caps.c 126 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
#include <linux/ceph/ceph_debug.h>
Sage Weil's avatar
Sage Weil committed
3 4 5

#include <linux/fs.h>
#include <linux/kernel.h>
6
#include <linux/sched/signal.h>
7
#include <linux/slab.h>
Sage Weil's avatar
Sage Weil committed
8 9
#include <linux/vmalloc.h>
#include <linux/wait.h>
10
#include <linux/writeback.h>
11
#include <linux/iversion.h>
Sage Weil's avatar
Sage Weil committed
12 13

#include "super.h"
14
#include "mds_client.h"
15
#include "cache.h"
16 17
#include <linux/ceph/decode.h>
#include <linux/ceph/messenger.h>
Sage Weil's avatar
Sage Weil committed
18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

/*
 * Capability management
 *
 * The Ceph metadata servers control client access to inode metadata
 * and file data by issuing capabilities, granting clients permission
 * to read and/or write both inode field and file data to OSDs
 * (storage nodes).  Each capability consists of a set of bits
 * indicating which operations are allowed.
 *
 * If the client holds a *_SHARED cap, the client has a coherent value
 * that can be safely read from the cached inode.
 *
 * In the case of a *_EXCL (exclusive) or FILE_WR capabilities, the
 * client is allowed to change inode attributes (e.g., file size,
 * mtime), note its dirty state in the ceph_cap, and asynchronously
 * flush that metadata change to the MDS.
 *
 * In the event of a conflicting operation (perhaps by another
 * client), the MDS will revoke the conflicting client capabilities.
 *
 * In order for a client to cache an inode, it must hold a capability
 * with at least one MDS server.  When inodes are released, release
 * notifications are batched and periodically sent en masse to the MDS
 * cluster to release server state.
 */

45
static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc);
46 47 48 49
static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
				 struct ceph_mds_session *session,
				 struct ceph_inode_info *ci,
				 u64 oldest_flush_tid);
Sage Weil's avatar
Sage Weil committed
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

/*
 * Generate readable cap strings for debugging output.
 */
#define MAX_CAP_STR 20
static char cap_str[MAX_CAP_STR][40];
static DEFINE_SPINLOCK(cap_str_lock);
static int last_cap_str;

static char *gcap_string(char *s, int c)
{
	if (c & CEPH_CAP_GSHARED)
		*s++ = 's';
	if (c & CEPH_CAP_GEXCL)
		*s++ = 'x';
	if (c & CEPH_CAP_GCACHE)
		*s++ = 'c';
	if (c & CEPH_CAP_GRD)
		*s++ = 'r';
	if (c & CEPH_CAP_GWR)
		*s++ = 'w';
	if (c & CEPH_CAP_GBUFFER)
		*s++ = 'b';
73 74
	if (c & CEPH_CAP_GWREXTEND)
		*s++ = 'a';
Sage Weil's avatar
Sage Weil committed
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
	if (c & CEPH_CAP_GLAZYIO)
		*s++ = 'l';
	return s;
}

const char *ceph_cap_string(int caps)
{
	int i;
	char *s;
	int c;

	spin_lock(&cap_str_lock);
	i = last_cap_str++;
	if (last_cap_str == MAX_CAP_STR)
		last_cap_str = 0;
	spin_unlock(&cap_str_lock);

	s = cap_str[i];

	if (caps & CEPH_CAP_PIN)
		*s++ = 'p';

	c = (caps >> CEPH_CAP_SAUTH) & 3;
	if (c) {
		*s++ = 'A';
		s = gcap_string(s, c);
	}

	c = (caps >> CEPH_CAP_SLINK) & 3;
	if (c) {
		*s++ = 'L';
		s = gcap_string(s, c);
	}

	c = (caps >> CEPH_CAP_SXATTR) & 3;
	if (c) {
		*s++ = 'X';
		s = gcap_string(s, c);
	}

	c = caps >> CEPH_CAP_SFILE;
	if (c) {
		*s++ = 'F';
		s = gcap_string(s, c);
	}

	if (s == cap_str[i])
		*s++ = '-';
	*s = 0;
	return cap_str[i];
}

127
void ceph_caps_init(struct ceph_mds_client *mdsc)
Sage Weil's avatar
Sage Weil committed
128
{
129 130
	INIT_LIST_HEAD(&mdsc->caps_list);
	spin_lock_init(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
131 132
}

133
void ceph_caps_finalize(struct ceph_mds_client *mdsc)
Sage Weil's avatar
Sage Weil committed
134 135 136
{
	struct ceph_cap *cap;

137 138 139 140
	spin_lock(&mdsc->caps_list_lock);
	while (!list_empty(&mdsc->caps_list)) {
		cap = list_first_entry(&mdsc->caps_list,
				       struct ceph_cap, caps_item);
Sage Weil's avatar
Sage Weil committed
141 142 143
		list_del(&cap->caps_item);
		kmem_cache_free(ceph_cap_cachep, cap);
	}
144 145 146 147 148 149
	mdsc->caps_total_count = 0;
	mdsc->caps_avail_count = 0;
	mdsc->caps_use_count = 0;
	mdsc->caps_reserve_count = 0;
	mdsc->caps_min_count = 0;
	spin_unlock(&mdsc->caps_list_lock);
150 151
}

152 153
void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
			      struct ceph_mount_options *fsopt)
154
{
155
	spin_lock(&mdsc->caps_list_lock);
156 157 158 159 160 161 162
	mdsc->caps_min_count = fsopt->max_readdir;
	if (mdsc->caps_min_count < 1024)
		mdsc->caps_min_count = 1024;
	mdsc->caps_use_max = fsopt->caps_max;
	if (mdsc->caps_use_max > 0 &&
	    mdsc->caps_use_max < mdsc->caps_min_count)
		mdsc->caps_use_max = mdsc->caps_min_count;
163
	spin_unlock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
164 165
}

166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196
static void __ceph_unreserve_caps(struct ceph_mds_client *mdsc, int nr_caps)
{
	struct ceph_cap *cap;
	int i;

	if (nr_caps) {
		BUG_ON(mdsc->caps_reserve_count < nr_caps);
		mdsc->caps_reserve_count -= nr_caps;
		if (mdsc->caps_avail_count >=
		    mdsc->caps_reserve_count + mdsc->caps_min_count) {
			mdsc->caps_total_count -= nr_caps;
			for (i = 0; i < nr_caps; i++) {
				cap = list_first_entry(&mdsc->caps_list,
					struct ceph_cap, caps_item);
				list_del(&cap->caps_item);
				kmem_cache_free(ceph_cap_cachep, cap);
			}
		} else {
			mdsc->caps_avail_count += nr_caps;
		}

		dout("%s: caps %d = %d used + %d resv + %d avail\n",
		     __func__,
		     mdsc->caps_total_count, mdsc->caps_use_count,
		     mdsc->caps_reserve_count, mdsc->caps_avail_count);
		BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
						 mdsc->caps_reserve_count +
						 mdsc->caps_avail_count);
	}
}

197 198 199 200
/*
 * Called under mdsc->mutex.
 */
int ceph_reserve_caps(struct ceph_mds_client *mdsc,
201
		      struct ceph_cap_reservation *ctx, int need)
Sage Weil's avatar
Sage Weil committed
202
{
203
	int i, j;
Sage Weil's avatar
Sage Weil committed
204 205 206
	struct ceph_cap *cap;
	int have;
	int alloc = 0;
207
	int max_caps;
208
	int err = 0;
209 210
	bool trimmed = false;
	struct ceph_mds_session *s;
Sage Weil's avatar
Sage Weil committed
211 212 213 214 215
	LIST_HEAD(newcaps);

	dout("reserve caps ctx=%p need=%d\n", ctx, need);

	/* first reserve any caps that are already allocated */
216 217
	spin_lock(&mdsc->caps_list_lock);
	if (mdsc->caps_avail_count >= need)
Sage Weil's avatar
Sage Weil committed
218 219
		have = need;
	else
220 221 222 223 224 225 226
		have = mdsc->caps_avail_count;
	mdsc->caps_avail_count -= have;
	mdsc->caps_reserve_count += have;
	BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
					 mdsc->caps_reserve_count +
					 mdsc->caps_avail_count);
	spin_unlock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
227

228
	for (i = have; i < need; ) {
Sage Weil's avatar
Sage Weil committed
229
		cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
		if (cap) {
			list_add(&cap->caps_item, &newcaps);
			alloc++;
			i++;
			continue;
		}

		if (!trimmed) {
			for (j = 0; j < mdsc->max_sessions; j++) {
				s = __ceph_lookup_mds_session(mdsc, j);
				if (!s)
					continue;
				mutex_unlock(&mdsc->mutex);

				mutex_lock(&s->s_mutex);
				max_caps = s->s_nr_caps - (need - i);
				ceph_trim_caps(mdsc, s, max_caps);
				mutex_unlock(&s->s_mutex);

				ceph_put_mds_session(s);
				mutex_lock(&mdsc->mutex);
251
			}
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270
			trimmed = true;

			spin_lock(&mdsc->caps_list_lock);
			if (mdsc->caps_avail_count) {
				int more_have;
				if (mdsc->caps_avail_count >= need - i)
					more_have = need - i;
				else
					more_have = mdsc->caps_avail_count;

				i += more_have;
				have += more_have;
				mdsc->caps_avail_count -= more_have;
				mdsc->caps_reserve_count += more_have;

			}
			spin_unlock(&mdsc->caps_list_lock);

			continue;
271
		}
272 273 274

		pr_warn("reserve caps ctx=%p ENOMEM need=%d got=%d\n",
			ctx, need, have + alloc);
275 276 277 278 279 280 281
		err = -ENOMEM;
		break;
	}

	if (!err) {
		BUG_ON(have + alloc != need);
		ctx->count = need;
282
		ctx->used = 0;
Sage Weil's avatar
Sage Weil committed
283 284
	}

285 286 287 288
	spin_lock(&mdsc->caps_list_lock);
	mdsc->caps_total_count += alloc;
	mdsc->caps_reserve_count += alloc;
	list_splice(&newcaps, &mdsc->caps_list);
Sage Weil's avatar
Sage Weil committed
289

290 291 292
	BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
					 mdsc->caps_reserve_count +
					 mdsc->caps_avail_count);
293 294 295 296

	if (err)
		__ceph_unreserve_caps(mdsc, have + alloc);

297
	spin_unlock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
298 299

	dout("reserve caps ctx=%p %d = %d used + %d resv + %d avail\n",
300 301
	     ctx, mdsc->caps_total_count, mdsc->caps_use_count,
	     mdsc->caps_reserve_count, mdsc->caps_avail_count);
302
	return err;
Sage Weil's avatar
Sage Weil committed
303 304
}

305
void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
306
			 struct ceph_cap_reservation *ctx)
Sage Weil's avatar
Sage Weil committed
307
{
308 309 310 311
	bool reclaim = false;
	if (!ctx->count)
		return;

Sage Weil's avatar
Sage Weil committed
312
	dout("unreserve caps ctx=%p count=%d\n", ctx, ctx->count);
313 314 315
	spin_lock(&mdsc->caps_list_lock);
	__ceph_unreserve_caps(mdsc, ctx->count);
	ctx->count = 0;
316 317 318 319

	if (mdsc->caps_use_max > 0 &&
	    mdsc->caps_use_count > mdsc->caps_use_max)
		reclaim = true;
320
	spin_unlock(&mdsc->caps_list_lock);
321 322 323

	if (reclaim)
		ceph_reclaim_caps_nr(mdsc, ctx->used);
Sage Weil's avatar
Sage Weil committed
324 325
}

326 327
struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
			      struct ceph_cap_reservation *ctx)
Sage Weil's avatar
Sage Weil committed
328 329 330 331
{
	struct ceph_cap *cap = NULL;

	/* temporary, until we do something about cap import/export */
332 333 334
	if (!ctx) {
		cap = kmem_cache_alloc(ceph_cap_cachep, GFP_NOFS);
		if (cap) {
335
			spin_lock(&mdsc->caps_list_lock);
336 337
			mdsc->caps_use_count++;
			mdsc->caps_total_count++;
338
			spin_unlock(&mdsc->caps_list_lock);
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
		} else {
			spin_lock(&mdsc->caps_list_lock);
			if (mdsc->caps_avail_count) {
				BUG_ON(list_empty(&mdsc->caps_list));

				mdsc->caps_avail_count--;
				mdsc->caps_use_count++;
				cap = list_first_entry(&mdsc->caps_list,
						struct ceph_cap, caps_item);
				list_del(&cap->caps_item);

				BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
				       mdsc->caps_reserve_count + mdsc->caps_avail_count);
			}
			spin_unlock(&mdsc->caps_list_lock);
354
		}
355

356 357
		return cap;
	}
Sage Weil's avatar
Sage Weil committed
358

359
	spin_lock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
360
	dout("get_cap ctx=%p (%d) %d = %d used + %d resv + %d avail\n",
361 362
	     ctx, ctx->count, mdsc->caps_total_count, mdsc->caps_use_count,
	     mdsc->caps_reserve_count, mdsc->caps_avail_count);
Sage Weil's avatar
Sage Weil committed
363
	BUG_ON(!ctx->count);
364 365
	BUG_ON(ctx->count > mdsc->caps_reserve_count);
	BUG_ON(list_empty(&mdsc->caps_list));
Sage Weil's avatar
Sage Weil committed
366 367

	ctx->count--;
368
	ctx->used++;
369 370
	mdsc->caps_reserve_count--;
	mdsc->caps_use_count++;
Sage Weil's avatar
Sage Weil committed
371

372
	cap = list_first_entry(&mdsc->caps_list, struct ceph_cap, caps_item);
Sage Weil's avatar
Sage Weil committed
373 374
	list_del(&cap->caps_item);

375 376 377
	BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
	       mdsc->caps_reserve_count + mdsc->caps_avail_count);
	spin_unlock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
378 379 380
	return cap;
}

381
void ceph_put_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap)
Sage Weil's avatar
Sage Weil committed
382
{
383
	spin_lock(&mdsc->caps_list_lock);
384
	dout("put_cap %p %d = %d used + %d resv + %d avail\n",
385 386 387
	     cap, mdsc->caps_total_count, mdsc->caps_use_count,
	     mdsc->caps_reserve_count, mdsc->caps_avail_count);
	mdsc->caps_use_count--;
Sage Weil's avatar
Sage Weil committed
388
	/*
389 390
	 * Keep some preallocated caps around (ceph_min_count), to
	 * avoid lots of free/alloc churn.
Sage Weil's avatar
Sage Weil committed
391
	 */
392 393 394
	if (mdsc->caps_avail_count >= mdsc->caps_reserve_count +
				      mdsc->caps_min_count) {
		mdsc->caps_total_count--;
Sage Weil's avatar
Sage Weil committed
395 396
		kmem_cache_free(ceph_cap_cachep, cap);
	} else {
397 398
		mdsc->caps_avail_count++;
		list_add(&cap->caps_item, &mdsc->caps_list);
Sage Weil's avatar
Sage Weil committed
399 400
	}

401 402 403
	BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
	       mdsc->caps_reserve_count + mdsc->caps_avail_count);
	spin_unlock(&mdsc->caps_list_lock);
Sage Weil's avatar
Sage Weil committed
404 405
}

406
void ceph_reservation_status(struct ceph_fs_client *fsc,
407 408
			     int *total, int *avail, int *used, int *reserved,
			     int *min)
Sage Weil's avatar
Sage Weil committed
409
{
410
	struct ceph_mds_client *mdsc = fsc->mdsc;
411

412 413
	spin_lock(&mdsc->caps_list_lock);

Sage Weil's avatar
Sage Weil committed
414
	if (total)
415
		*total = mdsc->caps_total_count;
Sage Weil's avatar
Sage Weil committed
416
	if (avail)
417
		*avail = mdsc->caps_avail_count;
Sage Weil's avatar
Sage Weil committed
418
	if (used)
419
		*used = mdsc->caps_use_count;
Sage Weil's avatar
Sage Weil committed
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);
Sage Weil's avatar
Sage Weil committed
426 427 428 429 430
}

/*
 * Find ceph_cap for given mds, if any.
 *
431
 * Called with i_ceph_lock held.
Sage Weil's avatar
Sage Weil committed
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449
 */
static struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, int mds)
{
	struct ceph_cap *cap;
	struct rb_node *n = ci->i_caps.rb_node;

	while (n) {
		cap = rb_entry(n, struct ceph_cap, ci_node);
		if (mds < cap->mds)
			n = n->rb_left;
		else if (mds > cap->mds)
			n = n->rb_right;
		else
			return cap;
	}
	return NULL;
}

450 451 452 453
struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, int mds)
{
	struct ceph_cap *cap;

454
	spin_lock(&ci->i_ceph_lock);
455
	cap = __get_cap_for_mds(ci, mds);
456
	spin_unlock(&ci->i_ceph_lock);
457 458 459
	return cap;
}

Sage Weil's avatar
Sage Weil committed
460
/*
461
 * Called under i_ceph_lock.
Sage Weil's avatar
Sage Weil committed
462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491
 */
static void __insert_cap_node(struct ceph_inode_info *ci,
			      struct ceph_cap *new)
{
	struct rb_node **p = &ci->i_caps.rb_node;
	struct rb_node *parent = NULL;
	struct ceph_cap *cap = NULL;

	while (*p) {
		parent = *p;
		cap = rb_entry(parent, struct ceph_cap, ci_node);
		if (new->mds < cap->mds)
			p = &(*p)->rb_left;
		else if (new->mds > cap->mds)
			p = &(*p)->rb_right;
		else
			BUG();
	}

	rb_link_node(&new->ci_node, parent, p);
	rb_insert_color(&new->ci_node, &ci->i_caps);
}

/*
 * (re)set cap hold timeouts, which control the delayed release
 * of unused caps back to the MDS.  Should be called on cap use.
 */
static void __cap_set_timeouts(struct ceph_mds_client *mdsc,
			       struct ceph_inode_info *ci)
{
492
	struct ceph_mount_options *opt = mdsc->fsc->mount_options;
Sage Weil's avatar
Sage Weil committed
493
	ci->i_hold_caps_max = round_jiffies(jiffies +
494
					    opt->caps_wanted_delay_max * HZ);
495 496
	dout("__cap_set_timeouts %p %lu\n", &ci->vfs_inode,
	     ci->i_hold_caps_max - jiffies);
Sage Weil's avatar
Sage Weil committed
497 498 499 500 501 502 503
}

/*
 * (Re)queue cap at the end of the delayed cap release list.
 *
 * If I_FLUSH is set, leave the inode at the front of the list.
 *
504
 * Caller holds i_ceph_lock
Sage Weil's avatar
Sage Weil committed
505 506 507
 *    -> we take mdsc->cap_delay_lock
 */
static void __cap_delay_requeue(struct ceph_mds_client *mdsc,
508
				struct ceph_inode_info *ci)
Sage Weil's avatar
Sage Weil committed
509
{
510
	dout("__cap_delay_requeue %p flags 0x%lx at %lu\n", &ci->vfs_inode,
Sage Weil's avatar
Sage Weil committed
511 512 513 514 515 516 517 518
	     ci->i_ceph_flags, ci->i_hold_caps_max);
	if (!mdsc->stopping) {
		spin_lock(&mdsc->cap_delay_lock);
		if (!list_empty(&ci->i_cap_delay_list)) {
			if (ci->i_ceph_flags & CEPH_I_FLUSH)
				goto no_change;
			list_del_init(&ci->i_cap_delay_list);
		}
519
		__cap_set_timeouts(mdsc, ci);
Sage Weil's avatar
Sage Weil committed
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
		list_add_tail(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
no_change:
		spin_unlock(&mdsc->cap_delay_lock);
	}
}

/*
 * Queue an inode for immediate writeback.  Mark inode with I_FLUSH,
 * indicating we should send a cap message to flush dirty metadata
 * asap, and move to the front of the delayed cap list.
 */
static void __cap_delay_requeue_front(struct ceph_mds_client *mdsc,
				      struct ceph_inode_info *ci)
{
	dout("__cap_delay_requeue_front %p\n", &ci->vfs_inode);
	spin_lock(&mdsc->cap_delay_lock);
	ci->i_ceph_flags |= CEPH_I_FLUSH;
	if (!list_empty(&ci->i_cap_delay_list))
		list_del_init(&ci->i_cap_delay_list);
	list_add(&ci->i_cap_delay_list, &mdsc->cap_delay_list);
	spin_unlock(&mdsc->cap_delay_lock);
}

/*
 * Cancel delayed work on cap.
 *
546
 * Caller must hold i_ceph_lock.
Sage Weil's avatar
Sage Weil committed
547 548 549 550 551 552 553 554 555 556 557 558
 */
static void __cap_delay_cancel(struct ceph_mds_client *mdsc,
			       struct ceph_inode_info *ci)
{
	dout("__cap_delay_cancel %p\n", &ci->vfs_inode);
	if (list_empty(&ci->i_cap_delay_list))
		return;
	spin_lock(&mdsc->cap_delay_lock);
	list_del_init(&ci->i_cap_delay_list);
	spin_unlock(&mdsc->cap_delay_lock);
}

559
/* Common issue checks for add_cap, handle_cap_grant. */
Sage Weil's avatar
Sage Weil committed
560 561 562 563 564
static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap,
			      unsigned issued)
{
	unsigned had = __ceph_caps_issued(ci, NULL);

565 566
	lockdep_assert_held(&ci->i_ceph_lock);

Sage Weil's avatar
Sage Weil committed
567 568 569 570
	/*
	 * Each time we receive FILE_CACHE anew, we increment
	 * i_rdcache_gen.
	 */
571 572
	if (S_ISREG(ci->vfs_inode.i_mode) &&
	    (issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
573
	    (had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) {
Sage Weil's avatar
Sage Weil committed
574
		ci->i_rdcache_gen++;
575
	}
Sage Weil's avatar
Sage Weil committed
576 577

	/*
578 579 580 581
	 * If FILE_SHARED is newly issued, mark dir not complete. We don't
	 * know what happened to this directory while we didn't have the cap.
	 * If FILE_SHARED is being revoked, also mark dir not complete. It
	 * stops on-going cached readdir.
Sage Weil's avatar
Sage Weil committed
582
	 */
583 584
	if ((issued & CEPH_CAP_FILE_SHARED) != (had & CEPH_CAP_FILE_SHARED)) {
		if (issued & CEPH_CAP_FILE_SHARED)
585
			atomic_inc(&ci->i_shared_gen);
586 587
		if (S_ISDIR(ci->vfs_inode.i_mode)) {
			dout(" marking %p NOT complete\n", &ci->vfs_inode);
588
			__ceph_dir_clear_complete(ci);
589
		}
Sage Weil's avatar
Sage Weil committed
590
	}
591 592 593 594 595 596 597

	/* Wipe saved layout if we're losing DIR_CREATE caps */
	if (S_ISDIR(ci->vfs_inode.i_mode) && (had & CEPH_CAP_DIR_CREATE) &&
		!(issued & CEPH_CAP_DIR_CREATE)) {
	     ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
	     memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
	}
Sage Weil's avatar
Sage Weil committed
598 599
}

