u64 oldest_flush_tid)2463 static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
2464 struct ceph_mds_session *session,
2465 struct ceph_inode_info *ci,
2466 u64 oldest_flush_tid)
2467 __releases(ci->i_ceph_lock)
2468 __acquires(ci->i_ceph_lock)
2469 {
2470 struct inode *inode = &ci->netfs.inode;
2471 struct ceph_cap *cap;
2472 struct ceph_cap_flush *cf;
2473 int ret;
2474 u64 first_tid = 0;
2475 u64 last_snap_flush = 0;
2476
2477 /* Don't do anything until create reply comes in */
2478 if (ci->i_ceph_flags & CEPH_I_ASYNC_CREATE)
2479 return;
2480
2481 ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
2482
2483 list_for_each_entry_reverse(cf, &ci->i_cap_flush_list, i_list) {
2484 if (cf->is_capsnap) {
2485 last_snap_flush = cf->tid;
2486 break;
2487 }
2488 }
2489
2490 list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
2491 if (cf->tid < first_tid)
2492 continue;
2493
2494 cap = ci->i_auth_cap;
2495 if (!(cap && cap->session == session)) {
2496 pr_err("%p auth cap %p not mds%d ???\n",
2497 inode, cap, session->s_mds);
2498 break;
2499 }
2500
2501 first_tid = cf->tid + 1;
2502
2503 if (!cf->is_capsnap) {
2504 struct cap_msg_args arg;
2505
2506 dout("kick_flushing_caps %p cap %p tid %llu %s\n",
2507 inode, cap, cf->tid, ceph_cap_string(cf->caps));
2508 __prep_cap(&arg, cap, CEPH_CAP_OP_FLUSH,
2509 (cf->tid < last_snap_flush ?
2510 CEPH_CLIENT_CAPS_PENDING_CAPSNAP : 0),
2511 __ceph_caps_used(ci),
2512 __ceph_caps_wanted(ci),
2513 (cap->issued | cap->implemented),
2514 cf->caps, cf->tid, oldest_flush_tid);
2515 spin_unlock(&ci->i_ceph_lock);
2516 __send_cap(&arg, ci);
2517 } else {
2518 struct ceph_cap_snap *capsnap =
2519 container_of(cf, struct ceph_cap_snap,
2520 cap_flush);
2521 dout("kick_flushing_caps %p capsnap %p tid %llu %s\n",
2522 inode, capsnap, cf->tid,
2523 ceph_cap_string(capsnap->dirty));
2524
2525 refcount_inc(&capsnap->nref);
2526 spin_unlock(&ci->i_ceph_lock);
2527
2528 ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
2529 oldest_flush_tid);
2530 if (ret < 0) {
2531 pr_err("kick_flushing_caps: error sending "
2532 "cap flushsnap, ino (%llx.%llx) "
2533 "tid %llu follows %llu\n",
2534 ceph_vinop(inode), cf->tid,
2535 capsnap->follows);
2536 }
2537
2538 ceph_put_cap_snap(capsnap);
2539 }
2540
2541 spin_lock(&ci->i_ceph_lock);
2542 }
2543 }
2544
ceph_early_kick_flushing_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)2545 void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
2546 struct ceph_mds_session *session)
2547 {
2548 struct ceph_inode_info *ci;
2549 struct ceph_cap *cap;
2550 u64 oldest_flush_tid;
2551
2552 dout("early_kick_flushing_caps mds%d\n", session->s_mds);
2553
2554 spin_lock(&mdsc->cap_dirty_lock);
2555 oldest_flush_tid = __get_oldest_flush_tid(mdsc);
2556 spin_unlock(&mdsc->cap_dirty_lock);
2557
2558 list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
2559 spin_lock(&ci->i_ceph_lock);
2560 cap = ci->i_auth_cap;
2561 if (!(cap && cap->session == session)) {
2562 pr_err("%p auth cap %p not mds%d ???\n",
2563 &ci->netfs.inode, cap, session->s_mds);
2564 spin_unlock(&ci->i_ceph_lock);
2565 continue;
2566 }
2567
2568
2569 /*
2570 * if flushing caps were revoked, we re-send the cap flush
2571 * in client reconnect stage. This guarantees MDS * processes
2572 * the cap flush message before issuing the flushing caps to
2573 * other client.
2574 */
2575 if ((cap->issued & ci->i_flushing_caps) !=
2576 ci->i_flushing_caps) {
2577 /* encode_caps_cb() also will reset these sequence
2578 * numbers. make sure sequence numbers in cap flush
2579 * message match later reconnect message */
2580 cap->seq = 0;
2581 cap->issue_seq = 0;
2582 cap->mseq = 0;
2583 __kick_flushing_caps(mdsc, session, ci,
2584 oldest_flush_tid);
2585 } else {
2586 ci->i_ceph_flags |= CEPH_I_KICK_FLUSH;
2587 }
2588
2589 spin_unlock(&ci->i_ceph_lock);
2590 }
2591 }
2592
ceph_kick_flushing_caps(struct ceph_mds_client * mdsc,struct ceph_mds_session * session)2593 void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
2594 struct ceph_mds_session *session)
2595 {
2596 struct ceph_inode_info *ci;
2597 struct ceph_cap *cap;
2598 u64 oldest_flush_tid;
2599
2600 lockdep_assert_held(&session->s_mutex);
2601
2602 dout("kick_flushing_caps mds%d\n", session->s_mds);
2603
2604 spin_lock(&mdsc->cap_dirty_lock);
2605 oldest_flush_tid = __get_oldest_flush_tid(mdsc);
2606 spin_unlock(&mdsc->cap_dirty_lock);
2607
2608 list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
2609 spin_lock(&ci->i_ceph_lock);
2610 cap = ci->i_auth_cap;
2611 if (!(cap && cap->session == session)) {
2612 pr_err("%p auth cap %p not mds%d ???\n",
2613 &ci->netfs.inode, cap, session->s_mds);
2614 spin_unlock(&ci->i_ceph_lock);
2615 continue;
2616 }
2617 if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH) {
2618 __kick_flushing_caps(mdsc, session, ci,
2619 oldest_flush_tid);
2620 }
2621 spin_unlock(&ci->i_ceph_lock);
2622 }
2623 }
2624
ceph_kick_flushing_inode_caps(struct ceph_mds_session * session,struct ceph_inode_info * ci)2625 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
2626 struct ceph_inode_info *ci)
2627 {
2628 struct ceph_mds_client *mdsc = session->s_mdsc;
2629 struct ceph_cap *cap = ci->i_auth_cap;
2630
2631 lockdep_assert_held(&ci->i_ceph_lock);
2632
2633 dout("%s %p flushing %s\n", __func__, &ci->netfs.inode,
2634 ceph_cap_string(ci->i_flushing_caps));
2635
2636 if (!list_empty(&ci->i_cap_flush_list)) {
2637 u64 oldest_flush_tid;
2638 spin_lock(&mdsc->cap_dirty_lock);
2639 list_move_tail(&ci->i_flushing_item,
2640 &cap->session->s_cap_flushing);
2641 oldest_flush_tid = __get_oldest_flush_tid(mdsc);
2642 spin_unlock(&mdsc->cap_dirty_lock);
2643
2644 __kick_flushing_caps(mdsc, session, ci, oldest_flush_tid);
2645 }
2646 }
2647
2648
2649 /*
2650 * Take references to capabilities we hold, so that we don't release
2651 * them to the MDS prematurely.
2652 */
ceph_take_cap_refs(struct ceph_inode_info * ci,int got,bool snap_rwsem_locked)2653 void ceph_take_cap_refs(struct ceph_inode_info *ci, int got,
2654 bool snap_rwsem_locked)
2655 {
2656 lockdep_assert_held(&ci->i_ceph_lock);
2657
2658 if (got & CEPH_CAP_PIN)
2659 ci->i_pin_ref++;
2660 if (got & CEPH_CAP_FILE_RD)
2661 ci->i_rd_ref++;
2662 if (got & CEPH_CAP_FILE_CACHE)
2663 ci->i_rdcache_ref++;
2664 if (got & CEPH_CAP_FILE_EXCL)
2665 ci->i_fx_ref++;
2666 if (got & CEPH_CAP_FILE_WR) {
2667 if (ci->i_wr_ref == 0 && !ci->i_head_snapc) {
2668 BUG_ON(!snap_rwsem_locked);
2669 ci->i_head_snapc = ceph_get_snap_context(
2670 ci->i_snap_realm->cached_context);
2671 }
2672 ci->i_wr_ref++;
2673 }
2674 if (got & CEPH_CAP_FILE_BUFFER) {
2675 if (ci->i_wb_ref == 0)
2676 ihold(&ci->netfs.inode);
2677 ci->i_wb_ref++;
2678 dout("%s %p wb %d -> %d (?)\n", __func__,
2679 &ci->netfs.inode, ci->i_wb_ref-1, ci->i_wb_ref);
2680 }
2681 }
2682
2683 /*
2684 * Try to grab cap references. Specify those refs we @want, and the
2685 * minimal set we @need. Also include the larger offset we are writing
2686 * to (when applicable), and check against max_size here as well.
2687 * Note that caller is responsible for ensuring max_size increases are
2688 * requested from the MDS.
2689 *
2690 * Returns 0 if caps were not able to be acquired (yet), 1 if succeed,
2691 * or a negative error code. There are 3 speical error codes:
2692 * -EAGAIN: need to sleep but non-blocking is specified
2693 * -EFBIG: ask caller to call check_max_size() and try again.
2694 * -EUCLEAN: ask caller to call ceph_renew_caps() and try again.
2695 */
2696 enum {
2697 /* first 8 bits are reserved for CEPH_FILE_MODE_FOO */
2698 NON_BLOCKING = (1 << 8),
2699 CHECK_FILELOCK = (1 << 9),
2700 };
2701
try_get_cap_refs(struct inode * inode,int need,int want,loff_t endoff,int flags,int * got)2702 static int try_get_cap_refs(struct inode *inode, int need, int want,
2703 loff_t endoff, int flags, int *got)
2704 {
2705 struct ceph_inode_info *ci = ceph_inode(inode);
2706 struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
2707 int ret = 0;
2708 int have, implemented;
2709 bool snap_rwsem_locked = false;
2710
2711 dout("get_cap_refs %p need %s want %s\n", inode,
2712 ceph_cap_string(need), ceph_cap_string(want));
2713
2714 again:
2715 spin_lock(&ci->i_ceph_lock);
2716
2717 if ((flags & CHECK_FILELOCK) &&
2718 (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK)) {
2719 dout("try_get_cap_refs %p error filelock\n", inode);
2720 ret = -EIO;
2721 goto out_unlock;
2722 }
2723
2724 /* finish pending truncate */
2725 while (ci->i_truncate_pending) {
2726 spin_unlock(&ci->i_ceph_lock);
2727 if (snap_rwsem_locked) {
2728 up_read(&mdsc->snap_rwsem);
2729 snap_rwsem_locked = false;
2730 }
2731 __ceph_do_pending_vmtruncate(inode);
2732 spin_lock(&ci->i_ceph_lock);
2733 }
2734
2735 have = __ceph_caps_issued(ci, &implemented);
2736
2737 if (have & need & CEPH_CAP_FILE_WR) {
2738 if (endoff >= 0 && endoff > (loff_t)ci->i_max_size) {
2739 dout("get_cap_refs %p endoff %llu > maxsize %llu\n",
2740 inode, endoff, ci->i_max_size);
2741 if (endoff > ci->i_requested_max_size)
2742 ret = ci->i_auth_cap ? -EFBIG : -EUCLEAN;
2743 goto out_unlock;
2744 }
2745 /*
2746 * If a sync write is in progress, we must wait, so that we
2747 * can get a final snapshot value for size+mtime.