600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
/**
 * change_auth_cap_ses - move inode to appropriate lists when auth caps change
 * @ci: inode to be moved
 * @session: new auth caps session
 */
static void change_auth_cap_ses(struct ceph_inode_info *ci,
				struct ceph_mds_session *session)
{
	lockdep_assert_held(&ci->i_ceph_lock);

	if (list_empty(&ci->i_dirty_item) && list_empty(&ci->i_flushing_item))
		return;

	spin_lock(&session->s_mdsc->cap_dirty_lock);
	if (!list_empty(&ci->i_dirty_item))
		list_move(&ci->i_dirty_item, &session->s_cap_dirty);
	if (!list_empty(&ci->i_flushing_item))
		list_move_tail(&ci->i_flushing_item, &session->s_cap_flushing);
	spin_unlock(&session->s_mdsc->cap_dirty_lock);
}

Sage Weil's avatar
Sage Weil committed
621 622 623
/*
 * Add a capability under the given MDS session.
 *
624
 * Caller should hold session snap_rwsem (read) and ci->i_ceph_lock
Sage Weil's avatar
Sage Weil committed
625 626 627 628 629
 *
 * @fmode is the open file mode, if we are opening a file, otherwise
 * it is < 0.  (This is so we can atomically add the cap and add an
 * open file reference to it.)
 */
630 631
void ceph_add_cap(struct inode *inode,
		  struct ceph_mds_session *session, u64 cap_id,
632
		  unsigned issued, unsigned wanted,
633 634
		  unsigned seq, unsigned mseq, u64 realmino, int flags,
		  struct ceph_cap **new_cap)
Sage Weil's avatar
Sage Weil committed
635
{
636
	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
Sage Weil's avatar
Sage Weil committed
637 638 639 640
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_cap *cap;
	int mds = session->s_mds;
	int actual_wanted;
641
	u32 gen;
Sage Weil's avatar
Sage Weil committed
642

643 644
	lockdep_assert_held(&ci->i_ceph_lock);

Sage Weil's avatar
Sage Weil committed
645 646 647
	dout("add_cap %p mds%d cap %llx %s seq %d\n", inode,
	     session->s_mds, cap_id, ceph_cap_string(issued), seq);

648
	gen = atomic_read(&session->s_cap_gen);
649

Sage Weil's avatar
Sage Weil committed
650 651
	cap = __get_cap_for_mds(ci, mds);
	if (!cap) {
652 653
		cap = *new_cap;
		*new_cap = NULL;
Sage Weil's avatar
Sage Weil committed
654 655 656 657 658

		cap->issued = 0;
		cap->implemented = 0;
		cap->mds = mds;
		cap->mds_wanted = 0;
659
		cap->mseq = 0;
Sage Weil's avatar
Sage Weil committed
660 661 662 663 664 665 666 667 668

		cap->ci = ci;
		__insert_cap_node(ci, cap);

		/* add to session cap list */
		cap->session = session;
		spin_lock(&session->s_cap_lock);
		list_add_tail(&cap->session_caps, &session->s_caps);
		session->s_nr_caps++;
669
		atomic64_inc(&mdsc->metric.total_caps);
Sage Weil's avatar
Sage Weil committed
670
		spin_unlock(&session->s_cap_lock);
671
	} else {
672 673 674 675
		spin_lock(&session->s_cap_lock);
		list_move_tail(&cap->session_caps, &session->s_caps);
		spin_unlock(&session->s_cap_lock);

676
		if (cap->cap_gen < gen)
677 678
			cap->issued = cap->implemented = CEPH_CAP_PIN;

679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
		/*
		 * auth mds of the inode changed. we received the cap export
		 * message, but still haven't received the cap import message.
		 * handle_cap_export() updated the new auth MDS' cap.
		 *
		 * "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing
		 * a message that was send before the cap import message. So
		 * don't remove caps.
		 */
		if (ceph_seq_cmp(seq, cap->seq) <= 0) {
			WARN_ON(cap != ci->i_auth_cap);
			WARN_ON(cap->cap_id != cap_id);
			seq = cap->seq;
			mseq = cap->mseq;
			issued |= cap->issued;
			flags |= CEPH_CAP_FLAG_AUTH;
		}
	}
Sage Weil's avatar
Sage Weil committed
697

698 699 700
	if (!ci->i_snap_realm ||
	    ((flags & CEPH_CAP_FLAG_AUTH) &&
	     realmino != (u64)-1 && ci->i_snap_realm->ino != realmino)) {
Sage Weil's avatar
Sage Weil committed
701 702 703 704 705
		/*
		 * add this inode to the appropriate snap realm
		 */
		struct ceph_snap_realm *realm = ceph_lookup_snap_realm(mdsc,
							       realmino);
706
		if (realm)
707
			ceph_change_snap_realm(inode, realm);
708 709 710 711
		else
			WARN(1, "%s: couldn't find snap realm 0x%llx (ino 0x%llx oldrealm 0x%llx)\n",
			     __func__, realmino, ci->i_vino.ino,
			     ci->i_snap_realm ? ci->i_snap_realm->ino : 0);
Sage Weil's avatar
Sage Weil committed
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726
	}

	__check_cap_issue(ci, cap, issued);

	/*
	 * If we are issued caps we don't want, or the mds' wanted
	 * value appears to be off, queue a check so we'll release
	 * later and/or update the mds wanted value.
	 */
	actual_wanted = __ceph_caps_wanted(ci);
	if ((wanted & ~actual_wanted) ||
	    (issued & ~actual_wanted & CEPH_CAP_ANY_WR)) {
		dout(" issued %s, mds wanted %s, actual %s, queueing\n",
		     ceph_cap_string(issued), ceph_cap_string(wanted),
		     ceph_cap_string(actual_wanted));
727
		__cap_delay_requeue(mdsc, ci);
Sage Weil's avatar
Sage Weil committed
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) {
733 734 735
			if (ci->i_auth_cap &&
			    ci->i_auth_cap->session != cap->session)
				change_auth_cap_ses(ci, cap->session);
736
			ci->i_auth_cap = cap;
737 738
			cap->mds_wanted = wanted;
		}
739 740
	} else {
		WARN_ON(ci->i_auth_cap == cap);
741
	}
Sage Weil's avatar
Sage Weil committed
742 743 744 745 746 747 748

	dout("add_cap inode %p (%llx.%llx) cap %p %s now %s seq %d mds%d\n",
	     inode, ceph_vinop(inode), cap, ceph_cap_string(issued),
	     ceph_cap_string(issued|cap->issued), seq, mds);
	cap->cap_id = cap_id;
	cap->issued = issued;
	cap->implemented |= issued;
749
	if (ceph_seq_cmp(mseq, cap->mseq) > 0)
750 751 752
		cap->mds_wanted = wanted;
	else
		cap->mds_wanted |= wanted;
Sage Weil's avatar
Sage Weil committed
753 754 755
	cap->seq = seq;
	cap->issue_seq = seq;
	cap->mseq = mseq;
756
	cap->cap_gen = gen;
Sage Weil's avatar
Sage Weil committed
757 758 759 760 761 762 763 764 765 766
}

/*
 * Return true if cap has not timed out and belongs to the current
 * generation of the MDS session (i.e. has not gone 'stale' due to
 * us losing touch with the mds).
 */
static int __cap_is_valid(struct ceph_cap *cap)
{
	unsigned long ttl;
Sage Weil's avatar
Sage Weil committed
767
	u32 gen;
Sage Weil's avatar
Sage Weil committed
768

769
	gen = atomic_read(&cap->session->s_cap_gen);
Sage Weil's avatar
Sage Weil committed
770 771
	ttl = cap->session->s_cap_ttl;

772
	if (cap->cap_gen < gen || time_after_eq(jiffies, ttl)) {
Sage Weil's avatar
Sage Weil committed
773 774
		dout("__cap_is_valid %p cap %p issued %s "
		     "but STALE (gen %u vs %u)\n", &cap->ci->vfs_inode,
775
		     cap, ceph_cap_string(cap->issued), cap->cap_gen, gen);
Sage Weil's avatar
Sage Weil committed
776 777 778 779 780 781 782 783 784 785 786 787 788
		return 0;
	}

	return 1;
}

/*
 * Return set of valid cap bits issued to us.  Note that caps time
 * out, and may be invalidated in bulk if the client session times out
 * and session->s_cap_gen is bumped.
 */
int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented)
{
789
	int have = ci->i_snap_caps;
Sage Weil's avatar
Sage Weil committed
790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
	struct ceph_cap *cap;
	struct rb_node *p;

	if (implemented)
		*implemented = 0;
	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
		cap = rb_entry(p, struct ceph_cap, ci_node);
		if (!__cap_is_valid(cap))
			continue;
		dout("__ceph_caps_issued %p cap %p issued %s\n",
		     &ci->vfs_inode, cap, ceph_cap_string(cap->issued));
		have |= cap->issued;
		if (implemented)
			*implemented |= cap->implemented;
	}
Yan, Zheng's avatar
Yan, Zheng committed
805 806 807 808 809 810 811 812 813
	/*
	 * exclude caps issued by non-auth MDS, but are been revoking
	 * by the auth MDS. The non-auth MDS should be revoking/exporting
	 * these caps, but the message is delayed.
	 */
	if (ci->i_auth_cap) {
		cap = ci->i_auth_cap;
		have &= ~cap->implemented | cap->issued;
	}
Sage Weil's avatar
Sage Weil committed
814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
	return have;
}

/*
 * Get cap bits issued by caps other than @ocap
 */
int __ceph_caps_issued_other(struct ceph_inode_info *ci, struct ceph_cap *ocap)
{
	int have = ci->i_snap_caps;
	struct ceph_cap *cap;
	struct rb_node *p;

	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
		cap = rb_entry(p, struct ceph_cap, ci_node);
		if (cap == ocap)
			continue;
		if (!__cap_is_valid(cap))
			continue;
		have |= cap->issued;
	}
	return have;
}

/*
 * Move a cap to the end of the LRU (oldest caps at list head, newest
 * at list tail).
 */
static void __touch_cap(struct ceph_cap *cap)
{
	struct ceph_mds_session *s = cap->session;

	spin_lock(&s->s_cap_lock);
846
	if (!s->s_cap_iterator) {
847 848 849 850 851 852 853
		dout("__touch_cap %p cap %p mds%d\n", &cap->ci->vfs_inode, cap,
		     s->s_mds);
		list_move_tail(&cap->session_caps, &s->s_caps);
	} else {
		dout("__touch_cap %p cap %p mds%d NOP, iterating over caps\n",
		     &cap->ci->vfs_inode, cap, s->s_mds);
	}
Sage Weil's avatar
Sage Weil committed
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
	spin_unlock(&s->s_cap_lock);
}

/*
 * Check if we hold the given mask.  If so, move the cap(s) to the
 * front of their respective LRUs.  (This is the preferred way for
 * callers to check for caps they want.)
 */
int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int touch)
{
	struct ceph_cap *cap;
	struct rb_node *p;
	int have = ci->i_snap_caps;

	if ((have & mask) == mask) {
869 870
		dout("__ceph_caps_issued_mask ino 0x%llx snap issued %s"
		     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
Sage Weil's avatar
Sage Weil committed
871 872 873 874 875 876 877 878 879 880
		     ceph_cap_string(have),
		     ceph_cap_string(mask));
		return 1;
	}

	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
		cap = rb_entry(p, struct ceph_cap, ci_node);
		if (!__cap_is_valid(cap))
			continue;
		if ((cap->issued & mask) == mask) {
881 882
			dout("__ceph_caps_issued_mask ino 0x%llx cap %p issued %s"
			     " (mask %s)\n", ceph_ino(&ci->vfs_inode), cap,
Sage Weil's avatar
Sage Weil committed
883 884 885 886 887 888 889 890 891 892
			     ceph_cap_string(cap->issued),
			     ceph_cap_string(mask));
			if (touch)
				__touch_cap(cap);
			return 1;
		}

		/* does a combination of caps satisfy mask? */
		have |= cap->issued;
		if ((have & mask) == mask) {
893 894
			dout("__ceph_caps_issued_mask ino 0x%llx combo issued %s"
			     " (mask %s)\n", ceph_ino(&ci->vfs_inode),
Sage Weil's avatar
Sage Weil committed
895 896 897 898 899
			     ceph_cap_string(cap->issued),
			     ceph_cap_string(mask));
			if (touch) {
				struct rb_node *q;

Lucas De Marchi's avatar
Lucas De Marchi committed
900
				/* touch this + preceding caps */
Sage Weil's avatar
Sage Weil committed
901 902 903 904 905 906 907
				__touch_cap(cap);
				for (q = rb_first(&ci->i_caps); q != p;
				     q = rb_next(q)) {
					cap = rb_entry(q, struct ceph_cap,
						       ci_node);
					if (!__cap_is_valid(cap))
						continue;
908 909
					if (cap->issued & mask)
						__touch_cap(cap);
Sage Weil's avatar
Sage Weil committed
910 911 912 913 914 915 916 917 918
				}
			}
			return 1;
		}
	}

	return 0;
}

919 920 921 922 923 924 925 926 927 928 929 930 931 932
int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
				   int touch)
{
	struct ceph_fs_client *fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
	int r;

	r = __ceph_caps_issued_mask(ci, mask, touch);
	if (r)
		ceph_update_cap_hit(&fsc->mdsc->metric);
	else
		ceph_update_cap_mis(&fsc->mdsc->metric);
	return r;
}

Sage Weil's avatar
Sage Weil committed
933 934 935
/*
 * Return true if mask caps are currently being revoked by an MDS.
 */
936 937
int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
			       struct ceph_cap *ocap, int mask)
Sage Weil's avatar
Sage Weil committed
938 939 940 941 942 943
{
	struct ceph_cap *cap;
	struct rb_node *p;

	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
		cap = rb_entry(p, struct ceph_cap, ci_node);
Yan, Zheng's avatar
Yan, Zheng committed
944
		if (cap != ocap &&
945 946
		    (cap->implemented & ~cap->issued & mask))
			return 1;
Sage Weil's avatar
Sage Weil committed
947
	}
948 949 950 951 952 953 954 955 956 957
	return 0;
}

int ceph_caps_revoking(struct ceph_inode_info *ci, int mask)
{
	struct inode *inode = &ci->vfs_inode;
	int ret;

	spin_lock(&ci->i_ceph_lock);
	ret = __ceph_caps_revoking_other(ci, NULL, mask);
958
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
959 960 961 962 963 964 965 966 967 968 969 970
	dout("ceph_caps_revoking %p %s = %d\n", inode,
	     ceph_cap_string(mask), ret);
	return ret;
}

int __ceph_caps_used(struct ceph_inode_info *ci)
{
	int used = 0;
	if (ci->i_pin_ref)
		used |= CEPH_CAP_PIN;
	if (ci->i_rd_ref)
		used |= CEPH_CAP_FILE_RD;
Yan, Zheng's avatar
Yan, Zheng committed
971
	if (ci->i_rdcache_ref ||
972
	    (S_ISREG(ci->vfs_inode.i_mode) &&
Yan, Zheng's avatar
Yan, Zheng committed
973
	     ci->vfs_inode.i_data.nrpages))
Sage Weil's avatar
Sage Weil committed
974 975 976
		used |= CEPH_CAP_FILE_CACHE;
	if (ci->i_wr_ref)
		used |= CEPH_CAP_FILE_WR;
977
	if (ci->i_wb_ref || ci->i_wrbuffer_ref)
Sage Weil's avatar
Sage Weil committed
978
		used |= CEPH_CAP_FILE_BUFFER;
979 980
	if (ci->i_fx_ref)
		used |= CEPH_CAP_FILE_EXCL;
Sage Weil's avatar
Sage Weil committed
981 982 983
	return used;
}

984 985
#define FMODE_WAIT_BIAS 1000

Sage Weil's avatar
Sage Weil committed
986 987 988 989 990
/*
 * wanted, by virtue of open file modes
 */
int __ceph_caps_file_wanted(struct ceph_inode_info *ci)
{
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
	const int PIN_SHIFT = ffs(CEPH_FILE_MODE_PIN);
	const int RD_SHIFT = ffs(CEPH_FILE_MODE_RD);
	const int WR_SHIFT = ffs(CEPH_FILE_MODE_WR);
	const int LAZY_SHIFT = ffs(CEPH_FILE_MODE_LAZY);
	struct ceph_mount_options *opt =
		ceph_inode_to_client(&ci->vfs_inode)->mount_options;
	unsigned long used_cutoff = jiffies - opt->caps_wanted_delay_max * HZ;
	unsigned long idle_cutoff = jiffies - opt->caps_wanted_delay_min * HZ;

	if (S_ISDIR(ci->vfs_inode.i_mode)) {
		int want = 0;

		/* use used_cutoff here, to keep dir's wanted caps longer */
		if (ci->i_nr_by_mode[RD_SHIFT] > 0 ||
		    time_after(ci->i_last_rd, used_cutoff))
			want |= CEPH_CAP_ANY_SHARED;

		if (ci->i_nr_by_mode[WR_SHIFT] > 0 ||
		    time_after(ci->i_last_wr, used_cutoff)) {
			want |= CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_EXCL;
			if (opt->flags & CEPH_MOUNT_OPT_ASYNC_DIROPS)
				want |= CEPH_CAP_ANY_DIR_OPS;
		}

		if (want || ci->i_nr_by_mode[PIN_SHIFT] > 0)
			want |= CEPH_CAP_PIN;

		return want;
	} else {
		int bits = 0;

		if (ci->i_nr_by_mode[RD_SHIFT] > 0) {
			if (ci->i_nr_by_mode[RD_SHIFT] >= FMODE_WAIT_BIAS ||
			    time_after(ci->i_last_rd, used_cutoff))
				bits |= 1 << RD_SHIFT;
		} else if (time_after(ci->i_last_rd, idle_cutoff)) {
			bits |= 1 << RD_SHIFT;
		}

		if (ci->i_nr_by_mode[WR_SHIFT] > 0) {
			if (ci->i_nr_by_mode[WR_SHIFT] >= FMODE_WAIT_BIAS ||
			    time_after(ci->i_last_wr, used_cutoff))
				bits |= 1 << WR_SHIFT;
		} else if (time_after(ci->i_last_wr, idle_cutoff)) {
			bits |= 1 << WR_SHIFT;
		}

		/* check lazyio only when read/write is wanted */
		if ((bits & (CEPH_FILE_MODE_RDWR << 1)) &&
		    ci->i_nr_by_mode[LAZY_SHIFT] > 0)
			bits |= 1 << LAZY_SHIFT;

		return bits ? ceph_caps_for_mode(bits >> 1) : 0;
1044
	}
Sage Weil's avatar
Sage Weil committed
1045 1046
}

1047 1048 1049 1050 1051 1052
/*
 * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 */
int __ceph_caps_wanted(struct ceph_inode_info *ci)
{
	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
1053 1054 1055 1056 1057
	if (S_ISDIR(ci->vfs_inode.i_mode)) {
		/* we want EXCL if holding caps of dir ops */
		if (w & CEPH_CAP_ANY_DIR_OPS)
			w |= CEPH_CAP_FILE_EXCL;
	} else {
1058 1059 1060 1061 1062 1063 1064
		/* we want EXCL if dirty data */
		if (w & CEPH_CAP_FILE_BUFFER)
			w |= CEPH_CAP_FILE_EXCL;
	}
	return w;
}

Sage Weil's avatar
Sage Weil committed
1065 1066 1067
/*
 * Return caps we have registered with the MDS(s) as 'wanted'.
 */
1068
int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check)
Sage Weil's avatar
Sage Weil committed
1069 1070 1071 1072 1073 1074 1075
{
	struct ceph_cap *cap;
	struct rb_node *p;
	int mds_wanted = 0;

	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
		cap = rb_entry(p, struct ceph_cap, ci_node);
1076
		if (check && !__cap_is_valid(cap))
Sage Weil's avatar
Sage Weil committed
1077
			continue;
1078 1079 1080 1081
		if (cap == ci->i_auth_cap)
			mds_wanted |= cap->mds_wanted;
		else
			mds_wanted |= (cap->mds_wanted & ~CEPH_CAP_ANY_FILE_WR);
Sage Weil's avatar
Sage Weil committed
1082 1083 1084 1085
	}
	return mds_wanted;
}

1086 1087 1088 1089 1090 1091
int ceph_is_any_caps(struct inode *inode)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int ret;

	spin_lock(&ci->i_ceph_lock);
1092
	ret = __ceph_is_any_real_caps(ci);
1093 1094 1095 1096 1097
	spin_unlock(&ci->i_ceph_lock);

	return ret;
}

Sage Weil's avatar
Sage Weil committed
1098
/*
Sage Weil's avatar
Sage Weil committed
1099 1100
 * Remove a cap.  Take steps to deal with a racing iterate_session_caps.
 *
1101
 * caller should hold i_ceph_lock.
1102
 * caller will not hold session s_mutex if called from destroy_inode.
Sage Weil's avatar
Sage Weil committed
1103
 */
1104
void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
Sage Weil's avatar
Sage Weil committed
1105 1106 1107
{
	struct ceph_mds_session *session = cap->session;
	struct ceph_inode_info *ci = cap->ci;
1108
	struct ceph_mds_client *mdsc;
Sage Weil's avatar
Sage Weil committed
1109
	int removed = 0;
Sage Weil's avatar
Sage Weil committed
1110

1111 1112 1113 1114 1115 1116
	/* 'ci' being NULL means the remove have already occurred */
	if (!ci) {
		dout("%s: cap inode is NULL\n", __func__);
		return;
	}

1117 1118
	lockdep_assert_held(&ci->i_ceph_lock);

Sage Weil's avatar
Sage Weil committed
1119 1120
	dout("__ceph_remove_cap %p from %p\n", cap, &ci->vfs_inode);

1121 1122
	mdsc = ceph_inode_to_client(&ci->vfs_inode)->mdsc;

1123 1124
	/* remove from inode's cap rbtree, and clear auth cap */
	rb_erase(&cap->ci_node, &ci->i_caps);
1125
	if (ci->i_auth_cap == cap)
1126 1127
		ci->i_auth_cap = NULL;

1128 1129 1130 1131 1132 1133 1134 1135 1136
	/* remove from session list */
	spin_lock(&session->s_cap_lock);
	if (session->s_cap_iterator == cap) {
		/* not yet, we are iterating over this very cap */
		dout("__ceph_remove_cap  delaying %p removal from session %p\n",
		     cap, cap->session);
	} else {
		list_del_init(&cap->session_caps);
		session->s_nr_caps--;
1137
		atomic64_dec(&mdsc->metric.total_caps);
1138
		cap->session = NULL;
Sage Weil's avatar
Sage Weil committed
1139
		removed = 1;
1140
	}
Sage Weil's avatar
Sage Weil committed
1141 1142
	/* protect backpointer with s_cap_lock: see iterate_session_caps */
	cap->ci = NULL;
1143 1144 1145 1146 1147 1148

	/*
	 * s_cap_reconnect is protected by s_cap_lock. no one changes
	 * s_cap_gen while session is in the reconnect state.
	 */
	if (queue_release &&
1149 1150
	    (!session->s_cap_reconnect ||
	     cap->cap_gen == atomic_read(&session->s_cap_gen))) {
1151 1152
		cap->queue_release = 1;
		if (removed) {
1153
			__ceph_queue_cap_release(session, cap);
1154 1155 1156 1157 1158 1159 1160
			removed = 0;
		}
	} else {
		cap->queue_release = 0;
	}
	cap->cap_ino = ci->i_vino.ino;

1161 1162
	spin_unlock(&session->s_cap_lock);

Sage Weil's avatar
Sage Weil committed
1163
	if (removed)
1164
		ceph_put_cap(mdsc, cap);
Sage Weil's avatar
Sage Weil committed
1165

1166 1167 1168 1169 1170 1171
	if (!__ceph_is_any_real_caps(ci)) {
		/* when reconnect denied, we remove session caps forcibly,
		 * i_wr_ref can be non-zero. If there are ongoing write,
		 * keep i_snap_realm.
		 */
		if (ci->i_wr_ref == 0 && ci->i_snap_realm)
1172
			ceph_change_snap_realm(&ci->vfs_inode, NULL);
1173

Sage Weil's avatar
Sage Weil committed
1174
		__cap_delay_cancel(mdsc, ci);
1175
	}
Sage Weil's avatar
Sage Weil committed
1176 1177
}

1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
void ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
{
	struct ceph_inode_info *ci = cap->ci;
	struct ceph_fs_client *fsc;

	/* 'ci' being NULL means the remove have already occurred */
	if (!ci) {
		dout("%s: cap inode is NULL\n", __func__);
		return;
	}

	lockdep_assert_held(&ci->i_ceph_lock);

	fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
	WARN_ON_ONCE(ci->i_auth_cap == cap &&
		     !list_empty(&ci->i_dirty_item) &&
		     !fsc->blocklisted &&
		     READ_ONCE(fsc->mount_state) != CEPH_MOUNT_SHUTDOWN);

	__ceph_remove_cap(cap, queue_release);
}

1200 1201 1202 1203 1204
struct cap_msg_args {
	struct ceph_mds_session	*session;
	u64			ino, cid, follows;
	u64			flush_tid, oldest_flush_tid, size, max_size;
	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 1210
	int			op, caps, wanted, dirty;
	u32			seq, issue_seq, mseq, time_warp_seq;
1211
	u32			flags;
1212 1213 1214 1215
	kuid_t			uid;
	kgid_t			gid;
	umode_t			mode;
	bool			inline_data;
1216
	bool			wake;
1217 1218
};

Sage Weil's avatar
Sage Weil committed
1219
/*
Jeff Layton's avatar
Jeff Layton committed
1220 1221
 * cap struct size + flock buffer size + inline version + inline data size +
 * osd_epoch_barrier + oldest_flush_tid
Sage Weil's avatar
Sage Weil committed
1222
 */
Jeff Layton's avatar
Jeff Layton committed
1223 1224 1225 1226 1227
#define CAP_MSG_SIZE (sizeof(struct ceph_mds_caps) + \
		      4 + 8 + 4 + 4 + 8 + 4 + 4 + 4 + 8 + 8 + 4)

/* Marshal up the cap msg to the MDS */
static void encode_cap_msg(struct ceph_msg *msg, struct cap_msg_args *arg)
Sage Weil's avatar
Sage Weil committed
1228 1229
{
	struct ceph_mds_caps *fc;
Yan, Zheng's avatar
Yan, Zheng committed
1230
	void *p;
1231
	struct ceph_osd_client *osdc = &arg->session->s_mdsc->fsc->client->osdc;
Sage Weil's avatar
Sage Weil committed
1232

Jeff Layton's avatar
Jeff Layton committed
1233 1234 1235 1236 1237 1238
	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",
	     __func__, ceph_cap_op_name(arg->op), arg->cid, arg->ino,
	     ceph_cap_string(arg->caps), ceph_cap_string(arg->wanted),
	     ceph_cap_string(arg->dirty), arg->seq, arg->issue_seq,
	     arg->flush_tid, arg->oldest_flush_tid, arg->mseq, arg->follows,
	     arg->size, arg->max_size, arg->xattr_version,
1239
	     arg->xattr_buf ? (int)arg->xattr_buf->vec.iov_len : 0);
Sage Weil's avatar
Sage Weil committed
1240

1241
	msg->hdr.version = cpu_to_le16(10);
1242
	msg->hdr.tid = cpu_to_le64(arg->flush_tid);
Sage Weil's avatar
Sage Weil committed
1243

1244
	fc = msg->front.iov_base;
Sage Weil's avatar
Sage Weil committed
1245 1246
	memset(fc, 0, sizeof(*fc));

1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
	fc->cap_id = cpu_to_le64(arg->cid);
	fc->op = cpu_to_le32(arg->op);
	fc->seq = cpu_to_le32(arg->seq);
	fc->issue_seq = cpu_to_le32(arg->issue_seq);
	fc->migrate_seq = cpu_to_le32(arg->mseq);
	fc->caps = cpu_to_le32(arg->caps);
	fc->wanted = cpu_to_le32(arg->wanted);
	fc->dirty = cpu_to_le32(arg->dirty);
	fc->ino = cpu_to_le64(arg->ino);
	fc->snap_follows = cpu_to_le64(arg->follows);

	fc->size = cpu_to_le64(arg->size);
	fc->max_size = cpu_to_le64(arg->max_size);
1260 1261 1262
	ceph_encode_timespec64(&fc->mtime, &arg->mtime);
	ceph_encode_timespec64(&fc->atime, &arg->atime);
	ceph_encode_timespec64(&fc->ctime, &arg->ctime);
1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
	fc->time_warp_seq = cpu_to_le32(arg->time_warp_seq);

	fc->uid = cpu_to_le32(from_kuid(&init_user_ns, arg->uid));
	fc->gid = cpu_to_le32(from_kgid(&init_user_ns, arg->gid));
	fc->mode = cpu_to_le32(arg->mode);

	fc->xattr_version = cpu_to_le64(arg->xattr_version);
	if (arg->xattr_buf) {
		msg->middle = ceph_buffer_get(arg->xattr_buf);
		fc->xattr_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
		msg->hdr.middle_len = cpu_to_le32(arg->xattr_buf->vec.iov_len);
1274 1275
	}

Yan, Zheng's avatar
Yan, Zheng committed
1276
	p = fc + 1;
1277
	/* flock buffer size (version 2) */
Yan, Zheng's avatar
Yan, Zheng committed
1278
	ceph_encode_32(&p, 0);
1279
	/* inline version (version 4) */
1280
	ceph_encode_64(&p, arg->inline_data ? 0 : CEPH_INLINE_NONE);
Yan, Zheng's avatar
Yan, Zheng committed
1281 1282
	/* inline data size */
	ceph_encode_32(&p, 0);
1283 1284 1285 1286 1287 1288
	/*
	 * osd_epoch_barrier (version 5)
	 * The epoch_barrier is protected osdc->lock, so READ_ONCE here in
	 * case it was recently changed
	 */
	ceph_encode_32(&p, READ_ONCE(osdc->epoch_barrier));
1289
	/* oldest_flush_tid (version 6) */
1290
	ceph_encode_64(&p, arg->oldest_flush_tid);
Yan, Zheng's avatar
Yan, Zheng committed
1291

1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304
	/*
	 * caller_uid/caller_gid (version 7)
	 *
	 * Currently, we don't properly track which caller dirtied the caps
	 * last, and force a flush of them when there is a conflict. For now,
	 * just set this to 0:0, to emulate how the MDS has worked up to now.
	 */
	ceph_encode_32(&p, 0);
	ceph_encode_32(&p, 0);

	/* pool namespace (version 8) (mds always ignores this) */
	ceph_encode_32(&p, 0);

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);
Sage Weil's avatar
Sage Weil committed
1312 1313 1314
}

/*
1315
 * Queue cap releases when an inode is dropped from our cache.
Sage Weil's avatar
Sage Weil committed
1316
 */
1317
void __ceph_remove_caps(struct ceph_inode_info *ci)
Sage Weil's avatar
Sage Weil committed
1318 1319 1320
{
	struct rb_node *p;

1321 1322 1323
	/* lock i_ceph_lock, because ceph_d_revalidate(..., LOOKUP_RCU)
	 * may call __ceph_caps_issued_mask() on a freeing inode. */
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
1324 1325 1326 1327
	p = rb_first(&ci->i_caps);
	while (p) {
		struct ceph_cap *cap = rb_entry(p, struct ceph_cap, ci_node);
		p = rb_next(p);
1328
		ceph_remove_cap(cap, true);
Sage Weil's avatar
Sage Weil committed
1329
	}
1330
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
1331 1332 1333
}

/*
1334 1335 1336
 * Prepare to send a cap message to an MDS. Update the cap state, and populate
 * the arg struct with the parameters that will need to be sent. This should
 * be done under the i_ceph_lock to guard against changes to cap state.
Sage Weil's avatar
Sage Weil committed
1337 1338 1339 1340
 *
 * Make note of max_size reported/requested from mds, revoked caps
 * that have now been implemented.
 */
1341 1342 1343
static void __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)
Sage Weil's avatar
Sage Weil committed
1344 1345 1346
{
	struct ceph_inode_info *ci = cap->ci;
	struct inode *inode = &ci->vfs_inode;
1347
	int held, revoking;
Sage Weil's avatar
Sage Weil committed
1348

1349
	lockdep_assert_held(&ci->i_ceph_lock);
1350

1351 1352 1353 1354
	held = cap->issued | cap->implemented;
	revoking = cap->implemented & ~cap->issued;
	retain &= ~revoking;

1355 1356
	dout("%s %p cap %p session %p %s -> %s (revoking %s)\n",
	     __func__, inode, cap, cap->session,
Sage Weil's avatar
Sage Weil committed
1357 1358 1359 1360
	     ceph_cap_string(held), ceph_cap_string(held & retain),
	     ceph_cap_string(revoking));
	BUG_ON((retain & CEPH_CAP_PIN) == 0);

1361
	ci->i_ceph_flags &= ~CEPH_I_FLUSH;
Sage Weil's avatar
Sage Weil committed
1362 1363

	cap->issued &= retain;  /* drop bits we don't want */
1364 1365 1366 1367 1368 1369
	/*
	 * Wake up any waiters on wanted -> needed transition. This is due to
	 * the weird transition from buffered to sync IO... we need to flush
	 * dirty pages _before_ allowing sync writes to avoid reordering.
	 */
	arg->wake = cap->implemented & ~cap->issued;
Sage Weil's avatar
Sage Weil committed
1370 1371 1372
	cap->implemented &= cap->issued | used;
	cap->mds_wanted = want;

1373 1374 1375 1376 1377 1378
	arg->session = cap->session;
	arg->ino = ceph_vino(inode).ino;
	arg->cid = cap->cap_id;
	arg->follows = flushing ? ci->i_head_snapc->seq : 0;
	arg->flush_tid = flush_tid;
	arg->oldest_flush_tid = oldest_flush_tid;
1379

1380
	arg->size = i_size_read(inode);
1381 1382
	ci->i_reported_size = arg->size;
	arg->max_size = ci->i_wanted_max_size;
1383 1384 1385 1386 1387 1388
	if (cap == ci->i_auth_cap) {
		if (want & CEPH_CAP_ANY_FILE_WR)
			ci->i_requested_max_size = arg->max_size;
		else
			ci->i_requested_max_size = 0;
	}
Sage Weil's avatar
Sage Weil committed
1389

Sage Weil's avatar
Sage Weil committed
1390
	if (flushing & CEPH_CAP_XATTR_EXCL) {
1391 1392 1393
		arg->old_xattr_buf = __ceph_build_xattrs_blob(ci);
		arg->xattr_version = ci->i_xattrs.version;
		arg->xattr_buf = ci->i_xattrs.blob;
1394
	} else {
1395 1396
		arg->xattr_buf = NULL;
		arg->old_xattr_buf = NULL;
Sage Weil's avatar
Sage Weil committed
1397 1398
	}

1399 1400 1401 1402 1403
	arg->mtime = inode->i_mtime;
	arg->atime = inode->i_atime;
	arg->ctime = inode->i_ctime;
	arg->btime = ci->i_btime;
	arg->change_attr = inode_peek_iversion_raw(inode);
1404

1405 1406 1407 1408
	arg->op = op;
	arg->caps = cap->implemented;
	arg->wanted = want;
	arg->dirty = flushing;
1409

1410 1411 1412 1413
	arg->seq = cap->seq;
	arg->issue_seq = cap->issue_seq;
	arg->mseq = cap->mseq;
	arg->time_warp_seq = ci->i_time_warp_seq;
1414

1415 1416 1417
	arg->uid = inode->i_uid;
	arg->gid = inode->i_gid;
	arg->mode = inode->i_mode;
1418

1419
	arg->inline_data = ci->i_inline_version != CEPH_INLINE_NONE;
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
	if (!(flags & CEPH_CLIENT_CAPS_PENDING_CAPSNAP) &&
	    !list_empty(&ci->i_cap_snaps)) {
		struct ceph_cap_snap *capsnap;
		list_for_each_entry_reverse(capsnap, &ci->i_cap_snaps, ci_item) {
			if (capsnap->cap_flush.tid)
				break;
			if (capsnap->need_flush) {
				flags |= CEPH_CLIENT_CAPS_PENDING_CAPSNAP;
				break;
			}
		}
	}
1432 1433
	arg->flags = flags;
}
Sage Weil's avatar
Sage Weil committed
1434

1435 1436 1437 1438 1439
/*
 * Send a cap msg on the given inode.
 *
 * Caller should hold snap_rwsem (read), s_mutex.
 */
1440
static void __send_cap(struct cap_msg_args *arg, struct ceph_inode_info *ci)
1441
{
Jeff Layton's avatar
Jeff Layton committed
1442
	struct ceph_msg *msg;
1443
	struct inode *inode = &ci->vfs_inode;
1444

Jeff Layton's avatar
Jeff Layton committed
1445 1446 1447
	msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, CAP_MSG_SIZE, GFP_NOFS, false);
	if (!msg) {
		pr_err("error allocating cap msg: ino (%llx.%llx) flushing %s tid %llu, requeuing cap.\n",
1448 1449
		       ceph_vinop(inode), ceph_cap_string(arg->dirty),
		       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);
Jeff Layton's avatar
Jeff Layton committed
1453
		return;
Sage Weil's avatar
Sage Weil committed
1454 1455
	}

Jeff Layton's avatar
Jeff Layton committed
1456 1457
	encode_cap_msg(msg, arg);
	ceph_con_send(&arg->session->s_con, msg);
1458 1459 1460
	ceph_buffer_put(arg->old_xattr_buf);
	if (arg->wake)
		wake_up_all(&ci->i_cap_wq);
Sage Weil's avatar
Sage Weil committed
1461 1462
}

1463 1464 1465 1466 1467
static inline int __send_flush_snap(struct inode *inode,
				    struct ceph_mds_session *session,
				    struct ceph_cap_snap *capsnap,
				    u32 mseq, u64 oldest_flush_tid)
{
1468
	struct cap_msg_args	arg;
Jeff Layton's avatar
Jeff Layton committed
1469 1470 1471 1472 1473
	struct ceph_msg		*msg;

	msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPS, CAP_MSG_SIZE, GFP_NOFS, false);
	if (!msg)
		return -ENOMEM;
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485

	arg.session = session;
	arg.ino = ceph_vino(inode).ino;
	arg.cid = 0;
	arg.follows = capsnap->follows;
	arg.flush_tid = capsnap->cap_flush.tid;
	arg.oldest_flush_tid = oldest_flush_tid;

	arg.size = capsnap->size;
	arg.max_size = 0;
	arg.xattr_version = capsnap->xattr_version;
	arg.xattr_buf = capsnap->xattr_blob;
1486
	arg.old_xattr_buf = NULL;
1487 1488 1489 1490

	arg.atime = capsnap->atime;
	arg.mtime = capsnap->mtime;
	arg.ctime = capsnap->ctime;
1491
	arg.btime = capsnap->btime;
1492
	arg.change_attr = capsnap->change_attr;
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508

	arg.op = CEPH_CAP_OP_FLUSHSNAP;
	arg.caps = capsnap->issued;
	arg.wanted = 0;
	arg.dirty = capsnap->dirty;

	arg.seq = 0;
	arg.issue_seq = 0;
	arg.mseq = mseq;
	arg.time_warp_seq = capsnap->time_warp_seq;

	arg.uid = capsnap->uid;
	arg.gid = capsnap->gid;
	arg.mode = capsnap->mode;

	arg.inline_data = capsnap->inline_data;
1509
	arg.flags = 0;
1510
	arg.wake = false;
1511

Jeff Layton's avatar
Jeff Layton committed
1512 1513 1514
	encode_cap_msg(msg, &arg);
	ceph_con_send(&arg.session->s_con, msg);
	return 0;
1515 1516
}

Sage Weil's avatar
Sage Weil committed
1517 1518 1519 1520 1521 1522 1523
/*
 * When a snapshot is taken, clients accumulate dirty metadata on
 * inodes with capabilities in ceph_cap_snaps to describe the file
 * state at the time the snapshot was taken.  This must be flushed
 * asynchronously back to the MDS once sync writes complete and dirty
 * data is written out.
 *
1524
 * Called under i_ceph_lock.
Sage Weil's avatar
Sage Weil committed
1525
 */
1526 1527
static void __ceph_flush_snaps(struct ceph_inode_info *ci,
			       struct ceph_mds_session *session)
1528 1529
		__releases(ci->i_ceph_lock)
		__acquires(ci->i_ceph_lock)
Sage Weil's avatar
Sage Weil committed
1530 1531
{
	struct inode *inode = &ci->vfs_inode;
1532
	struct ceph_mds_client *mdsc = session->s_mdsc;
Sage Weil's avatar
Sage Weil committed
1533
	struct ceph_cap_snap *capsnap;
1534 1535
	u64 oldest_flush_tid = 0;
	u64 first_tid = 1, last_tid = 0;
Sage Weil's avatar
Sage Weil committed
1536

1537
	dout("__flush_snaps %p session %p\n", inode, session);
Sage Weil's avatar
Sage Weil committed
1538 1539 1540 1541 1542 1543 1544

	list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
		/*
		 * we need to wait for sync writes to complete and for dirty
		 * pages to be written out.
		 */
		if (capsnap->dirty_pages || capsnap->writing)
1545
			break;
Sage Weil's avatar
Sage Weil committed
1546

1547 1548
		/* should be removed by ceph_try_drop_cap_snap() */
		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 1554 1555
			continue;
		}

1556
		spin_lock(&mdsc->cap_dirty_lock);
1557 1558 1559
		capsnap->cap_flush.tid = ++mdsc->last_cap_flush_tid;
		list_add_tail(&capsnap->cap_flush.g_list,
			      &mdsc->cap_flush_list);
1560 1561
		if (oldest_flush_tid == 0)
			oldest_flush_tid = __get_oldest_flush_tid(mdsc);
1562 1563 1564 1565
		if (list_empty(&ci->i_flushing_item)) {
			list_add_tail(&ci->i_flushing_item,
				      &session->s_cap_flushing);
		}
1566 1567
		spin_unlock(&mdsc->cap_dirty_lock);

1568 1569 1570
		list_add_tail(&capsnap->cap_flush.i_list,
			      &ci->i_cap_flush_list);

1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
		if (first_tid == 1)
			first_tid = capsnap->cap_flush.tid;
		last_tid = capsnap->cap_flush.tid;
	}

	ci->i_ceph_flags &= ~CEPH_I_FLUSH_SNAPS;

	while (first_tid <= last_tid) {
		struct ceph_cap *cap = ci->i_auth_cap;
		struct ceph_cap_flush *cf;
		int ret;

		if (!(cap && cap->session == session)) {
			dout("__flush_snaps %p auth cap %p not mds%d, "
			     "stop\n", inode, cap, session->s_mds);
			break;
		}

		ret = -ENOENT;
		list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
			if (cf->tid >= first_tid) {
				ret = 0;
				break;
			}
		}
		if (ret < 0)
			break;

		first_tid = cf->tid + 1;

		capsnap = container_of(cf, struct ceph_cap_snap, cap_flush);
1602
		refcount_inc(&capsnap->nref);
1603
		spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
1604

1605 1606
		dout("__flush_snaps %p capsnap %p tid %llu %s\n",
		     inode, capsnap, cf->tid, ceph_cap_string(capsnap->dirty));
Sage Weil's avatar
Sage Weil committed
1607

1608 1609 1610 1611 1612 1613 1614
		ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
					oldest_flush_tid);
		if (ret < 0) {
			pr_err("__flush_snaps: error sending cap flushsnap, "
			       "ino (%llx.%llx) tid %llu follows %llu\n",
				ceph_vinop(inode), cf->tid, capsnap->follows);
		}
Sage Weil's avatar
Sage Weil committed
1615

1616
		ceph_put_cap_snap(capsnap);
1617
		spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
1618
	}
1619
}
Sage Weil's avatar
Sage Weil committed
1620

1621 1622 1623 1624 1625
void ceph_flush_snaps(struct ceph_inode_info *ci,
		      struct ceph_mds_session **psession)
{
	struct inode *inode = &ci->vfs_inode;
	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
1626
	struct ceph_mds_session *session = NULL;
1627
	int mds;
1628

1629
	dout("ceph_flush_snaps %p\n", inode);
1630 1631
	if (psession)
		session = *psession;
1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
retry:
	spin_lock(&ci->i_ceph_lock);
	if (!(ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)) {
		dout(" no capsnap needs flush, doing nothing\n");
		goto out;
	}
	if (!ci->i_auth_cap) {
		dout(" no auth cap (migrating?), doing nothing\n");
		goto out;
	}
Sage Weil's avatar
Sage Weil committed
1642

1643 1644 1645
	mds = ci->i_auth_cap->session->s_mds;
	if (session && session->s_mds != mds) {
		dout(" oops, wrong session %p mutex\n", session);
Sage Weil's avatar
Sage Weil committed
1646
		ceph_put_mds_session(session);
1647 1648 1649 1650 1651 1652 1653 1654
		session = NULL;
	}
	if (!session) {
		spin_unlock(&ci->i_ceph_lock);
		mutex_lock(&mdsc->mutex);
		session = __ceph_lookup_mds_session(mdsc, mds);
		mutex_unlock(&mdsc->mutex);
		goto retry;
Sage Weil's avatar
Sage Weil committed
1655 1656
	}

1657
	// make sure flushsnap messages are sent in proper order.
1658
	if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
1659 1660
		__kick_flushing_caps(mdsc, session, ci, 0);

1661 1662
	__ceph_flush_snaps(ci, session);
out:
1663
	spin_unlock(&ci->i_ceph_lock);
1664

1665
	if (psession)
1666
		*psession = session;
1667
	else
1668 1669 1670 1671 1672
		ceph_put_mds_session(session);
	/* we flushed them all; remove this inode from the queue */
	spin_lock(&mdsc->snap_flush_lock);
	list_del_init(&ci->i_snap_flush_item);
	spin_unlock(&mdsc->snap_flush_lock);
Sage Weil's avatar
Sage Weil committed
1673 1674
}

1675
/*
1676 1677 1678
 * Mark caps dirty.  If inode is newly dirty, return the dirty flags.
 * Caller is then responsible for calling __mark_inode_dirty with the
 * returned flags value.
1679
 */
1680 1681
int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
			   struct ceph_cap_flush **pcf)
1682
{
1683
	struct ceph_mds_client *mdsc =
1684
		ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc;
1685 1686 1687 1688
	struct inode *inode = &ci->vfs_inode;
	int was = ci->i_dirty_caps;
	int dirty = 0;

1689 1690
	lockdep_assert_held(&ci->i_ceph_lock);

1691 1692 1693 1694 1695 1696 1697
	if (!ci->i_auth_cap) {
		pr_warn("__mark_dirty_caps %p %llx mask %s, "
			"but no auth cap (session was closed?)\n",
			inode, ceph_ino(inode), ceph_cap_string(mask));
		return 0;
	}