2748 */
2749 if (__ceph_have_pending_cap_snap(ci)) {
2750 dout("get_cap_refs %p cap_snap_pending\n", inode);
2751 goto out_unlock;
2752 }
2753 }
2754
2755 if ((have & need) == need) {
2756 /*
2757 * Look at (implemented & ~have & not) so that we keep waiting
2758 * on transition from wanted -> needed caps. This is needed
2759 * for WRBUFFER|WR -> WR to avoid a new WR sync write from
2760 * going before a prior buffered writeback happens.
2761 */
2762 int not = want & ~(have & need);
2763 int revoking = implemented & ~have;
2764 dout("get_cap_refs %p have %s but not %s (revoking %s)\n",
2765 inode, ceph_cap_string(have), ceph_cap_string(not),
2766 ceph_cap_string(revoking));
2767 if ((revoking & not) == 0) {
2768 if (!snap_rwsem_locked &&
2769 !ci->i_head_snapc &&
2770 (need & CEPH_CAP_FILE_WR)) {
2771 if (!down_read_trylock(&mdsc->snap_rwsem)) {
2772 /*
2773 * we can not call down_read() when
2774 * task isn't in TASK_RUNNING state
2775 */
2776 if (flags & NON_BLOCKING) {
2777 ret = -EAGAIN;
2778 goto out_unlock;
2779 }
2780
2781 spin_unlock(&ci->i_ceph_lock);
2782 down_read(&mdsc->snap_rwsem);
2783 snap_rwsem_locked = true;
2784 goto again;
2785 }
2786 snap_rwsem_locked = true;
2787 }
2788 if ((have & want) == want)
2789 *got = need | want;
2790 else
2791 *got = need;
2792 ceph_take_cap_refs(ci, *got, true);
2793 ret = 1;
2794 }
2795 } else {
2796 int session_readonly = false;
2797 int mds_wanted;
2798 if (ci->i_auth_cap &&
2799 (need & (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_EXCL))) {
2800 struct ceph_mds_session *s = ci->i_auth_cap->session;
2801 spin_lock(&s->s_cap_lock);
2802 session_readonly = s->s_readonly;
2803 spin_unlock(&s->s_cap_lock);
2804 }
2805 if (session_readonly) {
2806 dout("get_cap_refs %p need %s but mds%d readonly\n",
2807 inode, ceph_cap_string(need), ci->i_auth_cap->mds);
2808 ret = -EROFS;
2809 goto out_unlock;
2810 }
2811
2812 if (ceph_inode_is_shutdown(inode)) {
2813 dout("get_cap_refs %p inode is shutdown\n", inode);
2814 ret = -ESTALE;
2815 goto out_unlock;
2816 }
2817 mds_wanted = __ceph_caps_mds_wanted(ci, false);
2818 if (need & ~mds_wanted) {
2819 dout("get_cap_refs %p need %s > mds_wanted %s\n",
2820 inode, ceph_cap_string(need),
2821 ceph_cap_string(mds_wanted));
2822 ret = -EUCLEAN;
2823 goto out_unlock;
2824 }
2825
2826 dout("get_cap_refs %p have %s need %s\n", inode,
2827 ceph_cap_string(have), ceph_cap_string(need));
2828 }
2829 out_unlock:
2830
2831 __ceph_touch_fmode(ci, mdsc, flags);
2832
2833 spin_unlock(&ci->i_ceph_lock);
2834 if (snap_rwsem_locked)
2835 up_read(&mdsc->snap_rwsem);
2836
2837 if (!ret)
2838 ceph_update_cap_mis(&mdsc->metric);
2839 else if (ret == 1)
2840 ceph_update_cap_hit(&mdsc->metric);
2841
2842 dout("get_cap_refs %p ret %d got %s\n", inode,
2843 ret, ceph_cap_string(*got));
2844 return ret;
2845 }
2846
2847 /*
2848 * Check the offset we are writing up to against our current
2849 * max_size. If necessary, tell the MDS we want to write to
2850 * a larger offset.
2851 */
check_max_size(struct inode * inode,loff_t endoff)2852 static void check_max_size(struct inode *inode, loff_t endoff)
2853 {
2854 struct ceph_inode_info *ci = ceph_inode(inode);
2855 int check = 0;
2856
2857 /* do we need to explicitly request a larger max_size? */
2858 spin_lock(&ci->i_ceph_lock);
2859 if (endoff >= ci->i_max_size && endoff > ci->i_wanted_max_size) {
2860 dout("write %p at large endoff %llu, req max_size\n",
2861 inode, endoff);
2862 ci->i_wanted_max_size = endoff;
2863 }
2864 /* duplicate ceph_check_caps()'s logic */
2865 if (ci->i_auth_cap &&
2866 (ci->i_auth_cap->issued & CEPH_CAP_FILE_WR) &&
2867 ci->i_wanted_max_size > ci->i_max_size &&
2868 ci->i_wanted_max_size > ci->i_requested_max_size)
2869 check = 1;
2870 spin_unlock(&ci->i_ceph_lock);
2871 if (check)
2872 ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
2873 }
2874
get_used_fmode(int caps)2875 static inline int get_used_fmode(int caps)
2876 {
2877 int fmode = 0;
2878 if (caps & CEPH_CAP_FILE_RD)
2879 fmode |= CEPH_FILE_MODE_RD;
2880 if (caps & CEPH_CAP_FILE_WR)
2881 fmode |= CEPH_FILE_MODE_WR;
2882 return fmode;
2883 }
2884
ceph_try_get_caps(struct inode * inode,int need,int want,bool nonblock,int * got)2885 int ceph_try_get_caps(struct inode *inode, int need, int want,
2886 bool nonblock, int *got)
2887 {
2888 int ret, flags;
2889
2890 BUG_ON(need & ~CEPH_CAP_FILE_RD);
2891 BUG_ON(want & ~(CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO |
2892 CEPH_CAP_FILE_SHARED | CEPH_CAP_FILE_EXCL |
2893 CEPH_CAP_ANY_DIR_OPS));
2894 if (need) {
2895 ret = ceph_pool_perm_check(inode, need);
2896 if (ret < 0)
2897 return ret;
2898 }
2899
2900 flags = get_used_fmode(need | want);
2901 if (nonblock)
2902 flags |= NON_BLOCKING;
2903
2904 ret = try_get_cap_refs(inode, need, want, 0, flags, got);
2905 /* three special error codes */
2906 if (ret == -EAGAIN || ret == -EFBIG || ret == -EUCLEAN)
2907 ret = 0;
2908 return ret;
2909 }
2910
2911 /*
2912 * Wait for caps, and take cap references. If we can't get a WR cap
2913 * due to a small max_size, make sure we check_max_size (and possibly
2914 * ask the mds) so we don't get hung up indefinitely.
2915 */
ceph_get_caps(struct file * filp,int need,int want,loff_t endoff,int * got)2916 int ceph_get_caps(struct file *filp, int need, int want, loff_t endoff, int *got)
2917 {
2918 struct ceph_file_info *fi = filp->private_data;
2919 struct inode *inode = file_inode(filp);
2920 struct ceph_inode_info *ci = ceph_inode(inode);
2921 struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
2922 int ret, _got, flags;
2923
2924 ret = ceph_pool_perm_check(inode, need);
2925 if (ret < 0)
2926 return ret;
2927
2928 if ((fi->fmode & CEPH_FILE_MODE_WR) &&
2929 fi->filp_gen != READ_ONCE(fsc->filp_gen))
2930 return -EBADF;
2931
2932 flags = get_used_fmode(need | want);
2933
2934 while (true) {
2935 flags &= CEPH_FILE_MODE_MASK;
2936 if (atomic_read(&fi->num_locks))
2937 flags |= CHECK_FILELOCK;
2938 _got = 0;
2939 ret = try_get_cap_refs(inode, need, want, endoff,
2940 flags, &_got);
2941 WARN_ON_ONCE(ret == -EAGAIN);
2942 if (!ret) {
2943 struct ceph_mds_client *mdsc = fsc->mdsc;
2944 struct cap_wait cw;
2945 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2946
2947 cw.ino = ceph_ino(inode);
2948 cw.tgid = current->tgid;
2949 cw.need = need;
2950 cw.want = want;
2951
2952 spin_lock(&mdsc->caps_list_lock);
2953 list_add(&cw.list, &mdsc->cap_wait_list);
2954 spin_unlock(&mdsc->caps_list_lock);
2955
2956 /* make sure used fmode not timeout */
2957 ceph_get_fmode(ci, flags, FMODE_WAIT_BIAS);
2958 add_wait_queue(&ci->i_cap_wq, &wait);
2959
2960 flags |= NON_BLOCKING;
2961 while (!(ret = try_get_cap_refs(inode, need, want,
2962 endoff, flags, &_got))) {
2963 if (signal_pending(current)) {
2964 ret = -ERESTARTSYS;
2965 break;
2966 }
2967 wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
2968 }
2969
2970 remove_wait_queue(&ci->i_cap_wq, &wait);
2971 ceph_put_fmode(ci, flags, FMODE_WAIT_BIAS);
2972
2973 spin_lock(&mdsc->caps_list_lock);
2974 list_del(&cw.list);
2975 spin_unlock(&mdsc->caps_list_lock);
2976
2977 if (ret == -EAGAIN)
2978 continue;
2979 }
2980
2981 if ((fi->fmode & CEPH_FILE_MODE_WR) &&
2982 fi->filp_gen != READ_ONCE(fsc->filp_gen)) {
2983 if (ret >= 0 && _got)
2984 ceph_put_cap_refs(ci, _got);
2985 return -EBADF;
2986 }
2987
2988 if (ret < 0) {
2989 if (ret == -EFBIG || ret == -EUCLEAN) {
2990 int ret2 = ceph_wait_on_async_create(inode);
2991 if (ret2 < 0)
2992 return ret2;
2993 }
2994 if (ret == -EFBIG) {
2995 check_max_size(inode, endoff);
2996 continue;
2997 }
2998 if (ret == -EUCLEAN) {
2999 /* session was killed, try renew caps */
3000 ret = ceph_renew_caps(inode, flags);
3001 if (ret == 0)
3002 continue;
3003 }
3004 return ret;
3005 }
3006
3007 if (S_ISREG(ci->netfs.inode.i_mode) &&
3008 ci->i_inline_version != CEPH_INLINE_NONE &&
3009 (_got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
3010 i_size_read(inode) > 0) {
3011 struct page *page =
3012 find_get_page(inode->i_mapping, 0);
3013 if (page) {
3014 bool uptodate = PageUptodate(page);
3015
3016 put_page(page);
3017 if (uptodate)
3018 break;
3019 }
3020 /*
3021 * drop cap refs first because getattr while
3022 * holding * caps refs can cause deadlock.
3023 */
3024 ceph_put_cap_refs(ci, _got);
3025 _got = 0;
3026
3027 /*
3028 * getattr request will bring inline data into
3029 * page cache
3030 */
3031 ret = __ceph_do_getattr(inode, NULL,
3032 CEPH_STAT_CAP_INLINE_DATA,
3033 true);
3034 if (ret < 0)
3035 return ret;
3036 continue;
3037 }
3038 break;
3039 }
3040 *got = _got;
3041 return 0;
3042 }
3043
3044 /*
3045 * Take cap refs. Caller must already know we hold at least one ref
3046 * on the caps in question or we don't know this is safe.
3047 */
ceph_get_cap_refs(struct ceph_inode_info * ci,int caps)3048 void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps)
3049 {
3050 spin_lock(&ci->i_ceph_lock);
3051 ceph_take_cap_refs(ci, caps, false);
3052 spin_unlock(&ci->i_ceph_lock);
3053 }
3054
3055
3056 /*
3057 * drop cap_snap that is not associated with any snapshot.
3058 * we don't need to send FLUSHSNAP message for it.