1698 1699 1700 1701 1702
	dout("__mark_dirty_caps %p %s dirty %s -> %s\n", &ci->vfs_inode,
	     ceph_cap_string(mask), ceph_cap_string(was),
	     ceph_cap_string(was | mask));
	ci->i_dirty_caps |= mask;
	if (was == 0) {
1703 1704
		struct ceph_mds_session *session = ci->i_auth_cap->session;

1705 1706 1707
		WARN_ON_ONCE(ci->i_prealloc_cap_flush);
		swap(ci->i_prealloc_cap_flush, *pcf);

1708 1709
		if (!ci->i_head_snapc) {
			WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem));
1710 1711
			ci->i_head_snapc = ceph_get_snap_context(
				ci->i_snap_realm->cached_context);
1712
		}
1713 1714
		dout(" inode %p now dirty snapc %p auth cap %p\n",
		     &ci->vfs_inode, ci->i_head_snapc, ci->i_auth_cap);
1715 1716
		BUG_ON(!list_empty(&ci->i_dirty_item));
		spin_lock(&mdsc->cap_dirty_lock);
1717
		list_add(&ci->i_dirty_item, &session->s_cap_dirty);
1718 1719
		spin_unlock(&mdsc->cap_dirty_lock);
		if (ci->i_flushing_caps == 0) {
1720
			ihold(inode);
1721 1722
			dirty |= I_DIRTY_SYNC;
		}
1723 1724
	} else {
		WARN_ON_ONCE(!ci->i_prealloc_cap_flush);
1725 1726 1727 1728 1729
	}
	BUG_ON(list_empty(&ci->i_dirty_item));
	if (((was | ci->i_flushing_caps) & CEPH_CAP_FILE_BUFFER) &&
	    (mask & CEPH_CAP_FILE_BUFFER))
		dirty |= I_DIRTY_DATASYNC;
1730
	__cap_delay_requeue(mdsc, ci);
1731
	return dirty;
1732 1733
}

1734 1735
struct ceph_cap_flush *ceph_alloc_cap_flush(void)
{
1736 1737 1738
	struct ceph_cap_flush *cf;

	cf = kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
1739 1740 1741
	if (!cf)
		return NULL;

1742 1743
	cf->is_capsnap = false;
	return cf;
1744 1745 1746 1747 1748 1749 1750 1751
}

void ceph_free_cap_flush(struct ceph_cap_flush *cf)
{
	if (cf)
		kmem_cache_free(ceph_cap_flush_cachep, cf);
}

1752 1753
static u64 __get_oldest_flush_tid(struct ceph_mds_client *mdsc)
{
1754
	if (!list_empty(&mdsc->cap_flush_list)) {
1755
		struct ceph_cap_flush *cf =
1756 1757
			list_first_entry(&mdsc->cap_flush_list,
					 struct ceph_cap_flush, g_list);
1758 1759 1760 1761 1762
		return cf->tid;
	}
	return 0;
}

1763 1764 1765 1766
/*
 * Remove cap_flush from the mdsc's or inode's flushing cap list.
 * Return true if caller needs to wake up flush waiters.
 */
1767 1768
static bool __detach_cap_flush_from_mdsc(struct ceph_mds_client *mdsc,
					 struct ceph_cap_flush *cf)
1769 1770 1771
{
	struct ceph_cap_flush *prev;
	bool wake = cf->wake;
1772 1773 1774 1775 1776 1777

	if (wake && cf->g_list.prev != &mdsc->cap_flush_list) {
		prev = list_prev_entry(cf, g_list);
		prev->wake = true;
		wake = false;
	}
1778
	list_del_init(&cf->g_list);
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791
	return wake;
}

static bool __detach_cap_flush_from_ci(struct ceph_inode_info *ci,
				       struct ceph_cap_flush *cf)
{
	struct ceph_cap_flush *prev;
	bool wake = cf->wake;

	if (wake && cf->i_list.prev != &ci->i_cap_flush_list) {
		prev = list_prev_entry(cf, i_list);
		prev->wake = true;
		wake = false;
1792
	}
1793
	list_del_init(&cf->i_list);
1794 1795 1796
	return wake;
}

Sage Weil's avatar
Sage Weil committed
1797 1798 1799
/*
 * Add dirty inode to the flushing list.  Assigned a seq number so we
 * can wait for caps to flush without starving.
1800
 *
1801
 * Called under i_ceph_lock. Returns the flush tid.
Sage Weil's avatar
Sage Weil committed
1802
 */
1803
static u64 __mark_caps_flushing(struct inode *inode,
1804
				struct ceph_mds_session *session, bool wake,
1805
				u64 *oldest_flush_tid)
Sage Weil's avatar
Sage Weil committed
1806
{
1807
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
Sage Weil's avatar
Sage Weil committed
1808
	struct ceph_inode_info *ci = ceph_inode(inode);
1809
	struct ceph_cap_flush *cf = NULL;
1810
	int flushing;
Sage Weil's avatar
Sage Weil committed
1811

1812
	lockdep_assert_held(&ci->i_ceph_lock);
1813
	BUG_ON(ci->i_dirty_caps == 0);
Sage Weil's avatar
Sage Weil committed
1814
	BUG_ON(list_empty(&ci->i_dirty_item));
1815
	BUG_ON(!ci->i_prealloc_cap_flush);
1816 1817 1818 1819 1820 1821 1822 1823

	flushing = ci->i_dirty_caps;
	dout("__mark_caps_flushing flushing %s, flushing_caps %s -> %s\n",
	     ceph_cap_string(flushing),
	     ceph_cap_string(ci->i_flushing_caps),
	     ceph_cap_string(ci->i_flushing_caps | flushing));
	ci->i_flushing_caps |= flushing;
	ci->i_dirty_caps = 0;
1824
	dout(" inode %p now !dirty\n", inode);
1825

1826
	swap(cf, ci->i_prealloc_cap_flush);
1827
	cf->caps = flushing;
1828
	cf->wake = wake;
1829

Sage Weil's avatar
Sage Weil committed
1830
	spin_lock(&mdsc->cap_dirty_lock);
1831 1832
	list_del_init(&ci->i_dirty_item);

1833
	cf->tid = ++mdsc->last_cap_flush_tid;
1834
	list_add_tail(&cf->g_list, &mdsc->cap_flush_list);
1835
	*oldest_flush_tid = __get_oldest_flush_tid(mdsc);
1836

Sage Weil's avatar
Sage Weil committed
1837 1838 1839 1840 1841
	if (list_empty(&ci->i_flushing_item)) {
		list_add_tail(&ci->i_flushing_item, &session->s_cap_flushing);
		mdsc->num_cap_flushing++;
	}
	spin_unlock(&mdsc->cap_dirty_lock);
1842

1843
	list_add_tail(&cf->i_list, &ci->i_cap_flush_list);
1844

1845
	return cf->tid;
Sage Weil's avatar
Sage Weil committed
1846 1847
}

1848 1849 1850 1851
/*
 * try to invalidate mapping pages without blocking.
 */
static int try_nonblocking_invalidate(struct inode *inode)
1852 1853
	__releases(ci->i_ceph_lock)
	__acquires(ci->i_ceph_lock)
1854 1855 1856 1857
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	u32 invalidating_gen = ci->i_rdcache_gen;

1858
	spin_unlock(&ci->i_ceph_lock);
1859
	ceph_fscache_invalidate(inode);
1860
	invalidate_mapping_pages(&inode->i_data, 0, -1);
1861
	spin_lock(&ci->i_ceph_lock);
1862

1863
	if (inode->i_data.nrpages == 0 &&
1864 1865 1866
	    invalidating_gen == ci->i_rdcache_gen) {
		/* success. */
		dout("try_nonblocking_invalidate %p success\n", inode);
1867 1868
		/* save any racing async invalidate some trouble */
		ci->i_rdcache_revoking = ci->i_rdcache_gen - 1;
1869 1870 1871 1872 1873 1874
		return 0;
	}
	dout("try_nonblocking_invalidate %p failed\n", inode);
	return -1;
}

1875 1876
bool __ceph_should_report_size(struct ceph_inode_info *ci)
{
1877
	loff_t size = i_size_read(&ci->vfs_inode);
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
	/* mds will adjust max size according to the reported size */
	if (ci->i_flushing_caps & CEPH_CAP_FILE_WR)
		return false;
	if (size >= ci->i_max_size)
		return true;
	/* half of previous max_size increment has been used */
	if (ci->i_max_size > ci->i_reported_size &&
	    (size << 1) >= ci->i_max_size + ci->i_reported_size)
		return true;
	return false;
}

Sage Weil's avatar
Sage Weil committed
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902
/*
 * Swiss army knife function to examine currently used and wanted
 * versus held caps.  Release, flush, ack revoked caps to mds as
 * appropriate.
 *
 *  CHECK_CAPS_AUTHONLY - we should only check the auth cap
 *  CHECK_CAPS_FLUSH - we should flush any dirty caps immediately, without
 *    further delay.
 */
void ceph_check_caps(struct ceph_inode_info *ci, int flags,
		     struct ceph_mds_session *session)
{
	struct inode *inode = &ci->vfs_inode;
1903
	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
Sage Weil's avatar
Sage Weil committed
1904
	struct ceph_cap *cap;
1905
	u64 flush_tid, oldest_flush_tid;
1906
	int file_wanted, used, cap_used;
Sage Weil's avatar
Sage Weil committed
1907
	int issued, implemented, want, retain, revoking, flushing = 0;
Sage Weil's avatar
Sage Weil committed
1908 1909 1910
	int mds = -1;   /* keep track of how far we've gone through i_caps list
			   to avoid an infinite loop on retry */
	struct rb_node *p;
1911 1912
	bool queue_invalidate = false;
	bool tried_invalidate = false;
Sage Weil's avatar
Sage Weil committed
1913

1914 1915 1916
	if (session)
		ceph_get_mds_session(session);

1917
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
1918 1919 1920
	if (ci->i_ceph_flags & CEPH_I_FLUSH)
		flags |= CHECK_CAPS_FLUSH;
retry:
1921
	/* Caps wanted by virtue of active open files. */
Sage Weil's avatar
Sage Weil committed
1922
	file_wanted = __ceph_caps_file_wanted(ci);
1923 1924

	/* Caps which have active references against them */
Sage Weil's avatar
Sage Weil committed
1925
	used = __ceph_caps_used(ci);
1926 1927 1928 1929 1930 1931 1932

	/*
	 * "issued" represents the current caps that the MDS wants us to have.
	 * "implemented" is the set that we have been granted, and includes the
	 * ones that have not yet been returned to the MDS (the "revoking" set,
	 * usually because they have outstanding references).
	 */
Sage Weil's avatar
Sage Weil committed
1933 1934
	issued = __ceph_caps_issued(ci, &implemented);
	revoking = implemented & ~issued;
Sage Weil's avatar
Sage Weil committed
1935

1936
	want = file_wanted;
1937 1938

	/* The ones we currently want to retain (may be adjusted below) */
1939
	retain = file_wanted | used | CEPH_CAP_PIN;
Sage Weil's avatar
Sage Weil committed
1940
	if (!mdsc->stopping && inode->i_nlink > 0) {
1941
		if (file_wanted) {
Sage Weil's avatar
Sage Weil committed
1942
			retain |= CEPH_CAP_ANY;       /* be greedy */
1943 1944
		} else if (S_ISDIR(inode->i_mode) &&
			   (issued & CEPH_CAP_FILE_SHARED) &&
1945
			   __ceph_dir_is_complete(ci)) {
1946 1947 1948 1949 1950 1951
			/*
			 * If a directory is complete, we want to keep
			 * the exclusive cap. So that MDS does not end up
			 * revoking the shared cap on every create/unlink
			 * operation.
			 */
1952
			if (IS_RDONLY(inode)) {
1953
				want = CEPH_CAP_ANY_SHARED;
1954
			} else {
1955
				want |= CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_EXCL;
1956
			}
1957
			retain |= want;
Sage Weil's avatar
Sage Weil committed
1958
		} else {
1959

Sage Weil's avatar
Sage Weil committed
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970
			retain |= CEPH_CAP_ANY_SHARED;
			/*
			 * keep RD only if we didn't have the file open RW,
			 * because then the mds would revoke it anyway to
			 * journal max_size=0.
			 */
			if (ci->i_max_size == 0)
				retain |= CEPH_CAP_ANY_RD;
		}
	}

1971 1972
	dout("check_caps %llx.%llx file_want %s used %s dirty %s flushing %s"
	     " issued %s revoking %s retain %s %s%s\n", ceph_vinop(inode),
Sage Weil's avatar
Sage Weil committed
1973 1974 1975
	     ceph_cap_string(file_wanted),
	     ceph_cap_string(used), ceph_cap_string(ci->i_dirty_caps),
	     ceph_cap_string(ci->i_flushing_caps),
Sage Weil's avatar
Sage Weil committed
1976
	     ceph_cap_string(issued), ceph_cap_string(revoking),
Sage Weil's avatar
Sage Weil committed
1977 1978 1979 1980 1981 1982 1983 1984 1985
	     ceph_cap_string(retain),
	     (flags & CHECK_CAPS_AUTHONLY) ? " AUTHONLY" : "",
	     (flags & CHECK_CAPS_FLUSH) ? " FLUSH" : "");

	/*
	 * If we no longer need to hold onto old our caps, and we may
	 * have cached pages, but don't want them, then try to invalidate.
	 * If we fail, it's because pages are locked.... try again later.
	 */
1986
	if ((!(flags & CHECK_CAPS_NOINVAL) || mdsc->stopping) &&
1987
	    S_ISREG(inode->i_mode) &&
1988
	    !(ci->i_wb_ref || ci->i_wrbuffer_ref) &&   /* no dirty pages... */
Yan, Zheng's avatar
Yan, Zheng committed
1989
	    inode->i_data.nrpages &&		/* have cached pages */
1990 1991
	    (revoking & (CEPH_CAP_FILE_CACHE|
			 CEPH_CAP_FILE_LAZYIO)) && /*  or revoking cache */
Sage Weil's avatar
Sage Weil committed
1992
	    !tried_invalidate) {
1993 1994
		dout("check_caps trying to invalidate on %llx.%llx\n",
		     ceph_vinop(inode));
1995
		if (try_nonblocking_invalidate(inode) < 0) {
1996 1997 1998
			dout("check_caps queuing invalidate\n");
			queue_invalidate = true;
			ci->i_rdcache_revoking = ci->i_rdcache_gen;
Sage Weil's avatar
Sage Weil committed
1999
		}
2000
		tried_invalidate = true;
2001
		goto retry;
Sage Weil's avatar
Sage Weil committed
2002 2003 2004
	}

	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
2005
		int mflags = 0;
2006 2007
		struct cap_msg_args arg;

Sage Weil's avatar
Sage Weil committed
2008 2009 2010 2011 2012 2013 2014
		cap = rb_entry(p, struct ceph_cap, ci_node);

		/* avoid looping forever */
		if (mds >= cap->mds ||
		    ((flags & CHECK_CAPS_AUTHONLY) && cap != ci->i_auth_cap))
			continue;

2015 2016 2017 2018
		/*
		 * If we have an auth cap, we don't need to consider any
		 * overlapping caps as used.
		 */
2019 2020 2021 2022
		cap_used = used;
		if (ci->i_auth_cap && cap != ci->i_auth_cap)
			cap_used &= ~ci->i_auth_cap->issued;

Sage Weil's avatar
Sage Weil committed
2023
		revoking = cap->implemented & ~cap->issued;
2024
		dout(" mds%d cap %p used %s issued %s implemented %s revoking %s\n",
2025 2026
		     cap->mds, cap, ceph_cap_string(cap_used),
		     ceph_cap_string(cap->issued),
2027 2028
		     ceph_cap_string(cap->implemented),
		     ceph_cap_string(revoking));
Sage Weil's avatar
Sage Weil committed
2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039

		if (cap == ci->i_auth_cap &&
		    (cap->issued & CEPH_CAP_FILE_WR)) {
			/* request larger max_size from MDS? */
			if (ci->i_wanted_max_size > ci->i_max_size &&
			    ci->i_wanted_max_size > ci->i_requested_max_size) {
				dout("requesting new max_size\n");
				goto ack;
			}

			/* approaching file_max? */
2040
			if (__ceph_should_report_size(ci)) {
Sage Weil's avatar
Sage Weil committed
2041 2042 2043 2044 2045
				dout("i_size approaching max_size\n");
				goto ack;
			}
		}
		/* flush anything dirty? */
2046 2047 2048 2049 2050 2051 2052 2053 2054
		if (cap == ci->i_auth_cap) {
			if ((flags & CHECK_CAPS_FLUSH) && ci->i_dirty_caps) {
				dout("flushing dirty caps\n");
				goto ack;
			}
			if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS) {
				dout("flushing snap caps\n");
				goto ack;
			}
Sage Weil's avatar
Sage Weil committed
2055 2056 2057
		}

		/* completed revocation? going down and there are no caps? */
2058
		if (revoking && (revoking & cap_used) == 0) {
Sage Weil's avatar
Sage Weil committed
2059 2060 2061 2062 2063 2064
			dout("completed revocation of %s\n",
			     ceph_cap_string(cap->implemented & ~cap->issued));
			goto ack;
		}

		/* want more caps from mds? */
2065 2066 2067 2068 2069 2070
		if (want & ~cap->mds_wanted) {
			if (want & ~(cap->mds_wanted | cap->issued))
				goto ack;
			if (!__cap_is_valid(cap))
				goto ack;
		}
Sage Weil's avatar
Sage Weil committed
2071 2072

		/* things we might delay */
2073
		if ((cap->issued & ~retain) == 0)
Sage Weil's avatar
Sage Weil committed
2074 2075 2076
			continue;     /* nope, all good */

ack:
2077 2078
		ceph_put_mds_session(session);
		session = ceph_get_mds_session(cap->session);
2079 2080 2081 2082 2083 2084

		/* kick flushing and flush snaps before sending normal
		 * cap message */
		if (cap == ci->i_auth_cap &&
		    (ci->i_ceph_flags &
		     (CEPH_I_KICK_FLUSH | CEPH_I_FLUSH_SNAPS))) {
2085
			if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
2086
				__kick_flushing_caps(mdsc, session, ci, 0);
2087 2088 2089
			if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
				__ceph_flush_snaps(ci, session);

2090
			goto retry;
Sage Weil's avatar
Sage Weil committed
2091 2092
		}

2093
		if (cap == ci->i_auth_cap && ci->i_dirty_caps) {
2094 2095 2096
			flushing = ci->i_dirty_caps;
			flush_tid = __mark_caps_flushing(inode, session, false,
							 &oldest_flush_tid);
2097 2098 2099
			if (flags & CHECK_CAPS_FLUSH &&
			    list_empty(&session->s_cap_dirty))
				mflags |= CEPH_CLIENT_CAPS_SYNC;
2100
		} else {
2101
			flushing = 0;
2102
			flush_tid = 0;
2103 2104 2105
			spin_lock(&mdsc->cap_dirty_lock);
			oldest_flush_tid = __get_oldest_flush_tid(mdsc);
			spin_unlock(&mdsc->cap_dirty_lock);
2106
		}
Sage Weil's avatar
Sage Weil committed
2107 2108 2109

		mds = cap->mds;  /* remember mds, so we don't repeat */

2110 2111
		__prep_cap(&arg, cap, CEPH_CAP_OP_UPDATE, mflags, cap_used,
			   want, retain, flushing, flush_tid, oldest_flush_tid);
2112

2113
		spin_unlock(&ci->i_ceph_lock);
2114
		__send_cap(&arg, ci);
2115
		spin_lock(&ci->i_ceph_lock);
2116

2117
		goto retry; /* retake i_ceph_lock and restart our cap scan. */
Sage Weil's avatar
Sage Weil committed
2118 2119
	}

2120 2121 2122 2123 2124 2125
	/* periodically re-calculate caps wanted by open files */
	if (__ceph_is_any_real_caps(ci) &&
	    list_empty(&ci->i_cap_delay_list) &&
	    (file_wanted & ~CEPH_CAP_PIN) &&
	    !(used & (CEPH_CAP_FILE_RD | CEPH_CAP_ANY_FILE_WR))) {
		__cap_delay_requeue(mdsc, ci);
2126
	}
Sage Weil's avatar
Sage Weil committed
2127

2128
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2129

2130
	ceph_put_mds_session(session);
Sage Weil's avatar
Sage Weil committed
2131
	if (queue_invalidate)
2132
		ceph_queue_invalidate(inode);
Sage Weil's avatar
Sage Weil committed
2133 2134 2135 2136 2137
}

/*
 * Try to flush dirty caps back to the auth mds.
 */
2138
static int try_flush_caps(struct inode *inode, u64 *ptid)
Sage Weil's avatar
Sage Weil committed
2139
{
2140
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
Sage Weil's avatar
Sage Weil committed
2141
	struct ceph_inode_info *ci = ceph_inode(inode);
Yan, Zheng's avatar
Yan, Zheng committed
2142
	int flushing = 0;
2143
	u64 flush_tid = 0, oldest_flush_tid = 0;
Sage Weil's avatar
Sage Weil committed
2144

2145
	spin_lock(&ci->i_ceph_lock);
2146
retry_locked:
Sage Weil's avatar
Sage Weil committed
2147 2148
	if (ci->i_dirty_caps && ci->i_auth_cap) {
		struct ceph_cap *cap = ci->i_auth_cap;
2149
		struct cap_msg_args arg;
2150
		struct ceph_mds_session *session = cap->session;
Sage Weil's avatar
Sage Weil committed
2151

2152
		if (session->s_state < CEPH_MDS_SESSION_OPEN) {
2153
			spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2154
			goto out;
2155
		}
Sage Weil's avatar
Sage Weil committed
2156

2157 2158 2159 2160 2161 2162 2163 2164 2165
		if (ci->i_ceph_flags &
		    (CEPH_I_KICK_FLUSH | CEPH_I_FLUSH_SNAPS)) {
			if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH)
				__kick_flushing_caps(mdsc, session, ci, 0);
			if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
				__ceph_flush_snaps(ci, session);
			goto retry_locked;
		}

2166 2167 2168
		flushing = ci->i_dirty_caps;
		flush_tid = __mark_caps_flushing(inode, session, true,
						 &oldest_flush_tid);
Sage Weil's avatar
Sage Weil committed
2169

2170
		__prep_cap(&arg, cap, CEPH_CAP_OP_FLUSH, CEPH_CLIENT_CAPS_SYNC,
2171 2172 2173
			   __ceph_caps_used(ci), __ceph_caps_wanted(ci),
			   (cap->issued | cap->implemented),
			   flushing, flush_tid, oldest_flush_tid);
2174 2175
		spin_unlock(&ci->i_ceph_lock);

2176
		__send_cap(&arg, ci);
2177
	} else {
2178
		if (!list_empty(&ci->i_cap_flush_list)) {
2179
			struct ceph_cap_flush *cf =
2180
				list_last_entry(&ci->i_cap_flush_list,
2181 2182
						struct ceph_cap_flush, i_list);
			cf->wake = true;
2183 2184 2185 2186
			flush_tid = cf->tid;
		}
		flushing = ci->i_flushing_caps;
		spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2187 2188
	}
out:
2189
	*ptid = flush_tid;
Sage Weil's avatar
Sage Weil committed
2190 2191 2192 2193 2194 2195
	return flushing;
}

/*
 * Return true if we've flushed caps through the given flush_tid.
 */
2196
static int caps_are_flushed(struct inode *inode, u64 flush_tid)
Sage Weil's avatar
Sage Weil committed
2197 2198
{
	struct ceph_inode_info *ci = ceph_inode(inode);
2199
	int ret = 1;
Sage Weil's avatar
Sage Weil committed
2200

2201
	spin_lock(&ci->i_ceph_lock);
2202 2203 2204 2205
	if (!list_empty(&ci->i_cap_flush_list)) {
		struct ceph_cap_flush * cf =
			list_first_entry(&ci->i_cap_flush_list,
					 struct ceph_cap_flush, i_list);
2206
		if (cf->tid <= flush_tid)
Sage Weil's avatar
Sage Weil committed
2207
			ret = 0;
Yan, Zheng's avatar
Yan, Zheng committed
2208
	}
2209
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2210 2211 2212
	return ret;
}

Yan, Zheng's avatar
Yan, Zheng committed
2213
/*
2214
 * wait for any unsafe requests to complete.
Yan, Zheng's avatar
Yan, Zheng committed
2215
 */
2216
static int unsafe_request_wait(struct inode *inode)
Yan, Zheng's avatar
Yan, Zheng committed
2217
{
2218
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
Yan, Zheng's avatar
Yan, Zheng committed
2219
	struct ceph_inode_info *ci = ceph_inode(inode);
2220 2221
	struct ceph_mds_request *req1 = NULL, *req2 = NULL;
	int ret, err = 0;
Yan, Zheng's avatar
Yan, Zheng committed
2222 2223

	spin_lock(&ci->i_unsafe_lock);
2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
	if (S_ISDIR(inode->i_mode) && !list_empty(&ci->i_unsafe_dirops)) {
		req1 = list_last_entry(&ci->i_unsafe_dirops,
					struct ceph_mds_request,
					r_unsafe_dir_item);
		ceph_mdsc_get_request(req1);
	}
	if (!list_empty(&ci->i_unsafe_iops)) {
		req2 = list_last_entry(&ci->i_unsafe_iops,
					struct ceph_mds_request,
					r_unsafe_target_item);
		ceph_mdsc_get_request(req2);
	}
	spin_unlock(&ci->i_unsafe_lock);
Yan, Zheng's avatar
Yan, Zheng committed
2237

2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
	/*
	 * Trigger to flush the journal logs in all the relevant MDSes
	 * manually, or in the worst case we must wait at most 5 seconds
	 * to wait the journal logs to be flushed by the MDSes periodically.
	 */
	if (req1 || req2) {
		struct ceph_mds_session **sessions = NULL;
		struct ceph_mds_session *s;
		struct ceph_mds_request *req;
		unsigned int max;
		int i;

		/*
		 * The mdsc->max_sessions is unlikely to be changed
		 * mostly, here we will retry it by reallocating the
		 * sessions arrary memory to get rid of the mdsc->mutex
		 * lock.
		 */
retry:
		max = mdsc->max_sessions;
		sessions = krealloc(sessions, max * sizeof(s), __GFP_ZERO);
		if (!sessions)
			return -ENOMEM;

		spin_lock(&ci->i_unsafe_lock);
		if (req1) {
			list_for_each_entry(req, &ci->i_unsafe_dirops,
					    r_unsafe_dir_item) {
				s = req->r_session;
2267
				if (unlikely(s->s_mds >= max)) {
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
					spin_unlock(&ci->i_unsafe_lock);
					goto retry;
				}
				if (!sessions[s->s_mds]) {
					s = ceph_get_mds_session(s);
					sessions[s->s_mds] = s;
				}
			}
		}
		if (req2) {
			list_for_each_entry(req, &ci->i_unsafe_iops,
					    r_unsafe_target_item) {
				s = req->r_session;
2281
				if (unlikely(s->s_mds >= max)) {
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
					spin_unlock(&ci->i_unsafe_lock);
					goto retry;
				}
				if (!sessions[s->s_mds]) {
					s = ceph_get_mds_session(s);
					sessions[s->s_mds] = s;
				}
			}
		}
		spin_unlock(&ci->i_unsafe_lock);

		/* the auth MDS */
		spin_lock(&ci->i_ceph_lock);
		if (ci->i_auth_cap) {
		      s = ci->i_auth_cap->session;
		      if (!sessions[s->s_mds])
			      sessions[s->s_mds] = ceph_get_mds_session(s);
		}
		spin_unlock(&ci->i_ceph_lock);

		/* send flush mdlog request to MDSes */
		for (i = 0; i < max; i++) {
			s = sessions[i];
			if (s) {
				send_flush_mdlog(s);
				ceph_put_mds_session(s);
			}
		}
		kfree(sessions);
	}

2313
	dout("unsafe_request_wait %p wait on tid %llu %llu\n",
2314 2315 2316 2317
	     inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
	if (req1) {
		ret = !wait_for_completion_timeout(&req1->r_safe_completion,
					ceph_timeout_jiffies(req1->r_timeout));
Yan, Zheng's avatar
Yan, Zheng committed
2318
		if (ret)
2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329
			err = -EIO;
		ceph_mdsc_put_request(req1);
	}
	if (req2) {
		ret = !wait_for_completion_timeout(&req2->r_safe_completion,
					ceph_timeout_jiffies(req2->r_timeout));
		if (ret)
			err = -EIO;
		ceph_mdsc_put_request(req2);
	}
	return err;
Yan, Zheng's avatar
Yan, Zheng committed
2330 2331
}

2332
int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync)
Sage Weil's avatar
Sage Weil committed
2333
{
2334
	struct inode *inode = file->f_mapping->host;
Sage Weil's avatar
Sage Weil committed
2335
	struct ceph_inode_info *ci = ceph_inode(inode);
2336
	u64 flush_tid;
2337
	int ret, err;
Sage Weil's avatar
Sage Weil committed
2338 2339 2340
	int dirty;

	dout("fsync %p%s\n", inode, datasync ? " datasync" : "");
2341

2342
	ret = file_write_and_wait_range(file, start, end);
Yan, Zheng's avatar
Yan, Zheng committed
2343 2344 2345
	if (datasync)
		goto out;

2346 2347 2348 2349
	ret = ceph_wait_on_async_create(inode);
	if (ret)
		goto out;

2350
	dirty = try_flush_caps(inode, &flush_tid);
Sage Weil's avatar
Sage Weil committed
2351 2352
	dout("fsync dirty caps are %s\n", ceph_cap_string(dirty));

2353
	err = unsafe_request_wait(inode);
Yan, Zheng's avatar
Yan, Zheng committed
2354

Sage Weil's avatar
Sage Weil committed
2355 2356 2357 2358 2359
	/*
	 * only wait on non-file metadata writeback (the mds
	 * can recover size and mtime, so we don't need to
	 * wait for that)
	 */
2360 2361
	if (!err && (dirty & ~CEPH_CAP_ANY_FILE_WR)) {
		err = wait_event_interruptible(ci->i_cap_wq,
Yan, Zheng's avatar
Yan, Zheng committed
2362
					caps_are_flushed(inode, flush_tid));
Sage Weil's avatar
Sage Weil committed
2363
	}
2364 2365 2366 2367

	if (err < 0)
		ret = err;

2368 2369 2370
	err = file_check_and_advance_wb_err(file);
	if (err < 0)
		ret = err;
Yan, Zheng's avatar
Yan, Zheng committed
2371 2372
out:
	dout("fsync %p%s result=%d\n", inode, datasync ? " datasync" : "", ret);
Sage Weil's avatar
Sage Weil committed
2373 2374 2375 2376 2377 2378 2379 2380 2381
	return ret;
}

/*
 * Flush any dirty caps back to the mds.  If we aren't asked to wait,
 * queue inode for flush but don't do so immediately, because we can
 * get by with fewer MDS messages if we wait for data writeback to
 * complete first.
 */
2382
int ceph_write_inode(struct inode *inode, struct writeback_control *wbc)
Sage Weil's avatar
Sage Weil committed
2383 2384
{
	struct ceph_inode_info *ci = ceph_inode(inode);
2385
	u64 flush_tid;
Sage Weil's avatar
Sage Weil committed
2386 2387
	int err = 0;
	int dirty;
2388
	int wait = (wbc->sync_mode == WB_SYNC_ALL && !wbc->for_sync);
Sage Weil's avatar
Sage Weil committed
2389 2390 2391

	dout("write_inode %p wait=%d\n", inode, wait);
	if (wait) {
2392
		dirty = try_flush_caps(inode, &flush_tid);
Sage Weil's avatar
Sage Weil committed
2393 2394 2395 2396
		if (dirty)
			err = wait_event_interruptible(ci->i_cap_wq,
				       caps_are_flushed(inode, flush_tid));
	} else {
2397
		struct ceph_mds_client *mdsc =
2398
			ceph_sb_to_client(inode->i_sb)->mdsc;
Sage Weil's avatar
Sage Weil committed
2399

2400
		spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2401 2402
		if (__ceph_caps_dirty(ci))
			__cap_delay_requeue_front(mdsc, ci);
2403
		spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2404 2405 2406 2407
	}
	return err;
}

2408 2409 2410 2411 2412 2413
static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
				 struct ceph_mds_session *session,
				 struct ceph_inode_info *ci,
				 u64 oldest_flush_tid)
	__releases(ci->i_ceph_lock)
	__acquires(ci->i_ceph_lock)
2414 2415 2416 2417
{
	struct inode *inode = &ci->vfs_inode;
	struct ceph_cap *cap;
	struct ceph_cap_flush *cf;
2418
	int ret;
2419
	u64 first_tid = 0;
2420
	u64 last_snap_flush = 0;
2421

2422 2423
	ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;

2424
	list_for_each_entry_reverse(cf, &ci->i_cap_flush_list, i_list) {
2425
		if (cf->is_capsnap) {
2426 2427 2428 2429 2430
			last_snap_flush = cf->tid;
			break;
		}
	}

2431 2432 2433 2434
	list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
		if (cf->tid < first_tid)
			continue;

2435 2436
		cap = ci->i_auth_cap;
		if (!(cap && cap->session == session)) {
2437 2438
			pr_err("%p auth cap %p not mds%d ???\n",
			       inode, cap, session->s_mds);
2439 2440 2441 2442 2443
			break;
		}

		first_tid = cf->tid + 1;

2444
		if (!cf->is_capsnap) {
2445 2446
			struct cap_msg_args arg;

2447 2448
			dout("kick_flushing_caps %p cap %p tid %llu %s\n",
			     inode, cap, cf->tid, ceph_cap_string(cf->caps));
2449
			__prep_cap(&arg, cap, CEPH_CAP_OP_FLUSH,
2450 2451 2452
					 (cf->tid < last_snap_flush ?
					  CEPH_CLIENT_CAPS_PENDING_CAPSNAP : 0),
					  __ceph_caps_used(ci),
2453
					  __ceph_caps_wanted(ci),
2454
					  (cap->issued | cap->implemented),
2455
					  cf->caps, cf->tid, oldest_flush_tid);
2456
			spin_unlock(&ci->i_ceph_lock);
2457
			__send_cap(&arg, ci);
2458 2459 2460 2461 2462 2463 2464 2465
		} else {
			struct ceph_cap_snap *capsnap =
					container_of(cf, struct ceph_cap_snap,
						    cap_flush);
			dout("kick_flushing_caps %p capsnap %p tid %llu %s\n",
			     inode, capsnap, cf->tid,
			     ceph_cap_string(capsnap->dirty));

2466
			refcount_inc(&capsnap->nref);
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480
			spin_unlock(&ci->i_ceph_lock);

			ret = __send_flush_snap(inode, session, capsnap, cap->mseq,
						oldest_flush_tid);
			if (ret < 0) {
				pr_err("kick_flushing_caps: error sending "
					"cap flushsnap, ino (%llx.%llx) "
					"tid %llu follows %llu\n",
					ceph_vinop(inode), cf->tid,
					capsnap->follows);
			}

			ceph_put_cap_snap(capsnap);
		}
2481 2482

		spin_lock(&ci->i_ceph_lock);
2483 2484 2485
	}
}

2486 2487 2488 2489 2490
void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
				   struct ceph_mds_session *session)
{
	struct ceph_inode_info *ci;
	struct ceph_cap *cap;
2491
	u64 oldest_flush_tid;
2492 2493

	dout("early_kick_flushing_caps mds%d\n", session->s_mds);
2494 2495 2496 2497 2498

	spin_lock(&mdsc->cap_dirty_lock);
	oldest_flush_tid = __get_oldest_flush_tid(mdsc);
	spin_unlock(&mdsc->cap_dirty_lock);

2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
	list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
		spin_lock(&ci->i_ceph_lock);
		cap = ci->i_auth_cap;
		if (!(cap && cap->session == session)) {
			pr_err("%p auth cap %p not mds%d ???\n",
				&ci->vfs_inode, cap, session->s_mds);
			spin_unlock(&ci->i_ceph_lock);
			continue;
		}


		/*
		 * if flushing caps were revoked, we re-send the cap flush
		 * in client reconnect stage. This guarantees MDS * processes
		 * the cap flush message before issuing the flushing caps to
		 * other client.
		 */
		if ((cap->issued & ci->i_flushing_caps) !=
		    ci->i_flushing_caps) {
2518 2519 2520 2521 2522 2523
			/* encode_caps_cb() also will reset these sequence
			 * numbers. make sure sequence numbers in cap flush
			 * message match later reconnect message */
			cap->seq = 0;
			cap->issue_seq = 0;
			cap->mseq = 0;
2524 2525
			__kick_flushing_caps(mdsc, session, ci,
					     oldest_flush_tid);
2526 2527
		} else {
			ci->i_ceph_flags |= CEPH_I_KICK_FLUSH;
2528 2529 2530 2531 2532 2533
		}

		spin_unlock(&ci->i_ceph_lock);
	}
}

Sage Weil's avatar
Sage Weil committed
2534 2535 2536 2537
void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
			     struct ceph_mds_session *session)
{
	struct ceph_inode_info *ci;
2538
	struct ceph_cap *cap;
2539
	u64 oldest_flush_tid;
Sage Weil's avatar
Sage Weil committed
2540

2541 2542
	lockdep_assert_held(&session->s_mutex);

Sage Weil's avatar
Sage Weil committed
2543
	dout("kick_flushing_caps mds%d\n", session->s_mds);
2544 2545 2546 2547 2548

	spin_lock(&mdsc->cap_dirty_lock);
	oldest_flush_tid = __get_oldest_flush_tid(mdsc);
	spin_unlock(&mdsc->cap_dirty_lock);

Sage Weil's avatar
Sage Weil committed
2549
	list_for_each_entry(ci, &session->s_cap_flushing, i_flushing_item) {
2550
		spin_lock(&ci->i_ceph_lock);
2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561
		cap = ci->i_auth_cap;
		if (!(cap && cap->session == session)) {
			pr_err("%p auth cap %p not mds%d ???\n",
				&ci->vfs_inode, cap, session->s_mds);
			spin_unlock(&ci->i_ceph_lock);
			continue;
		}
		if (ci->i_ceph_flags & CEPH_I_KICK_FLUSH) {
			__kick_flushing_caps(mdsc, session, ci,
					     oldest_flush_tid);
		}
2562
		spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2563 2564 2565
	}
}

2566 2567
void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
				   struct ceph_inode_info *ci)
2568
{
2569 2570 2571 2572
	struct ceph_mds_client *mdsc = session->s_mdsc;
	struct ceph_cap *cap = ci->i_auth_cap;

	lockdep_assert_held(&ci->i_ceph_lock);
2573

2574
	dout("%s %p flushing %s\n", __func__, &ci->vfs_inode,
2575
	     ceph_cap_string(ci->i_flushing_caps));
2576

2577 2578
	if (!list_empty(&ci->i_cap_flush_list)) {
		u64 oldest_flush_tid;
2579 2580 2581
		spin_lock(&mdsc->cap_dirty_lock);
		list_move_tail(&ci->i_flushing_item,
			       &cap->session->s_cap_flushing);
2582
		oldest_flush_tid = __get_oldest_flush_tid(mdsc);
2583 2584
		spin_unlock(&mdsc->cap_dirty_lock);

2585
		__kick_flushing_caps(mdsc, session, ci, oldest_flush_tid);
2586 2587 2588
	}
}

Sage Weil's avatar
Sage Weil committed
2589 2590 2591 2592 2593

/*
 * Take references to capabilities we hold, so that we don't release
 * them to the MDS prematurely.
 */
2594
void ceph_take_cap_refs(struct ceph_inode_info *ci, int got,
2595
			    bool snap_rwsem_locked)
Sage Weil's avatar
Sage Weil committed
2596
{
2597 2598
	lockdep_assert_held(&ci->i_ceph_lock);

Sage Weil's avatar
Sage Weil committed
2599 2600 2601 2602 2603 2604
	if (got & CEPH_CAP_PIN)
		ci->i_pin_ref++;
	if (got & CEPH_CAP_FILE_RD)
		ci->i_rd_ref++;
	if (got & CEPH_CAP_FILE_CACHE)
		ci->i_rdcache_ref++;
2605 2606
	if (got & CEPH_CAP_FILE_EXCL)
		ci->i_fx_ref++;
2607 2608 2609 2610 2611 2612
	if (got & CEPH_CAP_FILE_WR) {
		if (ci->i_wr_ref == 0 && !ci->i_head_snapc) {
			BUG_ON(!snap_rwsem_locked);
			ci->i_head_snapc = ceph_get_snap_context(
					ci->i_snap_realm->cached_context);
		}
Sage Weil's avatar
Sage Weil committed
2613
		ci->i_wr_ref++;
2614
	}
Sage Weil's avatar
Sage Weil committed
2615
	if (got & CEPH_CAP_FILE_BUFFER) {
2616
		if (ci->i_wb_ref == 0)
2617
			ihold(&ci->vfs_inode);
2618
		ci->i_wb_ref++;
2619
		dout("%s %p wb %d -> %d (?)\n", __func__,
2620
		     &ci->vfs_inode, ci->i_wb_ref-1, ci->i_wb_ref);
Sage Weil's avatar
Sage Weil committed
2621 2622 2623 2624 2625 2626 2627 2628 2629
	}
}

/*
 * Try to grab cap references.  Specify those refs we @want, and the
 * minimal set we @need.  Also include the larger offset we are writing
 * to (when applicable), and check against max_size here as well.
 * Note that caller is responsible for ensuring max_size increases are
 * requested from the MDS.
2630
 *
2631 2632
 * Returns 0 if caps were not able to be acquired (yet), 1 if succeed,
 * or a negative error code. There are 3 speical error codes:
2633 2634 2635
 *  -EAGAIN:  need to sleep but non-blocking is specified
 *  -EFBIG:   ask caller to call check_max_size() and try again.
 *  -EUCLEAN: ask caller to call ceph_renew_caps() and try again.
Sage Weil's avatar
Sage Weil committed
2636
 */
2637
enum {
2638 2639 2640
	/* first 8 bits are reserved for CEPH_FILE_MODE_FOO */
	NON_BLOCKING	= (1 << 8),
	CHECK_FILELOCK	= (1 << 9),
2641 2642
};

2643
static int try_get_cap_refs(struct inode *inode, int need, int want,
2644
			    loff_t endoff, int flags, int *got)
Sage Weil's avatar
Sage Weil committed
2645
{
2646
	struct ceph_inode_info *ci = ceph_inode(inode);
2647
	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
Sage Weil's avatar
Sage Weil committed
2648
	int ret = 0;
2649
	int have, implemented;
2650
	bool snap_rwsem_locked = false;
Sage Weil's avatar
Sage Weil committed
2651 2652 2653

	dout("get_cap_refs %p need %s want %s\n", inode,
	     ceph_cap_string(need), ceph_cap_string(want));
2654

2655
again:
2656
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2657

2658 2659 2660 2661 2662 2663 2664
	if ((flags & CHECK_FILELOCK) &&
	    (ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK)) {
		dout("try_get_cap_refs %p error filelock\n", inode);
		ret = -EIO;
		goto out_unlock;
	}

2665 2666 2667
	/* finish pending truncate */
	while (ci->i_truncate_pending) {
		spin_unlock(&ci->i_ceph_lock);
2668 2669 2670 2671
		if (snap_rwsem_locked) {
			up_read(&mdsc->snap_rwsem);
			snap_rwsem_locked = false;
		}
2672
		__ceph_do_pending_vmtruncate(inode);
2673 2674 2675
		spin_lock(&ci->i_ceph_lock);
	}

Yan, Zheng's avatar
Yan, Zheng committed
2676 2677 2678
	have = __ceph_caps_issued(ci, &implemented);

	if (have & need & CEPH_CAP_FILE_WR) {
Sage Weil's avatar
Sage Weil committed
2679 2680 2681
		if (endoff >= 0 && endoff > (loff_t)ci->i_max_size) {
			dout("get_cap_refs %p endoff %llu > maxsize %llu\n",
			     inode, endoff, ci->i_max_size);
2682
			if (endoff > ci->i_requested_max_size)
2683
				ret = ci->i_auth_cap ? -EFBIG : -EUCLEAN;
2684
			goto out_unlock;
Sage Weil's avatar
Sage Weil committed
2685 2686 2687 2688 2689 2690 2691
		}
		/*
		 * If a sync write is in progress, we must wait, so that we
		 * can get a final snapshot value for size+mtime.
		 */
		if (__ceph_have_pending_cap_snap(ci)) {
			dout("get_cap_refs %p cap_snap_pending\n", inode);
2692
			goto out_unlock;
Sage Weil's avatar
Sage Weil committed
2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708
		}
	}

	if ((have & need) == need) {
		/*
		 * Look at (implemented & ~have & not) so that we keep waiting
		 * on transition from wanted -> needed caps.  This is needed
		 * for WRBUFFER|WR -> WR to avoid a new WR sync write from
		 * going before a prior buffered writeback happens.
		 */
		int not = want & ~(have & need);
		int revoking = implemented & ~have;
		dout("get_cap_refs %p have %s but not %s (revoking %s)\n",
		     inode, ceph_cap_string(have), ceph_cap_string(not),
		     ceph_cap_string(revoking));
		if ((revoking & not) == 0) {
2709 2710 2711 2712 2713 2714 2715 2716
			if (!snap_rwsem_locked &&
			    !ci->i_head_snapc &&
			    (need & CEPH_CAP_FILE_WR)) {
				if (!down_read_trylock(&mdsc->snap_rwsem)) {
					/*
					 * we can not call down_read() when
					 * task isn't in TASK_RUNNING state
					 */
2717
					if (flags & NON_BLOCKING) {
2718
						ret = -EAGAIN;
2719 2720 2721 2722 2723 2724 2725 2726 2727 2728
						goto out_unlock;
					}

					spin_unlock(&ci->i_ceph_lock);
					down_read(&mdsc->snap_rwsem);
					snap_rwsem_locked = true;
					goto again;
				}
				snap_rwsem_locked = true;
			}
2729 2730 2731 2732
			if ((have & want) == want)
				*got = need | want;
			else
				*got = need;
2733
			ceph_take_cap_refs(ci, *got, true);
Sage Weil's avatar
Sage Weil committed
2734 2735 2736
			ret = 1;
		}
	} else {
2737
		int session_readonly = false;
2738
		int mds_wanted;
2739 2740
		if (ci->i_auth_cap &&
		    (need & (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_EXCL))) {
2741 2742 2743 2744 2745 2746
			struct ceph_mds_session *s = ci->i_auth_cap->session;
			spin_lock(&s->s_cap_lock);
			session_readonly = s->s_readonly;
			spin_unlock(&s->s_cap_lock);
		}
		if (session_readonly) {
2747
			dout("get_cap_refs %p need %s but mds%d readonly\n",
2748
			     inode, ceph_cap_string(need), ci->i_auth_cap->mds);
2749
			ret = -EROFS;
2750 2751 2752
			goto out_unlock;
		}