3059 */
ceph_try_drop_cap_snap(struct ceph_inode_info * ci,struct ceph_cap_snap * capsnap)3060 static int ceph_try_drop_cap_snap(struct ceph_inode_info *ci,
3061 struct ceph_cap_snap *capsnap)
3062 {
3063 if (!capsnap->need_flush &&
3064 !capsnap->writing && !capsnap->dirty_pages) {
3065 dout("dropping cap_snap %p follows %llu\n",
3066 capsnap, capsnap->follows);
3067 BUG_ON(capsnap->cap_flush.tid > 0);
3068 ceph_put_snap_context(capsnap->context);
3069 if (!list_is_last(&capsnap->ci_item, &ci->i_cap_snaps))
3070 ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;
3071
3072 list_del(&capsnap->ci_item);
3073 ceph_put_cap_snap(capsnap);
3074 return 1;
3075 }
3076 return 0;
3077 }
3078
3079 enum put_cap_refs_mode {
3080 PUT_CAP_REFS_SYNC = 0,
3081 PUT_CAP_REFS_NO_CHECK,
3082 PUT_CAP_REFS_ASYNC,
3083 };
3084
3085 /*
3086 * Release cap refs.
3087 *
3088 * If we released the last ref on any given cap, call ceph_check_caps
3089 * to release (or schedule a release).
3090 *
3091 * If we are releasing a WR cap (from a sync write), finalize any affected
3092 * cap_snap, and wake up any waiters.
3093 */
__ceph_put_cap_refs(struct ceph_inode_info * ci,int had,enum put_cap_refs_mode mode)3094 static void __ceph_put_cap_refs(struct ceph_inode_info *ci, int had,
3095 enum put_cap_refs_mode mode)
3096 {
3097 struct inode *inode = &ci->netfs.inode;
3098 int last = 0, put = 0, flushsnaps = 0, wake = 0;
3099 bool check_flushsnaps = false;
3100
3101 spin_lock(&ci->i_ceph_lock);
3102 if (had & CEPH_CAP_PIN)
3103 --ci->i_pin_ref;
3104 if (had & CEPH_CAP_FILE_RD)
3105 if (--ci->i_rd_ref == 0)
3106 last++;
3107 if (had & CEPH_CAP_FILE_CACHE)
3108 if (--ci->i_rdcache_ref == 0)
3109 last++;
3110 if (had & CEPH_CAP_FILE_EXCL)
3111 if (--ci->i_fx_ref == 0)
3112 last++;
3113 if (had & CEPH_CAP_FILE_BUFFER) {
3114 if (--ci->i_wb_ref == 0) {
3115 last++;
3116 /* put the ref held by ceph_take_cap_refs() */
3117 put++;
3118 check_flushsnaps = true;
3119 }
3120 dout("put_cap_refs %p wb %d -> %d (?)\n",
3121 inode, ci->i_wb_ref+1, ci->i_wb_ref);
3122 }
3123 if (had & CEPH_CAP_FILE_WR) {
3124 if (--ci->i_wr_ref == 0) {
3125 last++;
3126 check_flushsnaps = true;
3127 if (ci->i_wrbuffer_ref_head == 0 &&
3128 ci->i_dirty_caps == 0 &&
3129 ci->i_flushing_caps == 0) {
3130 BUG_ON(!ci->i_head_snapc);
3131 ceph_put_snap_context(ci->i_head_snapc);
3132 ci->i_head_snapc = NULL;
3133 }
3134 /* see comment in __ceph_remove_cap() */
3135 if (!__ceph_is_any_real_caps(ci) && ci->i_snap_realm)
3136 ceph_change_snap_realm(inode, NULL);
3137 }
3138 }
3139 if (check_flushsnaps && __ceph_have_pending_cap_snap(ci)) {
3140 struct ceph_cap_snap *capsnap =
3141 list_last_entry(&ci->i_cap_snaps,
3142 struct ceph_cap_snap,
3143 ci_item);
3144
3145 capsnap->writing = 0;
3146 if (ceph_try_drop_cap_snap(ci, capsnap))
3147 /* put the ref held by ceph_queue_cap_snap() */
3148 put++;
3149 else if (__ceph_finish_cap_snap(ci, capsnap))
3150 flushsnaps = 1;
3151 wake = 1;
3152 }
3153 spin_unlock(&ci->i_ceph_lock);
3154
3155 dout("put_cap_refs %p had %s%s%s\n", inode, ceph_cap_string(had),
3156 last ? " last" : "", put ? " put" : "");
3157
3158 switch (mode) {
3159 case PUT_CAP_REFS_SYNC:
3160 if (last)
3161 ceph_check_caps(ci, 0, NULL);
3162 else if (flushsnaps)
3163 ceph_flush_snaps(ci, NULL);
3164 break;
3165 case PUT_CAP_REFS_ASYNC:
3166 if (last)
3167 ceph_queue_check_caps(inode);
3168 else if (flushsnaps)
3169 ceph_queue_flush_snaps(inode);
3170 break;
3171 default:
3172 break;
3173 }
3174 if (wake)
3175 wake_up_all(&ci->i_cap_wq);
3176 while (put-- > 0)
3177 iput(inode);
3178 }
3179
ceph_put_cap_refs(struct ceph_inode_info * ci,int had)3180 void ceph_put_cap_refs(struct ceph_inode_info *ci, int had)
3181 {
3182 __ceph_put_cap_refs(ci, had, PUT_CAP_REFS_SYNC);
3183 }
3184
ceph_put_cap_refs_async(struct ceph_inode_info * ci,int had)3185 void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had)
3186 {
3187 __ceph_put_cap_refs(ci, had, PUT_CAP_REFS_ASYNC);
3188 }
3189
ceph_put_cap_refs_no_check_caps(struct ceph_inode_info * ci,int had)3190 void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, int had)
3191 {
3192 __ceph_put_cap_refs(ci, had, PUT_CAP_REFS_NO_CHECK);
3193 }
3194
3195 /*
3196 * Release @nr WRBUFFER refs on dirty pages for the given @snapc snap
3197 * context. Adjust per-snap dirty page accounting as appropriate.
3198 * Once all dirty data for a cap_snap is flushed, flush snapped file
3199 * metadata back to the MDS. If we dropped the last ref, call
3200 * ceph_check_caps.
3201 */
ceph_put_wrbuffer_cap_refs(struct ceph_inode_info * ci,int nr,struct ceph_snap_context * snapc)3202 void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
3203 struct ceph_snap_context *snapc)
3204 {
3205 struct inode *inode = &ci->netfs.inode;
3206 struct ceph_cap_snap *capsnap = NULL, *iter;
3207 int put = 0;
3208 bool last = false;
3209 bool flush_snaps = false;
3210 bool complete_capsnap = false;
3211
3212 spin_lock(&ci->i_ceph_lock);
3213 ci->i_wrbuffer_ref -= nr;
3214 if (ci->i_wrbuffer_ref == 0) {
3215 last = true;
3216 put++;
3217 }
3218
3219 if (ci->i_head_snapc == snapc) {
3220 ci->i_wrbuffer_ref_head -= nr;
3221 if (ci->i_wrbuffer_ref_head == 0 &&
3222 ci->i_wr_ref == 0 &&
3223 ci->i_dirty_caps == 0 &&
3224 ci->i_flushing_caps == 0) {
3225 BUG_ON(!ci->i_head_snapc);
3226 ceph_put_snap_context(ci->i_head_snapc);
3227 ci->i_head_snapc = NULL;
3228 }
3229 dout("put_wrbuffer_cap_refs on %p head %d/%d -> %d/%d %s\n",
3230 inode,
3231 ci->i_wrbuffer_ref+nr, ci->i_wrbuffer_ref_head+nr,
3232 ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
3233 last ? " LAST" : "");
3234 } else {
3235 list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) {
3236 if (iter->context == snapc) {
3237 capsnap = iter;
3238 break;
3239 }
3240 }
3241
3242 if (!capsnap) {
3243 /*
3244 * The capsnap should already be removed when removing
3245 * auth cap in the case of a forced unmount.
3246 */
3247 WARN_ON_ONCE(ci->i_auth_cap);
3248 goto unlock;
3249 }
3250
3251 capsnap->dirty_pages -= nr;
3252 if (capsnap->dirty_pages == 0) {
3253 complete_capsnap = true;
3254 if (!capsnap->writing) {
3255 if (ceph_try_drop_cap_snap(ci, capsnap)) {
3256 put++;
3257 } else {
3258 ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;
3259 flush_snaps = true;
3260 }
3261 }
3262 }
3263 dout("put_wrbuffer_cap_refs on %p cap_snap %p "
3264 " snap %lld %d/%d -> %d/%d %s%s\n",
3265 inode, capsnap, capsnap->context->seq,
3266 ci->i_wrbuffer_ref+nr, capsnap->dirty_pages + nr,
3267 ci->i_wrbuffer_ref, capsnap->dirty_pages,
3268 last ? " (wrbuffer last)" : "",
3269 complete_capsnap ? " (complete capsnap)" : "");
3270 }
3271
3272 unlock:
3273 spin_unlock(&ci->i_ceph_lock);
3274
3275 if (last) {
3276 ceph_check_caps(ci, 0, NULL);
3277 } else if (flush_snaps) {
3278 ceph_flush_snaps(ci, NULL);
3279 }
3280 if (complete_capsnap)
3281 wake_up_all(&ci->i_cap_wq);
3282 while (put-- > 0) {
3283 iput(inode);
3284 }
3285 }
3286
3287 /*
3288 * Invalidate unlinked inode's aliases, so we can drop the inode ASAP.
3289 */
invalidate_aliases(struct inode * inode)3290 static void invalidate_aliases(struct inode *inode)
3291 {
3292 struct dentry *dn, *prev = NULL;
3293
3294 dout("invalidate_aliases inode %p\n", inode);
3295 d_prune_aliases(inode);
3296 /*
3297 * For non-directory inode, d_find_alias() only returns
3298 * hashed dentry. After calling d_invalidate(), the
3299 * dentry becomes unhashed.
3300 *
3301 * For directory inode, d_find_alias() can return
3302 * unhashed dentry. But directory inode should have
3303 * one alias at most.
3304 */
3305 while ((dn = d_find_alias(inode))) {
3306 if (dn == prev) {
3307 dput(dn);
3308 break;
3309 }
3310 d_invalidate(dn);
3311 if (prev)
3312 dput(prev);
3313 prev = dn;
3314 }
3315 if (prev)
3316 dput(prev);
3317 }
3318
3319 struct cap_extra_info {
3320 struct ceph_string *pool_ns;
3321 /* inline data */
3322 u64 inline_version;
3323 void *inline_data;
3324 u32 inline_len;
3325 /* dirstat */
3326 bool dirstat_valid;
3327 u64 nfiles;
3328 u64 nsubdirs;
3329 u64 change_attr;
3330 /* currently issued */
3331 int issued;
3332 struct timespec64 btime;
3333 };
3334
3335 /*
3336 * Handle a cap GRANT message from the MDS. (Note that a GRANT may
3337 * actually be a revocation if it specifies a smaller cap set.)
3338 *
3339 * caller holds s_mutex and i_ceph_lock, we drop both.