2753
		if (READ_ONCE(mdsc->fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN) {
2754 2755 2756 2757 2758 2759 2760 2761 2762
			dout("get_cap_refs %p forced umount\n", inode);
			ret = -EIO;
			goto out_unlock;
		}
		mds_wanted = __ceph_caps_mds_wanted(ci, false);
		if (need & ~mds_wanted) {
			dout("get_cap_refs %p need %s > mds_wanted %s\n",
			     inode, ceph_cap_string(need),
			     ceph_cap_string(mds_wanted));
2763
			ret = -EUCLEAN;
2764
			goto out_unlock;
2765 2766
		}

2767
		dout("get_cap_refs %p have %s need %s\n", inode,
Sage Weil's avatar
Sage Weil committed
2768 2769
		     ceph_cap_string(have), ceph_cap_string(need));
	}
2770
out_unlock:
2771 2772 2773

	__ceph_touch_fmode(ci, mdsc, flags);

2774
	spin_unlock(&ci->i_ceph_lock);
2775 2776
	if (snap_rwsem_locked)
		up_read(&mdsc->snap_rwsem);
2777

2778 2779 2780 2781 2782
	if (!ret)
		ceph_update_cap_mis(&mdsc->metric);
	else if (ret == 1)
		ceph_update_cap_hit(&mdsc->metric);

Sage Weil's avatar
Sage Weil committed
2783
	dout("get_cap_refs %p ret %d got %s\n", inode,
2784
	     ret, ceph_cap_string(*got));
Sage Weil's avatar
Sage Weil committed
2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798
	return ret;
}

/*
 * Check the offset we are writing up to against our current
 * max_size.  If necessary, tell the MDS we want to write to
 * a larger offset.
 */
static void check_max_size(struct inode *inode, loff_t endoff)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int check = 0;

	/* do we need to explicitly request a larger max_size? */
2799
	spin_lock(&ci->i_ceph_lock);
Yan, Zheng's avatar
Yan, Zheng committed
2800
	if (endoff >= ci->i_max_size && endoff > ci->i_wanted_max_size) {
Sage Weil's avatar
Sage Weil committed
2801 2802 2803 2804
		dout("write %p at large endoff %llu, req max_size\n",
		     inode, endoff);
		ci->i_wanted_max_size = endoff;
	}
Yan, Zheng's avatar
Yan, Zheng committed
2805 2806 2807 2808 2809 2810
	/* duplicate ceph_check_caps()'s logic */
	if (ci->i_auth_cap &&
	    (ci->i_auth_cap->issued & CEPH_CAP_FILE_WR) &&
	    ci->i_wanted_max_size > ci->i_max_size &&
	    ci->i_wanted_max_size > ci->i_requested_max_size)
		check = 1;
2811
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2812 2813 2814 2815
	if (check)
		ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL);
}

2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
static inline int get_used_fmode(int caps)
{
	int fmode = 0;
	if (caps & CEPH_CAP_FILE_RD)
		fmode |= CEPH_FILE_MODE_RD;
	if (caps & CEPH_CAP_FILE_WR)
		fmode |= CEPH_FILE_MODE_WR;
	return fmode;
}

2826
int ceph_try_get_caps(struct inode *inode, int need, int want,
2827
		      bool nonblock, int *got)
2828
{
2829
	int ret, flags;
2830 2831

	BUG_ON(need & ~CEPH_CAP_FILE_RD);
2832 2833 2834 2835 2836 2837 2838 2839
	BUG_ON(want & ~(CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO |
			CEPH_CAP_FILE_SHARED | CEPH_CAP_FILE_EXCL |
			CEPH_CAP_ANY_DIR_OPS));
	if (need) {
		ret = ceph_pool_perm_check(inode, need);
		if (ret < 0)
			return ret;
	}
2840

2841 2842 2843 2844 2845
	flags = get_used_fmode(need | want);
	if (nonblock)
		flags |= NON_BLOCKING;

	ret = try_get_cap_refs(inode, need, want, 0, flags, got);
2846
	/* three special error codes */
2847
	if (ret == -EAGAIN || ret == -EFBIG || ret == -EUCLEAN)
2848 2849
		ret = 0;
	return ret;
2850 2851
}

Sage Weil's avatar
Sage Weil committed
2852 2853 2854 2855 2856
/*
 * Wait for caps, and take cap references.  If we can't get a WR cap
 * due to a small max_size, make sure we check_max_size (and possibly
 * ask the mds) so we don't get hung up indefinitely.
 */
2857
int ceph_get_caps(struct file *filp, int need, int want, loff_t endoff, int *got)
Sage Weil's avatar
Sage Weil committed
2858
{
2859
	struct ceph_file_info *fi = filp->private_data;
2860 2861
	struct inode *inode = file_inode(filp);
	struct ceph_inode_info *ci = ceph_inode(inode);
2862
	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
2863
	int ret, _got, flags;
Sage Weil's avatar
Sage Weil committed
2864

2865
	ret = ceph_pool_perm_check(inode, need);
2866 2867 2868
	if (ret < 0)
		return ret;

2869 2870 2871 2872
	if ((fi->fmode & CEPH_FILE_MODE_WR) &&
	    fi->filp_gen != READ_ONCE(fsc->filp_gen))
		return -EBADF;

2873 2874
	flags = get_used_fmode(need | want);

2875
	while (true) {
2876 2877 2878
		flags &= CEPH_FILE_MODE_MASK;
		if (atomic_read(&fi->num_locks))
			flags |= CHECK_FILELOCK;
2879
		_got = 0;
2880
		ret = try_get_cap_refs(inode, need, want, endoff,
2881
				       flags, &_got);
2882
		WARN_ON_ONCE(ret == -EAGAIN);
2883
		if (!ret) {
2884 2885
			struct ceph_mds_client *mdsc = fsc->mdsc;
			struct cap_wait cw;
2886
			DEFINE_WAIT_FUNC(wait, woken_wake_function);
2887

2888
			cw.ino = ceph_ino(inode);
2889 2890 2891 2892 2893 2894 2895 2896
			cw.tgid = current->tgid;
			cw.need = need;
			cw.want = want;

			spin_lock(&mdsc->caps_list_lock);
			list_add(&cw.list, &mdsc->cap_wait_list);
			spin_unlock(&mdsc->caps_list_lock);

2897 2898
			/* make sure used fmode not timeout */
			ceph_get_fmode(ci, flags, FMODE_WAIT_BIAS);
2899 2900
			add_wait_queue(&ci->i_cap_wq, &wait);

2901
			flags |= NON_BLOCKING;
2902
			while (!(ret = try_get_cap_refs(inode, need, want,
2903
							endoff, flags, &_got))) {
2904 2905 2906 2907
				if (signal_pending(current)) {
					ret = -ERESTARTSYS;
					break;
				}
2908
				wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
2909
			}
2910 2911

			remove_wait_queue(&ci->i_cap_wq, &wait);
2912
			ceph_put_fmode(ci, flags, FMODE_WAIT_BIAS);
2913 2914 2915 2916 2917

			spin_lock(&mdsc->caps_list_lock);
			list_del(&cw.list);
			spin_unlock(&mdsc->caps_list_lock);

2918
			if (ret == -EAGAIN)
2919
				continue;
2920
		}
2921 2922 2923 2924 2925 2926 2927 2928

		if ((fi->fmode & CEPH_FILE_MODE_WR) &&
		    fi->filp_gen != READ_ONCE(fsc->filp_gen)) {
			if (ret >= 0 && _got)
				ceph_put_cap_refs(ci, _got);
			return -EBADF;
		}

2929
		if (ret < 0) {
2930
			if (ret == -EFBIG || ret == -EUCLEAN) {
2931 2932 2933 2934
				int ret2 = ceph_wait_on_async_create(inode);
				if (ret2 < 0)
					return ret2;
			}
2935 2936 2937 2938
			if (ret == -EFBIG) {
				check_max_size(inode, endoff);
				continue;
			}
2939
			if (ret == -EUCLEAN) {
2940
				/* session was killed, try renew caps */
2941
				ret = ceph_renew_caps(inode, flags);
2942 2943 2944
				if (ret == 0)
					continue;
			}
2945
			return ret;
2946
		}
2947

2948 2949
		if (S_ISREG(ci->vfs_inode.i_mode) &&
		    ci->i_inline_version != CEPH_INLINE_NONE &&
2950
		    (_got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) &&
2951
		    i_size_read(inode) > 0) {
2952
			struct page *page =
2953
				find_get_page(inode->i_mapping, 0);
2954
			if (page) {
2955 2956
				bool uptodate = PageUptodate(page);

2957
				put_page(page);
2958 2959
				if (uptodate)
					break;
2960
			}
2961 2962 2963 2964 2965 2966
			/*
			 * drop cap refs first because getattr while
			 * holding * caps refs can cause deadlock.
			 */
			ceph_put_cap_refs(ci, _got);
			_got = 0;
2967

2968 2969 2970 2971
			/*
			 * getattr request will bring inline data into
			 * page cache
			 */
2972
			ret = __ceph_do_getattr(inode, NULL,
2973 2974 2975 2976 2977 2978 2979
						CEPH_STAT_CAP_INLINE_DATA,
						true);
			if (ret < 0)
				return ret;
			continue;
		}
		break;
2980 2981 2982
	}
	*got = _got;
	return 0;
Sage Weil's avatar
Sage Weil committed
2983 2984 2985 2986 2987 2988 2989 2990
}

/*
 * Take cap refs.  Caller must already know we hold at least one ref
 * on the caps in question or we don't know this is safe.
 */
void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps)
{
2991
	spin_lock(&ci->i_ceph_lock);
2992
	ceph_take_cap_refs(ci, caps, false);
2993
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
2994 2995
}

2996 2997 2998 2999 3000

/*
 * drop cap_snap that is not associated with any snapshot.
 * we don't need to send FLUSHSNAP message for it.
 */
3001 3002
static int ceph_try_drop_cap_snap(struct ceph_inode_info *ci,
				  struct ceph_cap_snap *capsnap)
3003 3004 3005 3006 3007
{
	if (!capsnap->need_flush &&
	    !capsnap->writing && !capsnap->dirty_pages) {
		dout("dropping cap_snap %p follows %llu\n",
		     capsnap, capsnap->follows);
3008
		BUG_ON(capsnap->cap_flush.tid > 0);
3009
		ceph_put_snap_context(capsnap->context);
3010 3011 3012
		if (!list_is_last(&capsnap->ci_item, &ci->i_cap_snaps))
			ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;

3013 3014 3015 3016 3017 3018 3019
		list_del(&capsnap->ci_item);
		ceph_put_cap_snap(capsnap);
		return 1;
	}
	return 0;
}

3020 3021 3022 3023 3024 3025
enum put_cap_refs_mode {
	PUT_CAP_REFS_SYNC = 0,
	PUT_CAP_REFS_NO_CHECK,
	PUT_CAP_REFS_ASYNC,
};

Sage Weil's avatar
Sage Weil committed
3026 3027 3028 3029 3030 3031 3032 3033 3034
/*
 * Release cap refs.
 *
 * If we released the last ref on any given cap, call ceph_check_caps
 * to release (or schedule a release).
 *
 * If we are releasing a WR cap (from a sync write), finalize any affected
 * cap_snap, and wake up any waiters.
 */
3035
static void __ceph_put_cap_refs(struct ceph_inode_info *ci, int had,
3036
				enum put_cap_refs_mode mode)
Sage Weil's avatar
Sage Weil committed
3037 3038 3039
{
	struct inode *inode = &ci->vfs_inode;
	int last = 0, put = 0, flushsnaps = 0, wake = 0;
3040
	bool check_flushsnaps = false;
Sage Weil's avatar
Sage Weil committed
3041

3042
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3043 3044 3045 3046 3047 3048 3049 3050
	if (had & CEPH_CAP_PIN)
		--ci->i_pin_ref;
	if (had & CEPH_CAP_FILE_RD)
		if (--ci->i_rd_ref == 0)
			last++;
	if (had & CEPH_CAP_FILE_CACHE)
		if (--ci->i_rdcache_ref == 0)
			last++;
3051 3052 3053
	if (had & CEPH_CAP_FILE_EXCL)
		if (--ci->i_fx_ref == 0)
			last++;
Sage Weil's avatar
Sage Weil committed
3054
	if (had & CEPH_CAP_FILE_BUFFER) {
3055
		if (--ci->i_wb_ref == 0) {
Sage Weil's avatar
Sage Weil committed
3056
			last++;
3057
			/* put the ref held by ceph_take_cap_refs() */
Sage Weil's avatar
Sage Weil committed
3058
			put++;
3059
			check_flushsnaps = true;
Sage Weil's avatar
Sage Weil committed
3060
		}
3061 3062
		dout("put_cap_refs %p wb %d -> %d (?)\n",
		     inode, ci->i_wb_ref+1, ci->i_wb_ref);
Sage Weil's avatar
Sage Weil committed
3063
	}
3064
	if (had & CEPH_CAP_FILE_WR) {
Sage Weil's avatar
Sage Weil committed
3065 3066
		if (--ci->i_wr_ref == 0) {
			last++;
3067
			check_flushsnaps = true;
3068 3069 3070 3071 3072 3073 3074
			if (ci->i_wrbuffer_ref_head == 0 &&
			    ci->i_dirty_caps == 0 &&
			    ci->i_flushing_caps == 0) {
				BUG_ON(!ci->i_head_snapc);
				ceph_put_snap_context(ci->i_head_snapc);
				ci->i_head_snapc = NULL;
			}
3075
			/* see comment in __ceph_remove_cap() */
3076
			if (!__ceph_is_any_real_caps(ci) && ci->i_snap_realm)
3077
				ceph_change_snap_realm(inode, NULL);
Sage Weil's avatar
Sage Weil committed
3078
		}
3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093
	}
	if (check_flushsnaps && __ceph_have_pending_cap_snap(ci)) {
		struct ceph_cap_snap *capsnap =
			list_last_entry(&ci->i_cap_snaps,
					struct ceph_cap_snap,
					ci_item);

		capsnap->writing = 0;
		if (ceph_try_drop_cap_snap(ci, capsnap))
			/* put the ref held by ceph_queue_cap_snap() */
			put++;
		else if (__ceph_finish_cap_snap(ci, capsnap))
			flushsnaps = 1;
		wake = 1;
	}
3094
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3095

3096 3097
	dout("put_cap_refs %p had %s%s%s\n", inode, ceph_cap_string(had),
	     last ? " last" : "", put ? " put" : "");
Sage Weil's avatar
Sage Weil committed
3098

3099 3100
	switch (mode) {
	case PUT_CAP_REFS_SYNC:
3101 3102 3103 3104
		if (last)
			ceph_check_caps(ci, 0, NULL);
		else if (flushsnaps)
			ceph_flush_snaps(ci, NULL);
3105 3106 3107 3108 3109 3110 3111 3112 3113
		break;
	case PUT_CAP_REFS_ASYNC:
		if (last)
			ceph_queue_check_caps(inode);
		else if (flushsnaps)
			ceph_queue_flush_snaps(inode);
		break;
	default:
		break;
3114
	}
Sage Weil's avatar
Sage Weil committed
3115
	if (wake)
3116
		wake_up_all(&ci->i_cap_wq);
3117
	while (put-- > 0)
Sage Weil's avatar
Sage Weil committed
3118 3119 3120
		iput(inode);
}

3121 3122
void ceph_put_cap_refs(struct ceph_inode_info *ci, int had)
{
3123 3124 3125 3126 3127 3128
	__ceph_put_cap_refs(ci, had, PUT_CAP_REFS_SYNC);
}

void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had)
{
	__ceph_put_cap_refs(ci, had, PUT_CAP_REFS_ASYNC);
3129 3130 3131 3132
}

void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, int had)
{
3133
	__ceph_put_cap_refs(ci, had, PUT_CAP_REFS_NO_CHECK);
3134 3135
}

Sage Weil's avatar
Sage Weil committed
3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147
/*
 * Release @nr WRBUFFER refs on dirty pages for the given @snapc snap
 * context.  Adjust per-snap dirty page accounting as appropriate.
 * Once all dirty data for a cap_snap is flushed, flush snapped file
 * metadata back to the MDS.  If we dropped the last ref, call
 * ceph_check_caps.
 */
void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
				struct ceph_snap_context *snapc)
{
	struct inode *inode = &ci->vfs_inode;
	struct ceph_cap_snap *capsnap = NULL;
3148 3149 3150 3151 3152
	int put = 0;
	bool last = false;
	bool found = false;
	bool flush_snaps = false;
	bool complete_capsnap = false;
Sage Weil's avatar
Sage Weil committed
3153

3154
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3155
	ci->i_wrbuffer_ref -= nr;
3156 3157 3158 3159
	if (ci->i_wrbuffer_ref == 0) {
		last = true;
		put++;
	}
Sage Weil's avatar
Sage Weil committed
3160 3161 3162

	if (ci->i_head_snapc == snapc) {
		ci->i_wrbuffer_ref_head -= nr;
3163
		if (ci->i_wrbuffer_ref_head == 0 &&
3164 3165 3166
		    ci->i_wr_ref == 0 &&
		    ci->i_dirty_caps == 0 &&
		    ci->i_flushing_caps == 0) {
3167
			BUG_ON(!ci->i_head_snapc);
Sage Weil's avatar
Sage Weil committed
3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
			ceph_put_snap_context(ci->i_head_snapc);
			ci->i_head_snapc = NULL;
		}
		dout("put_wrbuffer_cap_refs on %p head %d/%d -> %d/%d %s\n",
		     inode,
		     ci->i_wrbuffer_ref+nr, ci->i_wrbuffer_ref_head+nr,
		     ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
		     last ? " LAST" : "");
	} else {
		list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
			if (capsnap->context == snapc) {
3179
				found = true;
Sage Weil's avatar
Sage Weil committed
3180 3181 3182
				break;
			}
		}
3183 3184 3185 3186 3187 3188 3189 3190 3191 3192

		if (!found) {
			/*
			 * The capsnap should already be removed when removing
			 * auth cap in the case of a forced unmount.
			 */
			WARN_ON_ONCE(ci->i_auth_cap);
			goto unlock;
		}

3193 3194
		capsnap->dirty_pages -= nr;
		if (capsnap->dirty_pages == 0) {
3195 3196 3197 3198 3199 3200 3201 3202 3203
			complete_capsnap = true;
			if (!capsnap->writing) {
				if (ceph_try_drop_cap_snap(ci, capsnap)) {
					put++;
				} else {
					ci->i_ceph_flags |= CEPH_I_FLUSH_SNAPS;
					flush_snaps = true;
				}
			}
3204
		}
Sage Weil's avatar
Sage Weil committed
3205
		dout("put_wrbuffer_cap_refs on %p cap_snap %p "
3206
		     " snap %lld %d/%d -> %d/%d %s%s\n",
Sage Weil's avatar
Sage Weil committed
3207 3208 3209 3210
		     inode, capsnap, capsnap->context->seq,
		     ci->i_wrbuffer_ref+nr, capsnap->dirty_pages + nr,
		     ci->i_wrbuffer_ref, capsnap->dirty_pages,
		     last ? " (wrbuffer last)" : "",
3211
		     complete_capsnap ? " (complete capsnap)" : "");
Sage Weil's avatar
Sage Weil committed
3212 3213
	}

3214
unlock:
3215
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3216 3217

	if (last) {
3218
		ceph_check_caps(ci, 0, NULL);
3219
	} else if (flush_snaps) {
3220
		ceph_flush_snaps(ci, NULL);
Sage Weil's avatar
Sage Weil committed
3221
	}
3222 3223
	if (complete_capsnap)
		wake_up_all(&ci->i_cap_wq);
3224
	while (put-- > 0) {
3225
		iput(inode);
3226
	}
Sage Weil's avatar
Sage Weil committed
3227 3228
}

Yan, Zheng's avatar
Yan, Zheng committed
3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239
/*
 * Invalidate unlinked inode's aliases, so we can drop the inode ASAP.
 */
static void invalidate_aliases(struct inode *inode)
{
	struct dentry *dn, *prev = NULL;

	dout("invalidate_aliases inode %p\n", inode);
	d_prune_aliases(inode);
	/*
	 * For non-directory inode, d_find_alias() only returns
J. Bruce Fields's avatar
J. Bruce Fields committed
3240 3241
	 * hashed dentry. After calling d_invalidate(), the
	 * dentry becomes unhashed.
Yan, Zheng's avatar
Yan, Zheng committed
3242
	 *
3243
	 * For directory inode, d_find_alias() can return
J. Bruce Fields's avatar
J. Bruce Fields committed
3244
	 * unhashed dentry. But directory inode should have
Yan, Zheng's avatar
Yan, Zheng committed
3245 3246 3247 3248 3249 3250 3251
	 * one alias at most.
	 */
	while ((dn = d_find_alias(inode))) {
		if (dn == prev) {
			dput(dn);
			break;
		}
3252
		d_invalidate(dn);
Yan, Zheng's avatar
Yan, Zheng committed
3253 3254 3255 3256 3257 3258 3259 3260
		if (prev)
			dput(prev);
		prev = dn;
	}
	if (prev)
		dput(prev);
}

3261 3262 3263 3264 3265 3266
struct cap_extra_info {
	struct ceph_string *pool_ns;
	/* inline data */
	u64 inline_version;
	void *inline_data;
	u32 inline_len;
3267 3268 3269 3270
	/* dirstat */
	bool dirstat_valid;
	u64 nfiles;
	u64 nsubdirs;
3271
	u64 change_attr;
3272 3273
	/* currently issued */
	int issued;
3274
	struct timespec64 btime;
3275 3276
};

Sage Weil's avatar
Sage Weil committed
3277 3278 3279 3280
/*
 * Handle a cap GRANT message from the MDS.  (Note that a GRANT may
 * actually be a revocation if it specifies a smaller cap set.)
 *
3281
 * caller holds s_mutex and i_ceph_lock, we drop both.
Sage Weil's avatar
Sage Weil committed
3282
 */
3283
static void handle_cap_grant(struct inode *inode,
3284
			     struct ceph_mds_session *session,
3285 3286 3287 3288
			     struct ceph_cap *cap,
			     struct ceph_mds_caps *grant,
			     struct ceph_buffer *xattr_buf,
			     struct cap_extra_info *extra_info)
3289
	__releases(ci->i_ceph_lock)
3290
	__releases(session->s_mdsc->snap_rwsem)
Sage Weil's avatar
Sage Weil committed
3291 3292
{
	struct ceph_inode_info *ci = ceph_inode(inode);
3293
	int seq = le32_to_cpu(grant->seq);
Sage Weil's avatar
Sage Weil committed
3294
	int newcaps = le32_to_cpu(grant->caps);
3295
	int used, wanted, dirty;
Sage Weil's avatar
Sage Weil committed
3296 3297
	u64 size = le64_to_cpu(grant->size);
	u64 max_size = le64_to_cpu(grant->max_size);
3298
	unsigned char check_caps = 0;
3299
	bool was_stale = cap->cap_gen < atomic_read(&session->s_cap_gen);
3300 3301 3302 3303 3304
	bool wake = false;
	bool writeback = false;
	bool queue_trunc = false;
	bool queue_invalidate = false;
	bool deleted_inode = false;
3305
	bool fill_inline = false;
Sage Weil's avatar
Sage Weil committed
3306