3340 */
handle_cap_grant(struct inode * inode,struct ceph_mds_session * session,struct ceph_cap * cap,struct ceph_mds_caps * grant,struct ceph_buffer * xattr_buf,struct cap_extra_info * extra_info)3341 static void handle_cap_grant(struct inode *inode,
3342 struct ceph_mds_session *session,
3343 struct ceph_cap *cap,
3344 struct ceph_mds_caps *grant,
3345 struct ceph_buffer *xattr_buf,
3346 struct cap_extra_info *extra_info)
3347 __releases(ci->i_ceph_lock)
3348 __releases(session->s_mdsc->snap_rwsem)
3349 {
3350 struct ceph_inode_info *ci = ceph_inode(inode);
3351 int seq = le32_to_cpu(grant->seq);
3352 int newcaps = le32_to_cpu(grant->caps);
3353 int used, wanted, dirty;
3354 u64 size = le64_to_cpu(grant->size);
3355 u64 max_size = le64_to_cpu(grant->max_size);
3356 unsigned char check_caps = 0;
3357 bool was_stale = cap->cap_gen < atomic_read(&session->s_cap_gen);
3358 bool wake = false;
3359 bool writeback = false;
3360 bool queue_trunc = false;
3361 bool queue_invalidate = false;
3362 bool deleted_inode = false;
3363 bool fill_inline = false;
3364
3365 dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n",
3366 inode, cap, session->s_mds, seq, ceph_cap_string(newcaps));
3367 dout(" size %llu max_size %llu, i_size %llu\n", size, max_size,
3368 i_size_read(inode));
3369
3370
3371 /*
3372 * If CACHE is being revoked, and we have no dirty buffers,
3373 * try to invalidate (once). (If there are dirty buffers, we
3374 * will invalidate _after_ writeback.)
3375 */
3376 if (S_ISREG(inode->i_mode) && /* don't invalidate readdir cache */
3377 ((cap->issued & ~newcaps) & CEPH_CAP_FILE_CACHE) &&
3378 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0 &&
3379 !(ci->i_wrbuffer_ref || ci->i_wb_ref)) {
3380 if (try_nonblocking_invalidate(inode)) {
3381 /* there were locked pages.. invalidate later
3382 in a separate thread. */
3383 if (ci->i_rdcache_revoking != ci->i_rdcache_gen) {
3384 queue_invalidate = true;
3385 ci->i_rdcache_revoking = ci->i_rdcache_gen;
3386 }
3387 }
3388 }
3389
3390 if (was_stale)
3391 cap->issued = cap->implemented = CEPH_CAP_PIN;
3392
3393 /*
3394 * auth mds of the inode changed. we received the cap export message,
3395 * but still haven't received the cap import message. handle_cap_export
3396 * updated the new auth MDS' cap.
3397 *
3398 * "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing a message
3399 * that was sent before the cap import message. So don't remove caps.
3400 */
3401 if (ceph_seq_cmp(seq, cap->seq) <= 0) {
3402 WARN_ON(cap != ci->i_auth_cap);
3403 WARN_ON(cap->cap_id != le64_to_cpu(grant->cap_id));
3404 seq = cap->seq;
3405 newcaps |= cap->issued;
3406 }
3407
3408 /* side effects now are allowed */
3409 cap->cap_gen = atomic_read(&session->s_cap_gen);
3410 cap->seq = seq;
3411
3412 __check_cap_issue(ci, cap, newcaps);
3413
3414 inode_set_max_iversion_raw(inode, extra_info->change_attr);
3415
3416 if ((newcaps & CEPH_CAP_AUTH_SHARED) &&
3417 (extra_info->issued & CEPH_CAP_AUTH_EXCL) == 0) {
3418 umode_t mode = le32_to_cpu(grant->mode);
3419
3420 if (inode_wrong_type(inode, mode))
3421 pr_warn_once("inode type changed! (ino %llx.%llx is 0%o, mds says 0%o)\n",
3422 ceph_vinop(inode), inode->i_mode, mode);
3423 else
3424 inode->i_mode = mode;
3425 inode->i_uid = make_kuid(&init_user_ns, le32_to_cpu(grant->uid));
3426 inode->i_gid = make_kgid(&init_user_ns, le32_to_cpu(grant->gid));
3427 ci->i_btime = extra_info->btime;
3428 dout("%p mode 0%o uid.gid %d.%d\n", inode, inode->i_mode,
3429 from_kuid(&init_user_ns, inode->i_uid),
3430 from_kgid(&init_user_ns, inode->i_gid));
3431 }
3432
3433 if ((newcaps & CEPH_CAP_LINK_SHARED) &&
3434 (extra_info->issued & CEPH_CAP_LINK_EXCL) == 0) {
3435 set_nlink(inode, le32_to_cpu(grant->nlink));
3436 if (inode->i_nlink == 0)
3437 deleted_inode = true;
3438 }
3439
3440 if ((extra_info->issued & CEPH_CAP_XATTR_EXCL) == 0 &&
3441 grant->xattr_len) {
3442 int len = le32_to_cpu(grant->xattr_len);
3443 u64 version = le64_to_cpu(grant->xattr_version);
3444
3445 if (version > ci->i_xattrs.version) {
3446 dout(" got new xattrs v%llu on %p len %d\n",
3447 version, inode, len);
3448 if (ci->i_xattrs.blob)
3449 ceph_buffer_put(ci->i_xattrs.blob);
3450 ci->i_xattrs.blob = ceph_buffer_get(xattr_buf);
3451 ci->i_xattrs.version = version;
3452 ceph_forget_all_cached_acls(inode);
3453 ceph_security_invalidate_secctx(inode);
3454 }
3455 }
3456
3457 if (newcaps & CEPH_CAP_ANY_RD) {
3458 struct timespec64 mtime, atime, ctime;
3459 /* ctime/mtime/atime? */
3460 ceph_decode_timespec64(&mtime, &grant->mtime);
3461 ceph_decode_timespec64(&atime, &grant->atime);
3462 ceph_decode_timespec64(&ctime, &grant->ctime);
3463 ceph_fill_file_time(inode, extra_info->issued,
3464 le32_to_cpu(grant->time_warp_seq),
3465 &ctime, &mtime, &atime);
3466 }
3467
3468 if ((newcaps & CEPH_CAP_FILE_SHARED) && extra_info->dirstat_valid) {
3469 ci->i_files = extra_info->nfiles;
3470 ci->i_subdirs = extra_info->nsubdirs;
3471 }
3472
3473 if (newcaps & (CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR)) {
3474 /* file layout may have changed */
3475 s64 old_pool = ci->i_layout.pool_id;
3476 struct ceph_string *old_ns;
3477
3478 ceph_file_layout_from_legacy(&ci->i_layout, &grant->layout);
3479 old_ns = rcu_dereference_protected(ci->i_layout.pool_ns,
3480 lockdep_is_held(&ci->i_ceph_lock));
3481 rcu_assign_pointer(ci->i_layout.pool_ns, extra_info->pool_ns);
3482
3483 if (ci->i_layout.pool_id != old_pool ||
3484 extra_info->pool_ns != old_ns)
3485 ci->i_ceph_flags &= ~CEPH_I_POOL_PERM;
3486
3487 extra_info->pool_ns = old_ns;
3488
3489 /* size/truncate_seq? */
3490 queue_trunc = ceph_fill_file_size(inode, extra_info->issued,
3491 le32_to_cpu(grant->truncate_seq),
3492 le64_to_cpu(grant->truncate_size),
3493 size);
3494 }
3495
3496 if (ci->i_auth_cap == cap && (newcaps & CEPH_CAP_ANY_FILE_WR)) {
3497 if (max_size != ci->i_max_size) {
3498 dout("max_size %lld -> %llu\n",
3499 ci->i_max_size, max_size);
3500 ci->i_max_size = max_size;
3501 if (max_size >= ci->i_wanted_max_size) {
3502 ci->i_wanted_max_size = 0; /* reset */
3503 ci->i_requested_max_size = 0;
3504 }
3505 wake = true;
3506 }
3507 }
3508
3509 /* check cap bits */
3510 wanted = __ceph_caps_wanted(ci);
3511 used = __ceph_caps_used(ci);
3512 dirty = __ceph_caps_dirty(ci);
3513 dout(" my wanted = %s, used = %s, dirty %s\n",
3514 ceph_cap_string(wanted),
3515 ceph_cap_string(used),
3516 ceph_cap_string(dirty));
3517
3518 if ((was_stale || le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) &&
3519 (wanted & ~(cap->mds_wanted | newcaps))) {
3520 /*
3521 * If mds is importing cap, prior cap messages that update
3522 * 'wanted' may get dropped by mds (migrate seq mismatch).
3523 *
3524 * We don't send cap message to update 'wanted' if what we
3525 * want are already issued. If mds revokes caps, cap message
3526 * that releases caps also tells mds what we want. But if
3527 * caps got revoked by mds forcedly (session stale). We may
3528 * haven't told mds what we want.
3529 */
3530 check_caps = 1;
3531 }
3532
3533 /* revocation, grant, or no-op? */
3534 if (cap->issued & ~newcaps) {
3535 int revoking = cap->issued & ~newcaps;
3536
3537 dout("revocation: %s -> %s (revoking %s)\n",
3538 ceph_cap_string(cap->issued),
3539 ceph_cap_string(newcaps),
3540 ceph_cap_string(revoking));
3541 if (S_ISREG(inode->i_mode) &&
3542 (revoking & used & CEPH_CAP_FILE_BUFFER))
3543 writeback = true; /* initiate writeback; will delay ack */
3544 else if (queue_invalidate &&
3545 revoking == CEPH_CAP_FILE_CACHE &&
3546 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0)
3547 ; /* do nothing yet, invalidation will be queued */
3548 else if (cap == ci->i_auth_cap)
3549 check_caps = 1; /* check auth cap only */
3550 else
3551 check_caps = 2; /* check all caps */
3552 cap->issued = newcaps;
3553 cap->implemented |= newcaps;
3554 } else if (cap->issued == newcaps) {
3555 dout("caps unchanged: %s -> %s\n",
3556 ceph_cap_string(cap->issued), ceph_cap_string(newcaps));
3557 } else {
3558 dout("grant: %s -> %s\n", ceph_cap_string(cap->issued),
3559 ceph_cap_string(newcaps));
3560 /* non-auth MDS is revoking the newly grant caps ? */
3561 if (cap == ci->i_auth_cap &&
3562 __ceph_caps_revoking_other(ci, cap, newcaps))
3563 check_caps = 2;
3564
3565 cap->issued = newcaps;
3566 cap->implemented |= newcaps; /* add bits only, to
3567 * avoid stepping on a
3568 * pending revocation */
3569 wake = true;
3570 }
3571 BUG_ON(cap->issued & ~cap->implemented);
3572
3573 if (extra_info->inline_version > 0 &&
3574 extra_info->inline_version >= ci->i_inline_version) {
3575 ci->i_inline_version = extra_info->inline_version;
3576 if (ci->i_inline_version != CEPH_INLINE_NONE &&
3577 (newcaps & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)))
3578 fill_inline = true;
3579 }
3580
3581 if (le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) {
3582 if (ci->i_auth_cap == cap) {
3583 if (newcaps & ~extra_info->issued)
3584 wake = true;
3585
3586 if (ci->i_requested_max_size > max_size ||
3587 !(le32_to_cpu(grant->wanted) & CEPH_CAP_ANY_FILE_WR)) {
3588 /* re-request max_size if necessary */
3589 ci->i_requested_max_size = 0;
3590 wake = true;
3591 }
3592
3593 ceph_kick_flushing_inode_caps(session, ci);
3594 }
3595 up_read(&session->s_mdsc->snap_rwsem);
3596 }
3597 spin_unlock(&ci->i_ceph_lock);
3598
3599 if (fill_inline)
3600 ceph_fill_inline_data(inode, NULL, extra_info->inline_data,
3601 extra_info->inline_len);
3602
3603 if (queue_trunc)
3604 ceph_queue_vmtruncate(inode);
3605
3606 if (writeback)
3607 /*
3608 * queue inode for writeback: we can't actually call
3609 * filemap_write_and_wait, etc. from message handler
3610 * context.