3307
	dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n",
3308
	     inode, cap, session->s_mds, seq, ceph_cap_string(newcaps));
Sage Weil's avatar
Sage Weil committed
3309
	dout(" size %llu max_size %llu, i_size %llu\n", size, max_size,
3310
		i_size_read(inode));
Sage Weil's avatar
Sage Weil committed
3311

3312

Sage Weil's avatar
Sage Weil committed
3313 3314 3315 3316 3317
	/*
	 * If CACHE is being revoked, and we have no dirty buffers,
	 * try to invalidate (once).  (If there are dirty buffers, we
	 * will invalidate _after_ writeback.)
	 */
3318
	if (S_ISREG(inode->i_mode) && /* don't invalidate readdir cache */
Yan, Zheng's avatar
Yan, Zheng committed
3319
	    ((cap->issued & ~newcaps) & CEPH_CAP_FILE_CACHE) &&
3320
	    (newcaps & CEPH_CAP_FILE_LAZYIO) == 0 &&
3321
	    !(ci->i_wrbuffer_ref || ci->i_wb_ref)) {
3322
		if (try_nonblocking_invalidate(inode)) {
Sage Weil's avatar
Sage Weil committed
3323 3324 3325
			/* there were locked pages.. invalidate later
			   in a separate thread. */
			if (ci->i_rdcache_revoking != ci->i_rdcache_gen) {
3326
				queue_invalidate = true;
Sage Weil's avatar
Sage Weil committed
3327 3328 3329 3330 3331
				ci->i_rdcache_revoking = ci->i_rdcache_gen;
			}
		}
	}

3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349
	if (was_stale)
		cap->issued = cap->implemented = CEPH_CAP_PIN;

	/*
	 * auth mds of the inode changed. we received the cap export message,
	 * but still haven't received the cap import message. handle_cap_export
	 * updated the new auth MDS' cap.
	 *
	 * "ceph_seq_cmp(seq, cap->seq) <= 0" means we are processing a message
	 * that was sent before the cap import message. So don't remove caps.
	 */
	if (ceph_seq_cmp(seq, cap->seq) <= 0) {
		WARN_ON(cap != ci->i_auth_cap);
		WARN_ON(cap->cap_id != le64_to_cpu(grant->cap_id));
		seq = cap->seq;
		newcaps |= cap->issued;
	}

Sage Weil's avatar
Sage Weil committed
3350
	/* side effects now are allowed */
3351
	cap->cap_gen = atomic_read(&session->s_cap_gen);
3352
	cap->seq = seq;
Sage Weil's avatar
Sage Weil committed
3353 3354 3355

	__check_cap_issue(ci, cap, newcaps);

3356 3357
	inode_set_max_iversion_raw(inode, extra_info->change_attr);

3358
	if ((newcaps & CEPH_CAP_AUTH_SHARED) &&
3359
	    (extra_info->issued & CEPH_CAP_AUTH_EXCL) == 0) {
3360 3361 3362 3363 3364 3365 3366
		umode_t mode = le32_to_cpu(grant->mode);

		if (inode_wrong_type(inode, mode))
			pr_warn_once("inode type changed! (ino %llx.%llx is 0%o, mds says 0%o)\n",
				     ceph_vinop(inode), inode->i_mode, mode);
		else
			inode->i_mode = mode;
3367 3368
		inode->i_uid = make_kuid(&init_user_ns, le32_to_cpu(grant->uid));
		inode->i_gid = make_kgid(&init_user_ns, le32_to_cpu(grant->gid));
3369
		ci->i_btime = extra_info->btime;
Sage Weil's avatar
Sage Weil committed
3370
		dout("%p mode 0%o uid.gid %d.%d\n", inode, inode->i_mode,
3371 3372
		     from_kuid(&init_user_ns, inode->i_uid),
		     from_kgid(&init_user_ns, inode->i_gid));
Sage Weil's avatar
Sage Weil committed
3373 3374
	}

3375
	if ((newcaps & CEPH_CAP_LINK_SHARED) &&
3376
	    (extra_info->issued & CEPH_CAP_LINK_EXCL) == 0) {
3377
		set_nlink(inode, le32_to_cpu(grant->nlink));
Yan, Zheng's avatar
Yan, Zheng committed
3378 3379
		if (inode->i_nlink == 0 &&
		    (newcaps & (CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL)))
3380
			deleted_inode = true;
Yan, Zheng's avatar
Yan, Zheng committed
3381
	}
Sage Weil's avatar
Sage Weil committed
3382

3383 3384
	if ((extra_info->issued & CEPH_CAP_XATTR_EXCL) == 0 &&
	    grant->xattr_len) {
Sage Weil's avatar
Sage Weil committed
3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
		int len = le32_to_cpu(grant->xattr_len);
		u64 version = le64_to_cpu(grant->xattr_version);

		if (version > ci->i_xattrs.version) {
			dout(" got new xattrs v%llu on %p len %d\n",
			     version, inode, len);
			if (ci->i_xattrs.blob)
				ceph_buffer_put(ci->i_xattrs.blob);
			ci->i_xattrs.blob = ceph_buffer_get(xattr_buf);
			ci->i_xattrs.version = version;
Guangliang Zhao's avatar
Guangliang Zhao committed
3395
			ceph_forget_all_cached_acls(inode);
Yan, Zheng's avatar
Yan, Zheng committed
3396
			ceph_security_invalidate_secctx(inode);
Sage Weil's avatar
Sage Weil committed
3397 3398 3399
		}
	}

3400
	if (newcaps & CEPH_CAP_ANY_RD) {
3401
		struct timespec64 mtime, atime, ctime;
3402
		/* ctime/mtime/atime? */
3403 3404 3405
		ceph_decode_timespec64(&mtime, &grant->mtime);
		ceph_decode_timespec64(&atime, &grant->atime);
		ceph_decode_timespec64(&ctime, &grant->ctime);
3406
		ceph_fill_file_time(inode, extra_info->issued,
3407 3408 3409 3410
				    le32_to_cpu(grant->time_warp_seq),
				    &ctime, &mtime, &atime);
	}

3411 3412 3413 3414 3415
	if ((newcaps & CEPH_CAP_FILE_SHARED) && extra_info->dirstat_valid) {
		ci->i_files = extra_info->nfiles;
		ci->i_subdirs = extra_info->nsubdirs;
	}

3416 3417
	if (newcaps & (CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR)) {
		/* file layout may have changed */
3418
		s64 old_pool = ci->i_layout.pool_id;
3419 3420
		struct ceph_string *old_ns;

3421
		ceph_file_layout_from_legacy(&ci->i_layout, &grant->layout);
3422 3423
		old_ns = rcu_dereference_protected(ci->i_layout.pool_ns,
					lockdep_is_held(&ci->i_ceph_lock));
3424
		rcu_assign_pointer(ci->i_layout.pool_ns, extra_info->pool_ns);
3425

3426 3427
		if (ci->i_layout.pool_id != old_pool ||
		    extra_info->pool_ns != old_ns)
3428
			ci->i_ceph_flags &= ~CEPH_I_POOL_PERM;
3429

3430
		extra_info->pool_ns = old_ns;
3431

3432
		/* size/truncate_seq? */
3433
		queue_trunc = ceph_fill_file_size(inode, extra_info->issued,
3434 3435 3436
					le32_to_cpu(grant->truncate_seq),
					le64_to_cpu(grant->truncate_size),
					size);
3437 3438 3439 3440
	}

	if (ci->i_auth_cap == cap && (newcaps & CEPH_CAP_ANY_FILE_WR)) {
		if (max_size != ci->i_max_size) {
3441 3442 3443 3444 3445 3446 3447
			dout("max_size %lld -> %llu\n",
			     ci->i_max_size, max_size);
			ci->i_max_size = max_size;
			if (max_size >= ci->i_wanted_max_size) {
				ci->i_wanted_max_size = 0;  /* reset */
				ci->i_requested_max_size = 0;
			}
3448
			wake = true;
Sage Weil's avatar
Sage Weil committed
3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
		}
	}

	/* check cap bits */
	wanted = __ceph_caps_wanted(ci);
	used = __ceph_caps_used(ci);
	dirty = __ceph_caps_dirty(ci);
	dout(" my wanted = %s, used = %s, dirty %s\n",
	     ceph_cap_string(wanted),
	     ceph_cap_string(used),
	     ceph_cap_string(dirty));
3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473

	if ((was_stale || le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) &&
	    (wanted & ~(cap->mds_wanted | newcaps))) {
		/*
		 * If mds is importing cap, prior cap messages that update
		 * 'wanted' may get dropped by mds (migrate seq mismatch).
		 *
		 * We don't send cap message to update 'wanted' if what we
		 * want are already issued. If mds revokes caps, cap message
		 * that releases caps also tells mds what we want. But if
		 * caps got revoked by mds forcedly (session stale). We may
		 * haven't told mds what we want.
		 */
		check_caps = 1;
Sage Weil's avatar
Sage Weil committed
3474 3475 3476 3477
	}

	/* revocation, grant, or no-op? */
	if (cap->issued & ~newcaps) {
3478 3479 3480 3481 3482 3483
		int revoking = cap->issued & ~newcaps;

		dout("revocation: %s -> %s (revoking %s)\n",
		     ceph_cap_string(cap->issued),
		     ceph_cap_string(newcaps),
		     ceph_cap_string(revoking));
3484 3485
		if (S_ISREG(inode->i_mode) &&
		    (revoking & used & CEPH_CAP_FILE_BUFFER))
3486
			writeback = true;  /* initiate writeback; will delay ack */
3487 3488 3489
		else if (queue_invalidate &&
			 revoking == CEPH_CAP_FILE_CACHE &&
			 (newcaps & CEPH_CAP_FILE_LAZYIO) == 0)
3490 3491 3492 3493 3494
			; /* do nothing yet, invalidation will be queued */
		else if (cap == ci->i_auth_cap)
			check_caps = 1; /* check auth cap only */
		else
			check_caps = 2; /* check all caps */
Sage Weil's avatar
Sage Weil committed
3495
		cap->issued = newcaps;
3496
		cap->implemented |= newcaps;
Sage Weil's avatar
Sage Weil committed
3497 3498 3499 3500 3501 3502
	} else if (cap->issued == newcaps) {
		dout("caps unchanged: %s -> %s\n",
		     ceph_cap_string(cap->issued), ceph_cap_string(newcaps));
	} else {
		dout("grant: %s -> %s\n", ceph_cap_string(cap->issued),
		     ceph_cap_string(newcaps));
3503 3504 3505 3506 3507
		/* non-auth MDS is revoking the newly grant caps ? */
		if (cap == ci->i_auth_cap &&
		    __ceph_caps_revoking_other(ci, cap, newcaps))
		    check_caps = 2;

Sage Weil's avatar
Sage Weil committed
3508 3509 3510 3511
		cap->issued = newcaps;
		cap->implemented |= newcaps; /* add bits only, to
					      * avoid stepping on a
					      * pending revocation */
3512
		wake = true;
Sage Weil's avatar
Sage Weil committed
3513
	}
3514
	BUG_ON(cap->issued & ~cap->implemented);
Sage Weil's avatar
Sage Weil committed
3515

3516 3517 3518
	if (extra_info->inline_version > 0 &&
	    extra_info->inline_version >= ci->i_inline_version) {
		ci->i_inline_version = extra_info->inline_version;
3519 3520 3521 3522 3523
		if (ci->i_inline_version != CEPH_INLINE_NONE &&
		    (newcaps & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)))
			fill_inline = true;
	}

3524 3525
	if (ci->i_auth_cap == cap &&
	    le32_to_cpu(grant->op) == CEPH_CAP_OP_IMPORT) {
3526
		if (newcaps & ~extra_info->issued)
3527
			wake = true;
3528 3529 3530 3531 3532 3533 3534 3535

		if (ci->i_requested_max_size > max_size ||
		    !(le32_to_cpu(grant->wanted) & CEPH_CAP_ANY_FILE_WR)) {
			/* re-request max_size if necessary */
			ci->i_requested_max_size = 0;
			wake = true;
		}

3536 3537
		ceph_kick_flushing_inode_caps(session, ci);
		spin_unlock(&ci->i_ceph_lock);
3538
		up_read(&session->s_mdsc->snap_rwsem);
3539 3540
	} else {
		spin_unlock(&ci->i_ceph_lock);
3541 3542
	}

3543
	if (fill_inline)
3544 3545
		ceph_fill_inline_data(inode, NULL, extra_info->inline_data,
				      extra_info->inline_len);
3546

3547
	if (queue_trunc)
3548 3549
		ceph_queue_vmtruncate(inode);

3550
	if (writeback)
Sage Weil's avatar
Sage Weil committed
3551 3552 3553 3554 3555
		/*
		 * queue inode for writeback: we can't actually call
		 * filemap_write_and_wait, etc. from message handler
		 * context.
		 */
3556 3557 3558
		ceph_queue_writeback(inode);
	if (queue_invalidate)
		ceph_queue_invalidate(inode);
Yan, Zheng's avatar
Yan, Zheng committed
3559 3560
	if (deleted_inode)
		invalidate_aliases(inode);
Sage Weil's avatar
Sage Weil committed
3561
	if (wake)
3562
		wake_up_all(&ci->i_cap_wq);
3563

3564
	mutex_unlock(&session->s_mutex);
3565
	if (check_caps == 1)
3566
		ceph_check_caps(ci, CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL,
3567 3568
				session);
	else if (check_caps == 2)
3569
		ceph_check_caps(ci, CHECK_CAPS_NOINVAL, session);
Sage Weil's avatar
Sage Weil committed
3570 3571 3572 3573 3574 3575
}

/*
 * Handle FLUSH_ACK from MDS, indicating that metadata we sent to the
 * MDS has been safely committed.
 */
3576
static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
Sage Weil's avatar
Sage Weil committed
3577 3578 3579
				 struct ceph_mds_caps *m,
				 struct ceph_mds_session *session,
				 struct ceph_cap *cap)
3580
	__releases(ci->i_ceph_lock)
Sage Weil's avatar
Sage Weil committed
3581 3582
{
	struct ceph_inode_info *ci = ceph_inode(inode);
3583
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
3584
	struct ceph_cap_flush *cf, *tmp_cf;
3585
	LIST_HEAD(to_remove);
Sage Weil's avatar
Sage Weil committed
3586 3587 3588
	unsigned seq = le32_to_cpu(m->seq);
	int dirty = le32_to_cpu(m->dirty);
	int cleaned = 0;
3589
	bool drop = false;
3590 3591
	bool wake_ci = false;
	bool wake_mdsc = false;
Sage Weil's avatar
Sage Weil committed
3592

3593
	list_for_each_entry_safe(cf, tmp_cf, &ci->i_cap_flush_list, i_list) {
3594
		/* Is this the one that was flushed? */
3595 3596
		if (cf->tid == flush_tid)
			cleaned = cf->caps;
3597 3598

		/* Is this a capsnap? */
3599
		if (cf->is_capsnap)
3600
			continue;
3601

3602
		if (cf->tid <= flush_tid) {
3603 3604 3605 3606
			/*
			 * An earlier or current tid. The FLUSH_ACK should
			 * represent a superset of this flush's caps.
			 */
3607
			wake_ci |= __detach_cap_flush_from_ci(ci, cf);
3608
			list_add_tail(&cf->i_list, &to_remove);
3609
		} else {
3610 3611 3612 3613
			/*
			 * This is a later one. Any caps in it are still dirty
			 * so don't count them as cleaned.
			 */
3614 3615 3616 3617 3618
			cleaned &= ~cf->caps;
			if (!cleaned)
				break;
		}
	}
Sage Weil's avatar
Sage Weil committed
3619 3620 3621 3622 3623 3624 3625

	dout("handle_cap_flush_ack inode %p mds%d seq %d on %s cleaned %s,"
	     " flushing %s -> %s\n",
	     inode, session->s_mds, seq, ceph_cap_string(dirty),
	     ceph_cap_string(cleaned), ceph_cap_string(ci->i_flushing_caps),
	     ceph_cap_string(ci->i_flushing_caps & ~cleaned));

3626
	if (list_empty(&to_remove) && !cleaned)
Sage Weil's avatar
Sage Weil committed
3627 3628 3629 3630 3631
		goto out;

	ci->i_flushing_caps &= ~cleaned;

	spin_lock(&mdsc->cap_dirty_lock);
3632

3633 3634
	list_for_each_entry(cf, &to_remove, i_list)
		wake_mdsc |= __detach_cap_flush_from_mdsc(mdsc, cf);
3635

Sage Weil's avatar
Sage Weil committed
3636
	if (ci->i_flushing_caps == 0) {
3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
		if (list_empty(&ci->i_cap_flush_list)) {
			list_del_init(&ci->i_flushing_item);
			if (!list_empty(&session->s_cap_flushing)) {
				dout(" mds%d still flushing cap on %p\n",
				     session->s_mds,
				     &list_first_entry(&session->s_cap_flushing,
						struct ceph_inode_info,
						i_flushing_item)->vfs_inode);
			}
		}
Sage Weil's avatar
Sage Weil committed
3647 3648
		mdsc->num_cap_flushing--;
		dout(" inode %p now !flushing\n", inode);
3649 3650 3651 3652

		if (ci->i_dirty_caps == 0) {
			dout(" inode %p now clean\n", inode);
			BUG_ON(!list_empty(&ci->i_dirty_item));
3653
			drop = true;
3654 3655
			if (ci->i_wr_ref == 0 &&
			    ci->i_wrbuffer_ref_head == 0) {
3656 3657 3658 3659
				BUG_ON(!ci->i_head_snapc);
				ceph_put_snap_context(ci->i_head_snapc);
				ci->i_head_snapc = NULL;
			}
3660 3661
		} else {
			BUG_ON(list_empty(&ci->i_dirty_item));
3662
		}
Sage Weil's avatar
Sage Weil committed
3663 3664 3665 3666
	}
	spin_unlock(&mdsc->cap_dirty_lock);

out:
3667
	spin_unlock(&ci->i_ceph_lock);
3668 3669 3670

	while (!list_empty(&to_remove)) {
		cf = list_first_entry(&to_remove,
3671
				      struct ceph_cap_flush, i_list);
3672 3673 3674
		list_del_init(&cf->i_list);
		if (!cf->is_capsnap)
			ceph_free_cap_flush(cf);
3675
	}
3676 3677 3678 3679 3680

	if (wake_ci)
		wake_up_all(&ci->i_cap_wq);
	if (wake_mdsc)
		wake_up_all(&mdsc->cap_flushing_wq);
3681
	if (drop)
Sage Weil's avatar
Sage Weil committed
3682 3683 3684
		iput(inode);
}

3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721
void __ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
			   bool *wake_ci, bool *wake_mdsc)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
	bool ret;

	lockdep_assert_held(&ci->i_ceph_lock);

	dout("removing capsnap %p, inode %p ci %p\n", capsnap, inode, ci);

	list_del_init(&capsnap->ci_item);
	ret = __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
	if (wake_ci)
		*wake_ci = ret;

	spin_lock(&mdsc->cap_dirty_lock);
	if (list_empty(&ci->i_cap_flush_list))
		list_del_init(&ci->i_flushing_item);

	ret = __detach_cap_flush_from_mdsc(mdsc, &capsnap->cap_flush);
	if (wake_mdsc)
		*wake_mdsc = ret;
	spin_unlock(&mdsc->cap_dirty_lock);
}

void ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
			 bool *wake_ci, bool *wake_mdsc)
{
	struct ceph_inode_info *ci = ceph_inode(inode);

	lockdep_assert_held(&ci->i_ceph_lock);

	WARN_ON_ONCE(capsnap->dirty_pages || capsnap->writing);
	__ceph_remove_capsnap(inode, capsnap, wake_ci, wake_mdsc);
}

Sage Weil's avatar
Sage Weil committed
3722 3723 3724 3725 3726 3727
/*
 * Handle FLUSHSNAP_ACK.  MDS has flushed snap data to disk and we can
 * throw away our cap_snap.
 *
 * Caller hold s_mutex.
 */
3728
static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
Sage Weil's avatar
Sage Weil committed
3729 3730 3731 3732
				     struct ceph_mds_caps *m,
				     struct ceph_mds_session *session)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
3733
	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
Sage Weil's avatar
Sage Weil committed
3734 3735
	u64 follows = le64_to_cpu(m->snap_follows);
	struct ceph_cap_snap *capsnap;
3736 3737 3738
	bool flushed = false;
	bool wake_ci = false;
	bool wake_mdsc = false;
Sage Weil's avatar
Sage Weil committed
3739 3740 3741 3742

	dout("handle_cap_flushsnap_ack inode %p ci %p mds%d follows %lld\n",
	     inode, ci, session->s_mds, follows);

3743
	spin_lock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3744 3745
	list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
		if (capsnap->follows == follows) {
3746
			if (capsnap->cap_flush.tid != flush_tid) {
Sage Weil's avatar
Sage Weil committed
3747 3748
				dout(" cap_snap %p follows %lld tid %lld !="
				     " %lld\n", capsnap, follows,
3749
				     flush_tid, capsnap->cap_flush.tid);
Sage Weil's avatar
Sage Weil committed
3750 3751
				break;
			}
3752
			flushed = true;
Sage Weil's avatar
Sage Weil committed
3753 3754 3755 3756 3757 3758
			break;
		} else {
			dout(" skipping cap_snap %p follows %lld\n",
			     capsnap, capsnap->follows);
		}
	}
3759 3760
	if (flushed)
		ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc);
3761
	spin_unlock(&ci->i_ceph_lock);
3762

3763 3764 3765
	if (flushed) {
		ceph_put_snap_context(capsnap->context);
		ceph_put_cap_snap(capsnap);
3766 3767 3768 3769
		if (wake_ci)
			wake_up_all(&ci->i_cap_wq);
		if (wake_mdsc)
			wake_up_all(&mdsc->cap_flushing_wq);
Sage Weil's avatar
Sage Weil committed
3770
		iput(inode);
3771
	}
Sage Weil's avatar
Sage Weil committed
3772 3773 3774 3775 3776 3777 3778
}

/*
 * Handle TRUNC from MDS, indicating file truncation.
 *
 * caller hold s_mutex.
 */
3779
static bool handle_cap_trunc(struct inode *inode,
Sage Weil's avatar
Sage Weil committed
3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
			     struct ceph_mds_caps *trunc,
			     struct ceph_mds_session *session)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int mds = session->s_mds;
	int seq = le32_to_cpu(trunc->seq);
	u32 truncate_seq = le32_to_cpu(trunc->truncate_seq);
	u64 truncate_size = le64_to_cpu(trunc->truncate_size);
	u64 size = le64_to_cpu(trunc->size);
	int implemented = 0;
	int dirty = __ceph_caps_dirty(ci);
	int issued = __ceph_caps_issued(ceph_inode(inode), &implemented);
3792 3793 3794
	bool queue_trunc = false;

	lockdep_assert_held(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
3795 3796 3797 3798 3799 3800 3801

	issued |= implemented | dirty;

	dout("handle_cap_trunc inode %p mds%d seq %d to %lld seq %d\n",
	     inode, mds, seq, truncate_size, truncate_seq);
	queue_trunc = ceph_fill_file_size(inode, issued,
					  truncate_seq, truncate_size, size);
3802
	return queue_trunc;
Sage Weil's avatar
Sage Weil committed
3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813
}