3611 */
3612 ceph_queue_writeback(inode);
3613 if (queue_invalidate)
3614 ceph_queue_invalidate(inode);
3615 if (deleted_inode)
3616 invalidate_aliases(inode);
3617 if (wake)
3618 wake_up_all(&ci->i_cap_wq);
3619
3620 mutex_unlock(&session->s_mutex);
3621 if (check_caps == 1)
3622 ceph_check_caps(ci, CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL,
3623 session);
3624 else if (check_caps == 2)
3625 ceph_check_caps(ci, CHECK_CAPS_NOINVAL, session);
3626 }
3627
3628 /*
3629 * Handle FLUSH_ACK from MDS, indicating that metadata we sent to the
3630 * MDS has been safely committed.
3631 */
handle_cap_flush_ack(struct inode * inode,u64 flush_tid,struct ceph_mds_caps * m,struct ceph_mds_session * session,struct ceph_cap * cap)3632 static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
3633 struct ceph_mds_caps *m,
3634 struct ceph_mds_session *session,
3635 struct ceph_cap *cap)
3636 __releases(ci->i_ceph_lock)
3637 {
3638 struct ceph_inode_info *ci = ceph_inode(inode);
3639 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
3640 struct ceph_cap_flush *cf, *tmp_cf;
3641 LIST_HEAD(to_remove);
3642 unsigned seq = le32_to_cpu(m->seq);
3643 int dirty = le32_to_cpu(m->dirty);
3644 int cleaned = 0;
3645 bool drop = false;
3646 bool wake_ci = false;
3647 bool wake_mdsc = false;
3648
3649 list_for_each_entry_safe(cf, tmp_cf, &ci->i_cap_flush_list, i_list) {
3650 /* Is this the one that was flushed? */
3651 if (cf->tid == flush_tid)
3652 cleaned = cf->caps;
3653
3654 /* Is this a capsnap? */
3655 if (cf->is_capsnap)
3656 continue;
3657
3658 if (cf->tid <= flush_tid) {
3659 /*
3660 * An earlier or current tid. The FLUSH_ACK should
3661 * represent a superset of this flush's caps.
3662 */
3663 wake_ci |= __detach_cap_flush_from_ci(ci, cf);
3664 list_add_tail(&cf->i_list, &to_remove);
3665 } else {
3666 /*
3667 * This is a later one. Any caps in it are still dirty
3668 * so don't count them as cleaned.
3669 */
3670 cleaned &= ~cf->caps;
3671 if (!cleaned)
3672 break;
3673 }
3674 }
3675
3676 dout("handle_cap_flush_ack inode %p mds%d seq %d on %s cleaned %s,"
3677 " flushing %s -> %s\n",
3678 inode, session->s_mds, seq, ceph_cap_string(dirty),
3679 ceph_cap_string(cleaned), ceph_cap_string(ci->i_flushing_caps),
3680 ceph_cap_string(ci->i_flushing_caps & ~cleaned));
3681
3682 if (list_empty(&to_remove) && !cleaned)
3683 goto out;
3684
3685 ci->i_flushing_caps &= ~cleaned;
3686
3687 spin_lock(&mdsc->cap_dirty_lock);
3688
3689 list_for_each_entry(cf, &to_remove, i_list)
3690 wake_mdsc |= __detach_cap_flush_from_mdsc(mdsc, cf);
3691
3692 if (ci->i_flushing_caps == 0) {
3693 if (list_empty(&ci->i_cap_flush_list)) {
3694 list_del_init(&ci->i_flushing_item);
3695 if (!list_empty(&session->s_cap_flushing)) {
3696 dout(" mds%d still flushing cap on %p\n",
3697 session->s_mds,
3698 &list_first_entry(&session->s_cap_flushing,
3699 struct ceph_inode_info,
3700 i_flushing_item)->netfs.inode);
3701 }
3702 }
3703 mdsc->num_cap_flushing--;
3704 dout(" inode %p now !flushing\n", inode);
3705
3706 if (ci->i_dirty_caps == 0) {
3707 dout(" inode %p now clean\n", inode);
3708 BUG_ON(!list_empty(&ci->i_dirty_item));
3709 drop = true;
3710 if (ci->i_wr_ref == 0 &&
3711 ci->i_wrbuffer_ref_head == 0) {
3712 BUG_ON(!ci->i_head_snapc);
3713 ceph_put_snap_context(ci->i_head_snapc);
3714 ci->i_head_snapc = NULL;
3715 }
3716 } else {
3717 BUG_ON(list_empty(&ci->i_dirty_item));
3718 }
3719 }
3720 spin_unlock(&mdsc->cap_dirty_lock);
3721
3722 out:
3723 spin_unlock(&ci->i_ceph_lock);
3724
3725 while (!list_empty(&to_remove)) {
3726 cf = list_first_entry(&to_remove,
3727 struct ceph_cap_flush, i_list);
3728 list_del_init(&cf->i_list);
3729 if (!cf->is_capsnap)
3730 ceph_free_cap_flush(cf);
3731 }
3732
3733 if (wake_ci)
3734 wake_up_all(&ci->i_cap_wq);
3735 if (wake_mdsc)
3736 wake_up_all(&mdsc->cap_flushing_wq);
3737 if (drop)
3738 iput(inode);
3739 }
3740
__ceph_remove_capsnap(struct inode * inode,struct ceph_cap_snap * capsnap,bool * wake_ci,bool * wake_mdsc)3741 void __ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
3742 bool *wake_ci, bool *wake_mdsc)
3743 {
3744 struct ceph_inode_info *ci = ceph_inode(inode);
3745 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
3746 bool ret;
3747
3748 lockdep_assert_held(&ci->i_ceph_lock);
3749
3750 dout("removing capsnap %p, inode %p ci %p\n", capsnap, inode, ci);
3751
3752 list_del_init(&capsnap->ci_item);
3753 ret = __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
3754 if (wake_ci)
3755 *wake_ci = ret;
3756
3757 spin_lock(&mdsc->cap_dirty_lock);
3758 if (list_empty(&ci->i_cap_flush_list))
3759 list_del_init(&ci->i_flushing_item);
3760
3761 ret = __detach_cap_flush_from_mdsc(mdsc, &capsnap->cap_flush);
3762 if (wake_mdsc)
3763 *wake_mdsc = ret;
3764 spin_unlock(&mdsc->cap_dirty_lock);
3765 }
3766
ceph_remove_capsnap(struct inode * inode,struct ceph_cap_snap * capsnap,bool * wake_ci,bool * wake_mdsc)3767 void ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
3768 bool *wake_ci, bool *wake_mdsc)
3769 {
3770 struct ceph_inode_info *ci = ceph_inode(inode);
3771
3772 lockdep_assert_held(&ci->i_ceph_lock);
3773
3774 WARN_ON_ONCE(capsnap->dirty_pages || capsnap->writing);
3775 __ceph_remove_capsnap(inode, capsnap, wake_ci, wake_mdsc);
3776 }
3777
3778 /*
3779 * Handle FLUSHSNAP_ACK. MDS has flushed snap data to disk and we can
3780 * throw away our cap_snap.
3781 *
3782 * Caller hold s_mutex.
3783 */
handle_cap_flushsnap_ack(struct inode * inode,u64 flush_tid,struct ceph_mds_caps * m,struct ceph_mds_session * session)3784 static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
3785 struct ceph_mds_caps *m,
3786 struct ceph_mds_session *session)
3787 {
3788 struct ceph_inode_info *ci = ceph_inode(inode);
3789 struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
3790 u64 follows = le64_to_cpu(m->snap_follows);
3791 struct ceph_cap_snap *capsnap = NULL, *iter;
3792 bool wake_ci = false;
3793 bool wake_mdsc = false;
3794
3795 dout("handle_cap_flushsnap_ack inode %p ci %p mds%d follows %lld\n",
3796 inode, ci, session->s_mds, follows);
3797
3798 spin_lock(&ci->i_ceph_lock);
3799 list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) {
3800 if (iter->follows == follows) {
3801 if (iter->cap_flush.tid != flush_tid) {
3802 dout(" cap_snap %p follows %lld tid %lld !="
3803 " %lld\n", iter, follows,
3804 flush_tid, iter->cap_flush.tid);
3805 break;
3806 }
3807 capsnap = iter;
3808 break;
3809 } else {
3810 dout(" skipping cap_snap %p follows %lld\n",
3811 iter, iter->follows);
3812 }
3813 }
3814 if (capsnap)
3815 ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc);
3816 spin_unlock(&ci->i_ceph_lock);
3817
3818 if (capsnap) {
3819 ceph_put_snap_context(capsnap->context);
3820 ceph_put_cap_snap(capsnap);
3821 if (wake_ci)
3822 wake_up_all(&ci->i_cap_wq);
3823 if (wake_mdsc)
3824 wake_up_all(&mdsc->cap_flushing_wq);
3825 iput(inode);
3826 }
3827 }
3828
3829 /*
3830 * Handle TRUNC from MDS, indicating file truncation.
3831 *
3832 * caller hold s_mutex.
3833 */
handle_cap_trunc(struct inode * inode,struct ceph_mds_caps * trunc,struct ceph_mds_session * session)3834 static bool handle_cap_trunc(struct inode *inode,
3835 struct ceph_mds_caps *trunc,
3836 struct ceph_mds_session *session)
3837 {
3838 struct ceph_inode_info *ci = ceph_inode(inode);
3839 int mds = session->s_mds;
3840 int seq = le32_to_cpu(trunc->seq);
3841 u32 truncate_seq = le32_to_cpu(trunc->truncate_seq);
3842 u64 truncate_size = le64_to_cpu(trunc->truncate_size);
3843 u64 size = le64_to_cpu(trunc->size);
3844 int implemented = 0;
3845 int dirty = __ceph_caps_dirty(ci);
3846 int issued = __ceph_caps_issued(ceph_inode(inode), &implemented);
3847 bool queue_trunc = false;
3848
3849 lockdep_assert_held(&ci->i_ceph_lock);
3850
3851 issued |= implemented | dirty;
3852
3853 dout("handle_cap_trunc inode %p mds%d seq %d to %lld seq %d\n",
3854 inode, mds, seq, truncate_size, truncate_seq);
3855 queue_trunc = ceph_fill_file_size(inode, issued,
3856 truncate_seq, truncate_size, size);
3857 return queue_trunc;
3858 }
3859
3860 /*
3861 * Handle EXPORT from MDS. Cap is being migrated _from_ this mds to a
3862 * different one. If we are the most recent migration we've seen (as
3863 * indicated by mseq), make note of the migrating cap bits for the
3864 * duration (until we see the corresponding IMPORT).
3865 *
3866 * caller holds s_mutex
3867 */
handle_cap_export(struct inode * inode,struct ceph_mds_caps * ex,struct ceph_mds_cap_peer * ph,struct ceph_mds_session * session)3868 static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
3869 struct ceph_mds_cap_peer *ph,
3870 struct ceph_mds_session *session)
3871 {
3872 struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
3873 struct ceph_mds_session *tsession = NULL;
3874 struct ceph_cap *cap, *tcap, *new_cap = NULL;
3875 struct ceph_inode_info *ci = ceph_inode(inode);
3876 u64 t_cap_id;
3877 unsigned mseq = le32_to_cpu(ex->migrate_seq);
3878 unsigned t_seq, t_mseq;
3879 int target, issued;
3880 int mds = session->s_mds;
3881
3882 if (ph) {
3883 t_cap_id = le64_to_cpu(ph->cap_id);
3884 t_seq = le32_to_cpu(ph->seq);
3885 t_mseq = le32_to_cpu(ph->mseq);
3886 target = le32_to_cpu(ph->mds);
3887 } else {
3888 t_cap_id = t_seq = t_mseq = 0;
3889 target = -1;
3890 }
3891
3892 dout("handle_cap_export inode %p ci %p mds%d mseq %d target %d\n",
3893 inode, ci, mds, mseq, target);
3894 retry:
3895 down_read(&mdsc->snap_rwsem);
3896 spin_lock(&ci->i_ceph_lock);
3897 cap = __get_cap_for_mds(ci, mds);
3898 if (!cap || cap->cap_id != le64_to_cpu(ex->cap_id))
3899 goto out_unlock;
3900
3901 if (target < 0) {
3902 ceph_remove_cap(cap, false);
3903 goto out_unlock;
3904 }
3905
3906 /*
3907 * now we know we haven't received the cap import message yet
3908 * because the exported cap still exist.
3909 */
3910
3911 issued = cap->issued;
3912 if (issued != cap->implemented)
3913 pr_err_ratelimited("handle_cap_export: issued != implemented: "
3914 "ino (%llx.%llx) mds%d seq %d mseq %d "
3915 "issued %s implemented %s\n",
3916 ceph_vinop(inode), mds, cap->seq, cap->mseq,
3917 ceph_cap_string(issued),
3918 ceph_cap_string(cap->implemented));
3919
3920
3921 tcap = __get_cap_for_mds(ci, target);
3922 if (tcap) {
3923 /* already have caps from the target */
3924 if (tcap->cap_id == t_cap_id &&
3925 ceph_seq_cmp(tcap->seq, t_seq) < 0) {
3926 dout(" updating import cap %p mds%d\n", tcap, target);
3927 tcap->cap_id = t_cap_id;
3928 tcap->seq = t_seq - 1;
3929 tcap->issue_seq = t_seq - 1;
3930 tcap->issued |= issued;
3931 tcap->implemented |= issued;
3932 if (cap == ci->i_auth_cap) {
3933 ci->i_auth_cap = tcap;
3934 change_auth_cap_ses(ci, tcap->session);
3935 }
3936 }
3937 ceph_remove_cap(cap, false);
3938 goto out_unlock;
3939 } else if (tsession) {
3940 /* add placeholder for the export tagert */
3941 int flag = (cap == ci->i_auth_cap) ? CEPH_CAP_FLAG_AUTH : 0;
3942 tcap = new_cap;
3943 ceph_add_cap(inode, tsession, t_cap_id, issued, 0,
3944 t_seq - 1, t_mseq, (u64)-1, flag, &new_cap);
3945
3946 if (!list_empty(&ci->i_cap_flush_list) &&
3947 ci->i_auth_cap == tcap) {
3948 spin_lock(&mdsc->cap_dirty_lock);
3949 list_move_tail(&ci->i_flushing_item,
3950 &tcap->session->s_cap_flushing);
3951 spin_unlock(&mdsc->cap_dirty_lock);
3952 }
3953
3954 ceph_remove_cap(cap, false);
3955 goto out_unlock;
3956 }
3957
3958 spin_unlock(&ci->i_ceph_lock);
3959 up_read(&mdsc->snap_rwsem);
3960 mutex_unlock(&session->s_mutex);
3961
3962 /* open target session */
3963 tsession = ceph_mdsc_open_export_target_session(mdsc, target);
3964 if (!IS_ERR(tsession)) {
3965 if (mds > target) {
3966 mutex_lock(&session->s_mutex);
3967 mutex_lock_nested(&tsession->s_mutex,
3968 SINGLE_DEPTH_NESTING);
3969 } else {
3970 mutex_lock(&tsession->s_mutex);
3971 mutex_lock_nested(&session->s_mutex,
3972 SINGLE_DEPTH_NESTING);
3973 }
3974 new_cap = ceph_get_cap(mdsc, NULL);
3975 } else {
3976 WARN_ON(1);
3977 tsession = NULL;
3978 target = -1;
3979 mutex_lock(&session->s_mutex);
3980 }
3981 goto retry;
3982
3983 out_unlock:
3984 spin_unlock(&ci->i_ceph_lock);
3985 up_read(&mdsc->snap_rwsem);
3986 mutex_unlock(&session->s_mutex);
3987 if (tsession) {
3988 mutex_unlock(&tsession->s_mutex);
3989 ceph_put_mds_session(tsession);
3990 }
3991 if (new_cap)
3992 ceph_put_cap(mdsc, new_cap);
3993 }
3994
3995 /*
3996 * Handle cap IMPORT.
3997 *
3998 * caller holds s_mutex. acquires i_ceph_lock
3999 */
handle_cap_import(struct ceph_mds_client * mdsc,struct inode * inode,struct ceph_mds_caps * im,struct ceph_mds_cap_peer * ph,struct ceph_mds_session * session,struct ceph_cap ** target_cap,int * old_issued)4000 static void handle_cap_import(struct ceph_mds_client *mdsc,
4001 struct inode *inode, struct ceph_mds_caps *im,
4002 struct ceph_mds_cap_peer *ph,
4003 struct ceph_mds_session *session,
4004 struct ceph_cap **target_cap, int *old_issued)
4005 {
4006 struct ceph_inode_info *ci = ceph_inode(inode);
4007 struct ceph_cap *cap, *ocap, *new_cap = NULL;
4008 int mds = session->s_mds;
4009 int issued;
4010 unsigned caps = le32_to_cpu(im->caps);
4011 unsigned wanted = le32_to_cpu(im->wanted);
4012 unsigned seq = le32_to_cpu(im->seq);
4013 unsigned mseq = le32_to_cpu(im->migrate_seq);
4014 u64 realmino = le64_to_cpu(im->realm);
4015 u64 cap_id = le64_to_cpu(im->cap_id);
4016 u64 p_cap_id;
4017 int peer;
4018
4019 if (ph) {
4020 p_cap_id = le64_to_cpu(ph->cap_id);
4021 peer = le32_to_cpu(ph->mds);
4022 } else {
4023 p_cap_id = 0;
4024 peer = -1;
4025 }
4026
4027 dout("handle_cap_import inode %p ci %p mds%d mseq %d peer %d\n",
4028 inode, ci, mds, mseq, peer);
4029 retry:
4030 cap = __get_cap_for_mds(ci, mds);
4031 if (!cap) {
4032 if (!new_cap) {
4033 spin_unlock(&ci->i_ceph_lock);
4034 new_cap = ceph_get_cap(mdsc, NULL);
4035 spin_lock(&ci->i_ceph_lock);
4036 goto retry;
4037 }
4038 cap = new_cap;
4039 } else {
4040 if (new_cap) {
4041 ceph_put_cap(mdsc, new_cap);
4042 new_cap = NULL;
4043 }
4044 }
4045
4046 __ceph_caps_issued(ci, &issued);
4047 issued |= __ceph_caps_dirty(ci);
4048
4049 ceph_add_cap(inode, session, cap_id, caps, wanted, seq, mseq,
4050 realmino, CEPH_CAP_FLAG_AUTH, &new_cap);
4051
4052 ocap = peer >= 0 ? __get_cap_for_mds(ci, peer) : NULL;
4053 if (ocap && ocap->cap_id == p_cap_id) {
4054 dout(" remove export cap %p mds%d flags %d\n",
4055 ocap, peer, ph->flags);
4056 if ((ph->flags & CEPH_CAP_FLAG_AUTH) &&
4057 (ocap->seq != le32_to_cpu(ph->seq) ||
4058 ocap->mseq != le32_to_cpu(ph->mseq))) {
4059 pr_err_ratelimited("handle_cap_import: "
4060 "mismatched seq/mseq: ino (%llx.%llx) "
4061 "mds%d seq %d mseq %d importer mds%d "
4062 "has peer seq %d mseq %d\n",
4063 ceph_vinop(inode), peer, ocap->seq,
4064 ocap->mseq, mds, le32_to_cpu(ph->seq),
4065 le32_to_cpu(ph->mseq));
4066 }
4067 ceph_remove_cap(ocap, (ph->flags & CEPH_CAP_FLAG_RELEASE));
4068 }
4069
4070 *old_issued = issued;
4071 *target_cap = cap;
4072 }
4073
4074 /*
4075 * Handle a caps message from the MDS.
4076 *
4077 * Identify the appropriate session, inode, and call the right handler
4078 * based on the cap op.
4079 */
ceph_handle_caps(struct ceph_mds_session * session,struct ceph_msg * msg)4080 void ceph_handle_caps(struct ceph_mds_session *session,
4081 struct ceph_msg *msg)
4082 {
4083 struct ceph_mds_client *mdsc = session->s_mdsc;
4084 struct inode *inode;
4085 struct ceph_inode_info *ci;
4086 struct ceph_cap *cap;
4087 struct ceph_mds_caps *h;
4088 struct ceph_mds_cap_peer *peer = NULL;
4089 struct ceph_snap_realm *realm = NULL;
4090 int op;
4091 int msg_version = le16_to_cpu(msg->hdr.version);
4092 u32 seq, mseq;
4093 struct ceph_vino vino;
4094 void *snaptrace;
4095 size_t snaptrace_len;
4096 void *p, *end;
4097 struct cap_extra_info extra_info = {};
4098 bool queue_trunc;
4099
4100 dout("handle_caps from mds%d\n", session->s_mds);
4101
4102 /* decode */
4103 end = msg->front.iov_base + msg->front.iov_len;
4104 if (msg->front.iov_len < sizeof(*h))
4105 goto bad;
4106 h = msg->front.iov_base;
4107 op = le32_to_cpu(h->op);
4108 vino.ino = le64_to_cpu(h->ino);
4109 vino.snap = CEPH_NOSNAP;
4110 seq = le32_to_cpu(h->seq);
4111 mseq = le32_to_cpu(h->migrate_seq);
4112
4113 snaptrace = h + 1;
4114 snaptrace_len = le32_to_cpu(h->snap_trace_len);
4115 p = snaptrace + snaptrace_len;
4116
4117 if (msg_version >= 2) {
4118 u32 flock_len;
4119 ceph_decode_32_safe(&p, end, flock_len, bad);
4120 if (p + flock_len > end)
4121 goto bad;
4122 p += flock_len;
4123 }
4124
4125 if (msg_version >= 3) {
4126 if (op == CEPH_CAP_OP_IMPORT) {
4127 if (p + sizeof(*peer) > end)
4128 goto bad;
4129 peer = p;
4130 p += sizeof(*peer);
4131 } else if (op == CEPH_CAP_OP_EXPORT) {
4132 /* recorded in unused fields */
4133 peer = (void *)&h->size;
4134 }
4135 }
4136
4137 if (msg_version >= 4) {
4138 ceph_decode_64_safe(&p, end, extra_info.inline_version, bad);
4139 ceph_decode_32_safe(&p, end, extra_info.inline_len, bad);
4140 if (p + extra_info.inline_len > end)
4141 goto bad;
4142 extra_info.inline_data = p;
4143 p += extra_info.inline_len;
4144 }
4145
4146 if (msg_version >= 5) {
4147 struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
4148 u32 epoch_barrier;
4149
4150 ceph_decode_32_safe(&p, end, epoch_barrier, bad);
4151 ceph_osdc_update_epoch_barrier(osdc, epoch_barrier);
4152 }
4153
4154 if (msg_version >= 8) {
4155 u32 pool_ns_len;
4156
4157 /* version >= 6 */
4158 ceph_decode_skip_64(&p, end, bad); // flush_tid
4159 /* version >= 7 */
4160 ceph_decode_skip_32(&p, end, bad); // caller_uid
4161 ceph_decode_skip_32(&p, end, bad); // caller_gid
4162 /* version >= 8 */
4163 ceph_decode_32_safe(&p, end, pool_ns_len, bad);
4164 if (pool_ns_len > 0) {
4165 ceph_decode_need(&p, end, pool_ns_len, bad);
4166 extra_info.pool_ns =
4167 ceph_find_or_create_string(p, pool_ns_len);
4168 p += pool_ns_len;
4169 }
4170 }
4171
4172 if (msg_version >= 9) {
4173 struct ceph_timespec *btime;
4174
4175 if (p + sizeof(*btime) > end)
4176 goto bad;
4177 btime = p;
4178 ceph_decode_timespec64(&extra_info.btime, btime);
4179 p += sizeof(*btime);
4180 ceph_decode_64_safe(&p, end, extra_info.change_attr, bad);
4181 }
4182
4183 if (msg_version >= 11) {
4184 /* version >= 10 */
4185 ceph_decode_skip_32(&p, end, bad); // flags
4186 /* version >= 11 */
4187 extra_info.dirstat_valid = true;
4188 ceph_decode_64_safe(&p, end, extra_info.nfiles, bad);
4189 ceph_decode_64_safe(&p, end, extra_info.nsubdirs, bad);
4190 }
4191
4192 /* lookup ino */
4193 inode = ceph_find_inode(mdsc->fsc->sb, vino);
4194 dout(" op %s ino %llx.%llx inode %p\n", ceph_cap_op_name(op), vino.ino,
4195 vino.snap, inode);
4196
4197 mutex_lock(&session->s_mutex);
4198 inc_session_sequence(session);
4199 dout(" mds%d seq %lld cap seq %u\n", session->s_mds, session->s_seq,
4200 (unsigned)seq);
4201
4202 if (!inode) {
4203 dout(" i don't have ino %llx\n", vino.ino);
4204
4205 if (op == CEPH_CAP_OP_IMPORT) {
4206 cap = ceph_get_cap(mdsc, NULL);
4207 cap->cap_ino = vino.ino;
4208 cap->queue_release = 1;
4209 cap->cap_id = le64_to_cpu(h->cap_id);
4210 cap->mseq = mseq;
4211 cap->seq = seq;
4212 cap->issue_seq = seq;
4213 spin_lock(&session->s_cap_lock);
4214 __ceph_queue_cap_release(session, cap);
4215 spin_unlock(&session->s_cap_lock);
4216 }
4217 goto flush_cap_releases;
4218 }
4219 ci = ceph_inode(inode);
4220
4221 /* these will work even if we don't have a cap yet */
4222 switch (op) {
4223 case CEPH_CAP_OP_FLUSHSNAP_ACK:
4224 handle_cap_flushsnap_ack(inode, le64_to_cpu(msg->hdr.tid),
4225 h, session);
4226 goto done;
4227
4228 case CEPH_CAP_OP_EXPORT:
4229 handle_cap_export(inode, h, peer, session);
4230 goto done_unlocked;
4231
4232 case CEPH_CAP_OP_IMPORT:
4233 realm = NULL;
4234 if (snaptrace_len) {
4235 down_write(&mdsc->snap_rwsem);
4236 ceph_update_snap_trace(mdsc, snaptrace,
4237 snaptrace + snaptrace_len,
4238 false, &realm);
4239 downgrade_write(&mdsc->snap_rwsem);
4240 } else {
4241 down_read(&mdsc->snap_rwsem);
4242 }
4243 spin_lock(&ci->i_ceph_lock);
4244 handle_cap_import(mdsc, inode, h, peer, session,
4245 &cap, &extra_info.issued);
4246 handle_cap_grant(inode, session, cap,
4247 h, msg->middle, &extra_info);
4248 if (realm)
4249 ceph_put_snap_realm(mdsc, realm);
4250 goto done_unlocked;
4251 }
4252
4253 /* the rest require a cap */
4254 spin_lock(&ci->i_ceph_lock);
4255 cap = __get_cap_for_mds(ceph_inode(inode), session->s_mds);
4256 if (!cap) {
4257 dout(" no cap on %p ino %llx.%llx from mds%d\n",
4258 inode, ceph_ino(inode), ceph_snap(inode),
4259 session->s_mds);
4260 spin_unlock(&ci->i_ceph_lock);
4261 goto flush_cap_releases;
4262 }
4263
4264 /* note that each of these drops i_ceph_lock for us */
4265 switch (op) {
4266 case CEPH_CAP_OP_REVOKE:
4267 case CEPH_CAP_OP_GRANT:
4268 __ceph_caps_issued(ci, &extra_info.issued);
4269 extra_info.issued |= __ceph_caps_dirty(ci);
4270 handle_cap_grant(inode, session, cap,
4271 h, msg->middle, &extra_info);
4272 goto done_unlocked;
4273
4274 case CEPH_CAP_OP_FLUSH_ACK:
4275 handle_cap_flush_ack(inode, le64_to_cpu(msg->hdr.tid),
4276 h, session, cap);
4277 break;
4278
4279 case CEPH_CAP_OP_TRUNC:
4280 queue_trunc = handle_cap_trunc(inode, h, session);
4281 spin_unlock(&ci->i_ceph_lock);
4282 if (queue_trunc)
4283 ceph_queue_vmtruncate(inode);
4284 break;
4285
4286 default:
4287 spin_unlock(&ci->i_ceph_lock);
4288 pr_err("ceph_handle_caps: unknown cap op %d %s\n", op,
4289 ceph_cap_op_name(op));
4290 }
4291
4292 done:
4293 mutex_unlock(&session->s_mutex);
4294 done_unlocked:
4295 iput(inode);
4296 out:
4297 ceph_put_string(extra_info.pool_ns);
4298 return;
4299
4300 flush_cap_releases:
4301 /*
4302 * send any cap release message to try to move things
4303 * along for the mds (who clearly thinks we still have this
4304 * cap).
4305 */
4306 ceph_flush_cap_releases(mdsc, session);
4307 goto done;
4308
4309 bad:
4310 pr_err("ceph_handle_caps: corrupt message\n");
4311 ceph_msg_dump(msg);
4312 goto out;
4313 }
4314
4315 /*
4316 * Delayed work handler to process end of delayed cap release LRU list.
4317 *
4318 * If new caps are added to the list while processing it, these won't get
4319 * processed in this run. In this case, the ci->i_hold_caps_max will be
4320 * returned so that the work can be scheduled accordingly.
4321 */
ceph_check_delayed_caps(struct ceph_mds_client * mdsc)4322 unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
4323 {
4324 struct inode *inode;
4325 struct ceph_inode_info *ci;
4326 struct ceph_mount_options *opt = mdsc->fsc->mount_options;
4327 unsigned long delay_max = opt->caps_wanted_delay_max * HZ;
4328 unsigned long loop_start = jiffies;
4329 unsigned long delay = 0;
4330
4331 dout("check_delayed_caps\n");
4332 spin_lock(&mdsc->cap_delay_lock);
4333 while (!list_empty(&mdsc->cap_delay_list)) {
4334 ci = list_first_entry(&mdsc->cap_delay_list,
4335 struct ceph_inode_info,
4336 i_cap_delay_list);
4337 if (time_before(loop_start, ci->i_hold_caps_max - delay_max)) {
4338 dout("%s caps added recently. Exiting loop", __func__);
4339 delay = ci->i_hold_caps_max;
4340 break;
4341 }
4342 if ((ci->i_ceph_flags & CEPH_I_FLUSH) == 0 &&
4343 time_before(jiffies, ci->i_hold_caps_max))
4344 break;
4345 list_del_init(&ci->i_cap_delay_list);
4346
4347 inode = igrab(&ci->netfs.inode);
4348 if (inode) {
4349 spin_unlock(&mdsc->cap_delay_lock);
4350 dout("check_delayed_caps on %p\n", inode);
4351 ceph_check_caps(ci, 0, NULL);
4352 iput(inode);
4353 spin_lock(&mdsc->cap_delay_lock);
4354 }
4355 }
4356 spin_unlock(&mdsc->cap_delay_lock);
4357
4358 return delay;
4359 }
4360
4361 /*
4362 * Flush all dirty caps to the mds
4363 */
flush_dirty_session_caps(struct ceph_mds_session * s)4364 static void flush_dirty_session_caps(struct ceph_mds_session *s)
4365 {
4366 struct ceph_mds_client *mdsc = s->s_mdsc;
4367 struct ceph_inode_info *ci;
4368 struct inode *inode;
4369
4370 dout("flush_dirty_caps\n");
4371 spin_lock(&mdsc->cap_dirty_lock);
4372 while (!list_empty(&s->s_cap_dirty)) {
4373 ci = list_first_entry(&s->s_cap_dirty, struct ceph_inode_info,
4374 i_dirty_item);
4375 inode = &ci->netfs.inode;
4376 ihold(inode);
4377 dout("flush_dirty_caps %llx.%llx\n", ceph_vinop(inode));
4378 spin_unlock(&mdsc->cap_dirty_lock);
4379 ceph_wait_on_async_create(inode);
4380 ceph_check_caps(ci, CHECK_CAPS_FLUSH, NULL);
4381 iput(inode);
4382 spin_lock(&mdsc->cap_dirty_lock);
4383 }
4384 spin_unlock(&mdsc->cap_dirty_lock);
4385 dout("flush_dirty_caps done\n");
4386 }
4387
ceph_flush_dirty_caps(struct ceph_mds_client * mdsc)4388 void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
4389 {
4390 ceph_mdsc_iterate_sessions(mdsc, flush_dirty_session_caps, true);
4391 }
4392
__ceph_touch_fmode(struct ceph_inode_info * ci,struct ceph_mds_client * mdsc,int fmode)4393 void __ceph_touch_fmode(struct ceph_inode_info *ci,
4394 struct ceph_mds_client *mdsc, int fmode)
4395 {
4396 unsigned long now = jiffies;
4397 if (fmode & CEPH_FILE_MODE_RD)
4398 ci->i_last_rd = now;
4399 if (fmode & CEPH_FILE_MODE_WR)
4400 ci->i_last_wr = now;
4401 /* queue periodic check */
4402 if (fmode &&
4403 __ceph_is_any_real_caps(ci) &&
4404 list_empty(&ci->i_cap_delay_list))
4405 __cap_delay_requeue(mdsc, ci);
4406 }
4407
ceph_get_fmode(struct ceph_inode_info * ci,int fmode,int count)4408 void ceph_get_fmode(struct ceph_inode_info *ci, int fmode, int count)
4409 {
4410 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->netfs.inode.i_sb);
4411 int bits = (fmode << 1) | 1;
4412 bool already_opened = false;
4413 int i;
4414
4415 if (count == 1)
4416 atomic64_inc(&mdsc->metric.opened_files);
4417
4418 spin_lock(&ci->i_ceph_lock);
4419 for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
4420 /*
4421 * If any of the mode ref is larger than 0,
4422 * that means it has been already opened by
4423 * others. Just skip checking the PIN ref.
4424 */
4425 if (i && ci->i_nr_by_mode[i])
4426 already_opened = true;
4427
4428 if (bits & (1 << i))
4429 ci->i_nr_by_mode[i] += count;
4430 }
4431
4432 if (!already_opened)
4433 percpu_counter_inc(&mdsc->metric.opened_inodes);
4434 spin_unlock(&ci->i_ceph_lock);
4435 }
4436
4437 /*
4438 * Drop open file reference. If we were the last open file,
4439 * we may need to release capabilities to the MDS (or schedule
4440 * their delayed release).
4441 */
ceph_put_fmode(struct ceph_inode_info * ci,int fmode,int count)4442 void ceph_put_fmode(struct ceph_inode_info *ci, int fmode, int count)
4443 {
4444 struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->netfs.inode.i_sb);
4445 int bits = (fmode << 1) | 1;
4446 bool is_closed = true;
4447 int i;
4448
4449 if (count == 1)
4450 atomic64_dec(&mdsc->metric.opened_files);
4451
4452 spin_lock(&ci->i_ceph_lock);
4453 for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
4454 if (bits & (1 << i)) {
4455 BUG_ON(ci->i_nr_by_mode[i] < count);
4456 ci->i_nr_by_mode[i] -= count;
4457 }
4458
4459 /*
4460 * If any of the mode ref is not 0 after
4461 * decreased, that means it is still opened
4462 * by others. Just skip checking the PIN ref.
4463 */
4464 if (i && ci->i_nr_by_mode[i])
4465 is_closed = false;
4466 }
4467
4468 if (is_closed)
4469 percpu_counter_dec(&mdsc->metric.opened_inodes);
4470 spin_unlock(&ci->i_ceph_lock);
4471 }
4472
4473 /*
4474 * For a soon-to-be unlinked file, drop the LINK caps. If it
4475 * looks like the link count will hit 0, drop any other caps (other
4476 * than PIN) we don't specifically want (due to the file still being
4477 * open).
4478 */
ceph_drop_caps_for_unlink(struct inode * inode)4479 int ceph_drop_caps_for_unlink(struct inode *inode)
4480 {
4481 struct ceph_inode_info *ci = ceph_inode(inode);
4482 int drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;
4483
4484 spin_lock(&ci->i_ceph_lock);
4485 if (inode->i_nlink == 1) {
4486 drop |= ~(__ceph_caps_wanted(ci) | CEPH_CAP_PIN);
4487
4488 if (__ceph_caps_dirty(ci)) {
4489 struct ceph_mds_client *mdsc =
4490 ceph_inode_to_client(inode)->mdsc;
4491 __cap_delay_requeue_front(mdsc, ci);
4492 }
4493 }
4494 spin_unlock(&ci->i_ceph_lock);
4495 return drop;
4496 }
4497
4498 /*
4499 * Helpers for embedding cap and dentry lease releases into mds
4500 * requests.
4501 *
4502 * @force is used by dentry_release (below) to force inclusion of a
4503 * record for the directory inode, even when there aren't any caps to
4504 * drop.
4505 */
ceph_encode_inode_release(void ** p,struct inode * inode,int mds,int drop,int unless,int force)4506 int ceph_encode_inode_release(void **p, struct inode *inode,
4507 int mds, int drop, int unless, int force)
4508 {
4509 struct ceph_inode_info *ci = ceph_inode(inode);
4510 struct ceph_cap *cap;
4511 struct ceph_mds_request_release *rel = *p;
4512 int used, dirty;
4513 int ret = 0;
4514
4515 spin_lock(&ci->i_ceph_lock);
4516 used = __ceph_caps_used(ci);
4517 dirty = __ceph_caps_dirty(ci);
4518
4519 dout("encode_inode_release %p mds%d used|dirty %s drop %s unless %s\n",
4520 inode, mds, ceph_cap_string(used|dirty), ceph_cap_string(drop),
4521 ceph_cap_string(unless));
4522
4523 /* only drop unused, clean caps */
4524 drop &= ~(used | dirty);
4525
4526 cap = __get_cap_for_mds(ci, mds);
4527 if (cap && __cap_is_valid(cap)) {
4528 unless &= cap->issued;
4529 if (unless) {
4530 if (unless & CEPH_CAP_AUTH_EXCL)
4531 drop &= ~CEPH_CAP_AUTH_SHARED;
4532 if (unless & CEPH_CAP_LINK_EXCL)
4533 drop &= ~CEPH_CAP_LINK_SHARED;
4534 if (unless & CEPH_CAP_XATTR_EXCL)
4535 drop &= ~CEPH_CAP_XATTR_SHARED;
4536 if (unless & CEPH_CAP_FILE_EXCL)
4537 drop &= ~CEPH_CAP_FILE_SHARED;
4538 }
4539
4540 if (force || (cap->issued & drop)) {
4541 if (cap->issued & drop) {
4542 int wanted = __ceph_caps_wanted(ci);
4543 dout("encode_inode_release %p cap %p "
4544 "%s -> %s, wanted %s -> %s\n", inode, cap,
4545 ceph_cap_string(cap->issued),
4546 ceph_cap_string(cap->issued & ~drop),
4547 ceph_cap_string(cap->mds_wanted),
4548 ceph_cap_string(wanted));
4549
4550 cap->issued &= ~drop;
4551 cap->implemented &= ~drop;
4552 cap->mds_wanted = wanted;
4553 if (cap == ci->i_auth_cap &&
4554 !(wanted & CEPH_CAP_ANY_FILE_WR))
4555 ci->i_requested_max_size = 0;
4556 } else {
4557 dout("encode_inode_release %p cap %p %s"
4558 " (force)\n", inode, cap,
4559 ceph_cap_string(cap->issued));
4560 }
4561
4562 rel->ino = cpu_to_le64(ceph_ino(inode));
4563 rel->cap_id = cpu_to_le64(cap->cap_id);
4564 rel->seq = cpu_to_le32(cap->seq);
4565 rel->issue_seq = cpu_to_le32(cap->issue_seq);
4566 rel->mseq = cpu_to_le32(cap->mseq);
4567 rel->caps = cpu_to_le32(cap->implemented);
4568 rel->wanted = cpu_to_le32(cap->mds_wanted);
4569 rel->dname_len = 0;
4570 rel->dname_seq = 0;
4571 *p += sizeof(*rel);
4572 ret = 1;
4573 } else {
4574 dout("encode_inode_release %p cap %p %s (noop)\n",
4575 inode, cap, ceph_cap_string(cap->issued));
4576 }
4577 }
4578 spin_unlock(&ci->i_ceph_lock);
4579 return ret;
4580 }
4581
ceph_encode_dentry_release(void ** p,struct dentry * dentry,struct inode * dir,int mds,int drop,int unless)4582 int ceph_encode_dentry_release(void **p, struct dentry *dentry,
4583 struct inode *dir,
4584 int mds, int drop, int unless)
4585 {
4586 struct dentry *parent = NULL;
4587 struct ceph_mds_request_release *rel = *p;
4588 struct ceph_dentry_info *di = ceph_dentry(dentry);
4589 int force = 0;
4590 int ret;
4591
4592 /*
4593 * force an record for the directory caps if we have a dentry lease.
4594 * this is racy (can't take i_ceph_lock and d_lock together), but it
4595 * doesn't have to be perfect; the mds will revoke anything we don't
4596 * release.
4597 */
4598 spin_lock(&dentry->d_lock);
4599 if (di->lease_session && di->lease_session->s_mds == mds)
4600 force = 1;
4601 if (!dir) {
4602 parent = dget(dentry->d_parent);
4603 dir = d_inode(parent);
4604 }
4605 spin_unlock(&dentry->d_lock);
4606
4607 ret = ceph_encode_inode_release(p, dir, mds, drop, unless, force);
4608 dput(parent);
4609
4610 spin_lock(&dentry->d_lock);
4611 if (ret && di->lease_session && di->lease_session->s_mds == mds) {
4612 dout("encode_dentry_release %p mds%d seq %d\n",
4613 dentry, mds, (int)di->lease_seq);
4614 rel->dname_len = cpu_to_le32(dentry->d_name.len);
4615 memcpy(*p, dentry->d_name.name, dentry->d_name.len);
4616 *p += dentry->d_name.len;
4617 rel->dname_seq = cpu_to_le32(di->lease_seq);
4618 __ceph_mdsc_drop_dentry_lease(dentry);
4619 }
4620 spin_unlock(&dentry->d_lock);
4621 return ret;
4622 }
4623
remove_capsnaps(struct ceph_mds_client * mdsc,struct inode * inode)4624 static int remove_capsnaps(struct ceph_mds_client *mdsc, struct inode *inode)
4625 {
4626 struct ceph_inode_info *ci = ceph_inode(inode);
4627 struct ceph_cap_snap *capsnap;
4628 int capsnap_release = 0;
4629
4630 lockdep_assert_held(&ci->i_ceph_lock);
4631
4632 dout("removing capsnaps, ci is %p, inode is %p\n", ci, inode);
4633
4634 while (!list_empty(&ci->i_cap_snaps)) {
4635 capsnap = list_first_entry(&ci->i_cap_snaps,
4636 struct ceph_cap_snap, ci_item);
4637 __ceph_remove_capsnap(inode, capsnap, NULL, NULL);
4638 ceph_put_snap_context(capsnap->context);
4639 ceph_put_cap_snap(capsnap);
4640 capsnap_release++;
4641 }
4642 wake_up_all(&ci->i_cap_wq);
4643 wake_up_all(&mdsc->cap_flushing_wq);
4644 return capsnap_release;
4645 }
4646
ceph_purge_inode_cap(struct inode * inode,struct ceph_cap * cap,bool * invalidate)4647 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate)
4648 {
4649 struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
4650 struct ceph_mds_client *mdsc = fsc->mdsc;
4651 struct ceph_inode_info *ci = ceph_inode(inode);
4652 bool is_auth;
4653 bool dirty_dropped = false;
4654 int iputs = 0;
4655
4656 lockdep_assert_held(&ci->i_ceph_lock);
4657
4658 dout("removing cap %p, ci is %p, inode is %p\n",
4659 cap, ci, &ci->netfs.inode);
4660
4661 is_auth = (cap == ci->i_auth_cap);
4662 __ceph_remove_cap(cap, false);
4663 if (is_auth) {
4664 struct ceph_cap_flush *cf;
4665
4666 if (ceph_inode_is_shutdown(inode)) {
4667 if (inode->i_data.nrpages > 0)
4668 *invalidate = true;
4669 if (ci->i_wrbuffer_ref > 0)
4670 mapping_set_error(&inode->i_data, -EIO);
4671 }
4672
4673 spin_lock(&mdsc->cap_dirty_lock);
4674
4675 /* trash all of the cap flushes for this inode */
4676 while (!list_empty(&ci->i_cap_flush_list)) {
4677 cf = list_first_entry(&ci->i_cap_flush_list,
4678 struct ceph_cap_flush, i_list);
4679 list_del_init(&cf->g_list);
4680 list_del_init(&cf->i_list);
4681 if (!cf->is_capsnap)
4682 ceph_free_cap_flush(cf);
4683 }
4684
4685 if (!list_empty(&ci->i_dirty_item)) {
4686 pr_warn_ratelimited(
4687 " dropping dirty %s state for %p %lld\n",
4688 ceph_cap_string(ci->i_dirty_caps),
4689 inode, ceph_ino(inode));
4690 ci->i_dirty_caps = 0;
4691 list_del_init(&ci->i_dirty_item);
4692 dirty_dropped = true;
4693 }
4694 if (!list_empty(&ci->i_flushing_item)) {
4695 pr_warn_ratelimited(
4696 " dropping dirty+flushing %s state for %p %lld\n",
4697 ceph_cap_string(ci->i_flushing_caps),
4698 inode, ceph_ino(inode));
4699 ci->i_flushing_caps = 0;
4700 list_del_init(&ci->i_flushing_item);
4701 mdsc->num_cap_flushing--;
4702 dirty_dropped = true;
4703 }
4704 spin_unlock(&mdsc->cap_dirty_lock);
4705
4706 if (dirty_dropped) {
4707 mapping_set_error(inode->i_mapping, -EIO);
4708
4709 if (ci->i_wrbuffer_ref_head == 0 &&
4710 ci->i_wr_ref == 0 &&
4711 ci->i_dirty_caps == 0 &&
4712 ci->i_flushing_caps == 0) {
4713 ceph_put_snap_context(ci->i_head_snapc);
4714 ci->i_head_snapc = NULL;
4715 }
4716 }
4717
4718 if (atomic_read(&ci->i_filelock_ref) > 0) {
4719 /* make further file lock syscall return -EIO */
4720 ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
4721 pr_warn_ratelimited(" dropping file locks for %p %lld\n",
4722 inode, ceph_ino(inode));
4723 }
4724
4725 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
4726 cf = ci->i_prealloc_cap_flush;
4727 ci->i_prealloc_cap_flush = NULL;
4728 if (!cf->is_capsnap)
4729 ceph_free_cap_flush(cf);
4730 }
4731
4732 if (!list_empty(&ci->i_cap_snaps))
4733 iputs = remove_capsnaps(mdsc, inode);
4734 }
4735 if (dirty_dropped)
4736 ++iputs;
4737 return iputs;
4738 }
4739