/*
 * Handle EXPORT from MDS.  Cap is being migrated _from_ this mds to a
 * different one.  If we are the most recent migration we've seen (as
 * indicated by mseq), make note of the migrating cap bits for the
 * duration (until we see the corresponding IMPORT).
 *
 * caller holds s_mutex
 */
static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
3814 3815
			      struct ceph_mds_cap_peer *ph,
			      struct ceph_mds_session *session)
Sage Weil's avatar
Sage Weil committed
3816
{
3817
	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
3818
	struct ceph_mds_session *tsession = NULL;
3819
	struct ceph_cap *cap, *tcap, *new_cap = NULL;
Sage Weil's avatar
Sage Weil committed
3820
	struct ceph_inode_info *ci = ceph_inode(inode);
3821
	u64 t_cap_id;
Sage Weil's avatar
Sage Weil committed
3822
	unsigned mseq = le32_to_cpu(ex->migrate_seq);
3823 3824 3825
	unsigned t_seq, t_mseq;
	int target, issued;
	int mds = session->s_mds;
Sage Weil's avatar
Sage Weil committed
3826

3827 3828 3829 3830 3831 3832 3833 3834 3835
	if (ph) {
		t_cap_id = le64_to_cpu(ph->cap_id);
		t_seq = le32_to_cpu(ph->seq);
		t_mseq = le32_to_cpu(ph->mseq);
		target = le32_to_cpu(ph->mds);
	} else {
		t_cap_id = t_seq = t_mseq = 0;
		target = -1;
	}
Sage Weil's avatar
Sage Weil committed
3836

3837 3838 3839
	dout("handle_cap_export inode %p ci %p mds%d mseq %d target %d\n",
	     inode, ci, mds, mseq, target);
retry:
3840
	spin_lock(&ci->i_ceph_lock);
3841
	cap = __get_cap_for_mds(ci, mds);
3842
	if (!cap || cap->cap_id != le64_to_cpu(ex->cap_id))
3843
		goto out_unlock;
Sage Weil's avatar
Sage Weil committed
3844

3845
	if (target < 0) {
3846
		ceph_remove_cap(cap, false);
3847
		goto out_unlock;
Sage Weil's avatar
Sage Weil committed
3848 3849
	}

3850 3851 3852 3853
	/*
	 * now we know we haven't received the cap import message yet
	 * because the exported cap still exist.
	 */
3854

3855
	issued = cap->issued;
3856 3857 3858 3859 3860 3861 3862 3863
	if (issued != cap->implemented)
		pr_err_ratelimited("handle_cap_export: issued != implemented: "
				"ino (%llx.%llx) mds%d seq %d mseq %d "
				"issued %s implemented %s\n",
				ceph_vinop(inode), mds, cap->seq, cap->mseq,
				ceph_cap_string(issued),
				ceph_cap_string(cap->implemented));

3864 3865 3866 3867

	tcap = __get_cap_for_mds(ci, target);
	if (tcap) {
		/* already have caps from the target */
3868
		if (tcap->cap_id == t_cap_id &&
3869 3870 3871 3872 3873 3874 3875
		    ceph_seq_cmp(tcap->seq, t_seq) < 0) {
			dout(" updating import cap %p mds%d\n", tcap, target);
			tcap->cap_id = t_cap_id;
			tcap->seq = t_seq - 1;
			tcap->issue_seq = t_seq - 1;
			tcap->issued |= issued;
			tcap->implemented |= issued;
3876
			if (cap == ci->i_auth_cap) {
3877
				ci->i_auth_cap = tcap;
3878
				change_auth_cap_ses(ci, tcap->session);
3879
			}
Sage Weil's avatar
Sage Weil committed
3880
		}
3881
		ceph_remove_cap(cap, false);
3882
		goto out_unlock;
3883
	} else if (tsession) {
3884
		/* add placeholder for the export tagert */
3885
		int flag = (cap == ci->i_auth_cap) ? CEPH_CAP_FLAG_AUTH : 0;
3886
		tcap = new_cap;
3887
		ceph_add_cap(inode, tsession, t_cap_id, issued, 0,
3888 3889
			     t_seq - 1, t_mseq, (u64)-1, flag, &new_cap);

3890 3891 3892 3893 3894 3895 3896 3897
		if (!list_empty(&ci->i_cap_flush_list) &&
		    ci->i_auth_cap == tcap) {
			spin_lock(&mdsc->cap_dirty_lock);
			list_move_tail(&ci->i_flushing_item,
				       &tcap->session->s_cap_flushing);
			spin_unlock(&mdsc->cap_dirty_lock);
		}

3898
		ceph_remove_cap(cap, false);
3899
		goto out_unlock;
Sage Weil's avatar
Sage Weil committed
3900 3901
	}

3902
	spin_unlock(&ci->i_ceph_lock);
3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916
	mutex_unlock(&session->s_mutex);

	/* open target session */
	tsession = ceph_mdsc_open_export_target_session(mdsc, target);
	if (!IS_ERR(tsession)) {
		if (mds > target) {
			mutex_lock(&session->s_mutex);
			mutex_lock_nested(&tsession->s_mutex,
					  SINGLE_DEPTH_NESTING);
		} else {
			mutex_lock(&tsession->s_mutex);
			mutex_lock_nested(&session->s_mutex,
					  SINGLE_DEPTH_NESTING);
		}
3917
		new_cap = ceph_get_cap(mdsc, NULL);
3918 3919 3920 3921
	} else {
		WARN_ON(1);
		tsession = NULL;
		target = -1;
3922
		mutex_lock(&session->s_mutex);
3923 3924 3925 3926 3927 3928 3929 3930 3931 3932
	}
	goto retry;

out_unlock:
	spin_unlock(&ci->i_ceph_lock);
	mutex_unlock(&session->s_mutex);
	if (tsession) {
		mutex_unlock(&tsession->s_mutex);
		ceph_put_mds_session(tsession);
	}
3933 3934
	if (new_cap)
		ceph_put_cap(mdsc, new_cap);
Sage Weil's avatar
Sage Weil committed
3935 3936 3937
}

/*
3938
 * Handle cap IMPORT.
Sage Weil's avatar
Sage Weil committed
3939
 *
3940
 * caller holds s_mutex. acquires i_ceph_lock
Sage Weil's avatar
Sage Weil committed
3941 3942 3943
 */
static void handle_cap_import(struct ceph_mds_client *mdsc,
			      struct inode *inode, struct ceph_mds_caps *im,
3944
			      struct ceph_mds_cap_peer *ph,
Sage Weil's avatar
Sage Weil committed
3945
			      struct ceph_mds_session *session,
3946
			      struct ceph_cap **target_cap, int *old_issued)
Sage Weil's avatar
Sage Weil committed
3947 3948
{
	struct ceph_inode_info *ci = ceph_inode(inode);
3949
	struct ceph_cap *cap, *ocap, *new_cap = NULL;
Sage Weil's avatar
Sage Weil committed
3950
	int mds = session->s_mds;
3951 3952
	int issued;
	unsigned caps = le32_to_cpu(im->caps);
Sage Weil's avatar
Sage Weil committed
3953 3954 3955 3956 3957
	unsigned wanted = le32_to_cpu(im->wanted);
	unsigned seq = le32_to_cpu(im->seq);
	unsigned mseq = le32_to_cpu(im->migrate_seq);
	u64 realmino = le64_to_cpu(im->realm);
	u64 cap_id = le64_to_cpu(im->cap_id);
3958 3959
	u64 p_cap_id;
	int peer;
Sage Weil's avatar
Sage Weil committed
3960

3961 3962 3963 3964 3965 3966 3967
	if (ph) {
		p_cap_id = le64_to_cpu(ph->cap_id);
		peer = le32_to_cpu(ph->mds);
	} else {
		p_cap_id = 0;
		peer = -1;
	}
3968

3969 3970
	dout("handle_cap_import inode %p ci %p mds%d mseq %d peer %d\n",
	     inode, ci, mds, mseq, peer);
3971 3972 3973 3974 3975 3976
retry:
	cap = __get_cap_for_mds(ci, mds);
	if (!cap) {
		if (!new_cap) {
			spin_unlock(&ci->i_ceph_lock);
			new_cap = ceph_get_cap(mdsc, NULL);
3977
			spin_lock(&ci->i_ceph_lock);
3978 3979
			goto retry;
		}
3980 3981 3982 3983 3984 3985
		cap = new_cap;
	} else {
		if (new_cap) {
			ceph_put_cap(mdsc, new_cap);
			new_cap = NULL;
		}
3986 3987
	}

3988 3989 3990
	__ceph_caps_issued(ci, &issued);
	issued |= __ceph_caps_dirty(ci);

3991
	ceph_add_cap(inode, session, cap_id, caps, wanted, seq, mseq,
3992 3993
		     realmino, CEPH_CAP_FLAG_AUTH, &new_cap);

3994 3995
	ocap = peer >= 0 ? __get_cap_for_mds(ci, peer) : NULL;
	if (ocap && ocap->cap_id == p_cap_id) {
3996
		dout(" remove export cap %p mds%d flags %d\n",
3997
		     ocap, peer, ph->flags);
3998
		if ((ph->flags & CEPH_CAP_FLAG_AUTH) &&
3999 4000
		    (ocap->seq != le32_to_cpu(ph->seq) ||
		     ocap->mseq != le32_to_cpu(ph->mseq))) {
4001 4002 4003 4004 4005 4006 4007
			pr_err_ratelimited("handle_cap_import: "
					"mismatched seq/mseq: ino (%llx.%llx) "
					"mds%d seq %d mseq %d importer mds%d "
					"has peer seq %d mseq %d\n",
					ceph_vinop(inode), peer, ocap->seq,
					ocap->mseq, mds, le32_to_cpu(ph->seq),
					le32_to_cpu(ph->mseq));
4008
		}
4009
		ceph_remove_cap(ocap, (ph->flags & CEPH_CAP_FLAG_RELEASE));
Sage Weil's avatar
Sage Weil committed
4010 4011
	}

4012 4013
	*old_issued = issued;
	*target_cap = cap;
Sage Weil's avatar
Sage Weil committed
4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026
}

/*
 * Handle a caps message from the MDS.
 *
 * Identify the appropriate session, inode, and call the right handler
 * based on the cap op.
 */
void ceph_handle_caps(struct ceph_mds_session *session,
		      struct ceph_msg *msg)
{
	struct ceph_mds_client *mdsc = session->s_mdsc;
	struct inode *inode;
4027
	struct ceph_inode_info *ci;
Sage Weil's avatar
Sage Weil committed
4028 4029
	struct ceph_cap *cap;
	struct ceph_mds_caps *h;
4030
	struct ceph_mds_cap_peer *peer = NULL;
4031
	struct ceph_snap_realm *realm = NULL;
4032
	int op;
4033
	int msg_version = le16_to_cpu(msg->hdr.version);
4034
	u32 seq, mseq;
Sage Weil's avatar
Sage Weil committed
4035
	struct ceph_vino vino;
4036
	void *snaptrace;
4037
	size_t snaptrace_len;
4038
	void *p, *end;
4039
	struct cap_extra_info extra_info = {};
4040
	bool queue_trunc;
Sage Weil's avatar
Sage Weil committed
4041

4042
	dout("handle_caps from mds%d\n", session->s_mds);
Sage Weil's avatar
Sage Weil committed
4043 4044

	/* decode */
4045
	end = msg->front.iov_base + msg->front.iov_len;
Sage Weil's avatar
Sage Weil committed
4046 4047 4048 4049 4050 4051 4052
	if (msg->front.iov_len < sizeof(*h))
		goto bad;
	h = msg->front.iov_base;
	op = le32_to_cpu(h->op);
	vino.ino = le64_to_cpu(h->ino);
	vino.snap = CEPH_NOSNAP;
	seq = le32_to_cpu(h->seq);
4053
	mseq = le32_to_cpu(h->migrate_seq);
Sage Weil's avatar
Sage Weil committed
4054

4055 4056
	snaptrace = h + 1;
	snaptrace_len = le32_to_cpu(h->snap_trace_len);
4057
	p = snaptrace + snaptrace_len;
4058

4059
	if (msg_version >= 2) {
4060
		u32 flock_len;
4061
		ceph_decode_32_safe(&p, end, flock_len, bad);
4062 4063
		if (p + flock_len > end)
			goto bad;
4064
		p += flock_len;
4065 4066
	}

4067
	if (msg_version >= 3) {
4068 4069 4070 4071
		if (op == CEPH_CAP_OP_IMPORT) {
			if (p + sizeof(*peer) > end)
				goto bad;
			peer = p;
4072
			p += sizeof(*peer);
4073 4074 4075
		} else if (op == CEPH_CAP_OP_EXPORT) {
			/* recorded in unused fields */
			peer = (void *)&h->size;
4076 4077 4078
		}
	}

4079
	if (msg_version >= 4) {
4080 4081 4082
		ceph_decode_64_safe(&p, end, extra_info.inline_version, bad);
		ceph_decode_32_safe(&p, end, extra_info.inline_len, bad);
		if (p + extra_info.inline_len > end)
4083
			goto bad;
4084 4085
		extra_info.inline_data = p;
		p += extra_info.inline_len;
4086 4087
	}

4088
	if (msg_version >= 5) {
4089 4090 4091 4092 4093 4094 4095
		struct ceph_osd_client	*osdc = &mdsc->fsc->client->osdc;
		u32			epoch_barrier;

		ceph_decode_32_safe(&p, end, epoch_barrier, bad);
		ceph_osdc_update_epoch_barrier(osdc, epoch_barrier);
	}

4096
	if (msg_version >= 8) {
4097
		u32 pool_ns_len;
4098

4099
		/* version >= 6 */
4100
		ceph_decode_skip_64(&p, end, bad);	// flush_tid
4101
		/* version >= 7 */
4102 4103
		ceph_decode_skip_32(&p, end, bad);	// caller_uid
		ceph_decode_skip_32(&p, end, bad);	// caller_gid
4104 4105
		/* version >= 8 */
		ceph_decode_32_safe(&p, end, pool_ns_len, bad);
4106 4107
		if (pool_ns_len > 0) {
			ceph_decode_need(&p, end, pool_ns_len, bad);
4108 4109
			extra_info.pool_ns =
				ceph_find_or_create_string(p, pool_ns_len);
4110 4111
			p += pool_ns_len;
		}
4112 4113
	}

4114
	if (msg_version >= 9) {
4115 4116 4117 4118 4119
		struct ceph_timespec *btime;

		if (p + sizeof(*btime) > end)
			goto bad;
		btime = p;
4120
		ceph_decode_timespec64(&extra_info.btime, btime);
4121
		p += sizeof(*btime);
4122
		ceph_decode_64_safe(&p, end, extra_info.change_attr, bad);
4123 4124 4125
	}

	if (msg_version >= 11) {
4126
		/* version >= 10 */
4127
		ceph_decode_skip_32(&p, end, bad); // flags
4128 4129 4130 4131 4132 4133
		/* version >= 11 */
		extra_info.dirstat_valid = true;
		ceph_decode_64_safe(&p, end, extra_info.nfiles, bad);
		ceph_decode_64_safe(&p, end, extra_info.nsubdirs, bad);
	}

4134
	/* lookup ino */
4135
	inode = ceph_find_inode(mdsc->fsc->sb, vino);
4136 4137 4138 4139
	ci = ceph_inode(inode);
	dout(" op %s ino %llx.%llx inode %p\n", ceph_cap_op_name(op), vino.ino,
	     vino.snap, inode);

Sage Weil's avatar
Sage Weil committed
4140
	mutex_lock(&session->s_mutex);
4141
	inc_session_sequence(session);
Sage Weil's avatar
Sage Weil committed
4142 4143 4144 4145 4146
	dout(" mds%d seq %lld cap seq %u\n", session->s_mds, session->s_seq,
	     (unsigned)seq);

	if (!inode) {
		dout(" i don't have ino %llx\n", vino.ino);
4147

4148
		if (op == CEPH_CAP_OP_IMPORT) {
4149 4150 4151
			cap = ceph_get_cap(mdsc, NULL);
			cap->cap_ino = vino.ino;
			cap->queue_release = 1;
4152
			cap->cap_id = le64_to_cpu(h->cap_id);
4153 4154
			cap->mseq = mseq;
			cap->seq = seq;
4155
			cap->issue_seq = seq;
4156
			spin_lock(&session->s_cap_lock);
4157
			__ceph_queue_cap_release(session, cap);
4158 4159
			spin_unlock(&session->s_cap_lock);
		}
4160
		goto flush_cap_releases;
Sage Weil's avatar
Sage Weil committed
4161 4162 4163 4164 4165
	}

	/* these will work even if we don't have a cap yet */
	switch (op) {
	case CEPH_CAP_OP_FLUSHSNAP_ACK:
4166 4167
		handle_cap_flushsnap_ack(inode, le64_to_cpu(msg->hdr.tid),
					 h, session);
Sage Weil's avatar
Sage Weil committed
4168 4169 4170
		goto done;

	case CEPH_CAP_OP_EXPORT:
4171 4172
		handle_cap_export(inode, h, peer, session);
		goto done_unlocked;
Sage Weil's avatar
Sage Weil committed
4173 4174

	case CEPH_CAP_OP_IMPORT:
4175 4176 4177 4178 4179 4180 4181 4182 4183 4184
		realm = NULL;
		if (snaptrace_len) {
			down_write(&mdsc->snap_rwsem);
			ceph_update_snap_trace(mdsc, snaptrace,
					       snaptrace + snaptrace_len,
					       false, &realm);
			downgrade_write(&mdsc->snap_rwsem);
		} else {
			down_read(&mdsc->snap_rwsem);
		}
4185
		spin_lock(&ci->i_ceph_lock);
4186
		handle_cap_import(mdsc, inode, h, peer, session,
4187 4188 4189
				  &cap, &extra_info.issued);
		handle_cap_grant(inode, session, cap,
				 h, msg->middle, &extra_info);
4190 4191
		if (realm)
			ceph_put_snap_realm(mdsc, realm);
4192
		goto done_unlocked;
Sage Weil's avatar
Sage Weil committed
4193 4194 4195
	}

	/* the rest require a cap */
4196
	spin_lock(&ci->i_ceph_lock);
4197
	cap = __get_cap_for_mds(ceph_inode(inode), session->s_mds);
Sage Weil's avatar
Sage Weil committed
4198
	if (!cap) {
4199
		dout(" no cap on %p ino %llx.%llx from mds%d\n",
4200 4201
		     inode, ceph_ino(inode), ceph_snap(inode),
		     session->s_mds);
4202
		spin_unlock(&ci->i_ceph_lock);
4203
		goto flush_cap_releases;
Sage Weil's avatar
Sage Weil committed
4204 4205
	}

4206
	/* note that each of these drops i_ceph_lock for us */
Sage Weil's avatar
Sage Weil committed
4207 4208 4209
	switch (op) {
	case CEPH_CAP_OP_REVOKE:
	case CEPH_CAP_OP_GRANT:
4210 4211 4212 4213
		__ceph_caps_issued(ci, &extra_info.issued);
		extra_info.issued |= __ceph_caps_dirty(ci);
		handle_cap_grant(inode, session, cap,
				 h, msg->middle, &extra_info);
4214
		goto done_unlocked;
Sage Weil's avatar
Sage Weil committed
4215 4216

	case CEPH_CAP_OP_FLUSH_ACK:
4217 4218
		handle_cap_flush_ack(inode, le64_to_cpu(msg->hdr.tid),
				     h, session, cap);
Sage Weil's avatar
Sage Weil committed
4219 4220 4221
		break;

	case CEPH_CAP_OP_TRUNC:
4222 4223 4224 4225
		queue_trunc = handle_cap_trunc(inode, h, session);
		spin_unlock(&ci->i_ceph_lock);
		if (queue_trunc)
			ceph_queue_vmtruncate(inode);
Sage Weil's avatar
Sage Weil committed
4226 4227 4228
		break;

	default:
4229
		spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
4230 4231 4232 4233
		pr_err("ceph_handle_caps: unknown cap op %d %s\n", op,
		       ceph_cap_op_name(op));
	}

4234 4235 4236
done:
	mutex_unlock(&session->s_mutex);
done_unlocked:
4237
	iput(inode);
4238 4239
out:
	ceph_put_string(extra_info.pool_ns);
4240
	return;
4241 4242 4243

flush_cap_releases:
	/*
4244
	 * send any cap release message to try to move things
4245 4246 4247
	 * along for the mds (who clearly thinks we still have this
	 * cap).
	 */
4248 4249
	ceph_flush_cap_releases(mdsc, session);
	goto done;
Sage Weil's avatar
Sage Weil committed
4250 4251 4252

bad:
	pr_err("ceph_handle_caps: corrupt message\n");
4253
	ceph_msg_dump(msg);
4254
	goto out;
Sage Weil's avatar
Sage Weil committed
4255 4256 4257 4258
}

/*
 * Delayed work handler to process end of delayed cap release LRU list.
4259 4260 4261 4262
 *
 * If new caps are added to the list while processing it, these won't get
 * processed in this run.  In this case, the ci->i_hold_caps_max will be
 * returned so that the work can be scheduled accordingly.
Sage Weil's avatar
Sage Weil committed
4263
 */
4264
unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
Sage Weil's avatar
Sage Weil committed
4265
{
4266
	struct inode *inode;
Sage Weil's avatar
Sage Weil committed
4267
	struct ceph_inode_info *ci;
4268 4269 4270 4271
	struct ceph_mount_options *opt = mdsc->fsc->mount_options;
	unsigned long delay_max = opt->caps_wanted_delay_max * HZ;
	unsigned long loop_start = jiffies;
	unsigned long delay = 0;
Sage Weil's avatar
Sage Weil committed
4272 4273

	dout("check_delayed_caps\n");
4274 4275
	spin_lock(&mdsc->cap_delay_lock);
	while (!list_empty(&mdsc->cap_delay_list)) {
Sage Weil's avatar
Sage Weil committed
4276 4277 4278
		ci = list_first_entry(&mdsc->cap_delay_list,
				      struct ceph_inode_info,
				      i_cap_delay_list);
4279 4280 4281 4282 4283
		if (time_before(loop_start, ci->i_hold_caps_max - delay_max)) {
			dout("%s caps added recently.  Exiting loop", __func__);
			delay = ci->i_hold_caps_max;
			break;
		}
Sage Weil's avatar
Sage Weil committed
4284 4285 4286 4287
		if ((ci->i_ceph_flags & CEPH_I_FLUSH) == 0 &&
		    time_before(jiffies, ci->i_hold_caps_max))
			break;
		list_del_init(&ci->i_cap_delay_list);
4288 4289 4290

		inode = igrab(&ci->vfs_inode);
		if (inode) {
4291
			spin_unlock(&mdsc->cap_delay_lock);
4292
			dout("check_delayed_caps on %p\n", inode);
4293
			ceph_check_caps(ci, 0, NULL);
4294
			iput(inode);
4295
			spin_lock(&mdsc->cap_delay_lock);
4296
		}
Sage Weil's avatar
Sage Weil committed
4297 4298
	}
	spin_unlock(&mdsc->cap_delay_lock);
4299 4300

	return delay;
Sage Weil's avatar
Sage Weil committed
4301 4302
}

4303 4304 4305
/*
 * Flush all dirty caps to the mds
 */
4306
static void flush_dirty_session_caps(struct ceph_mds_session *s)
4307
{
4308
	struct ceph_mds_client *mdsc = s->s_mdsc;
4309 4310
	struct ceph_inode_info *ci;
	struct inode *inode;
4311 4312 4313

	dout("flush_dirty_caps\n");
	spin_lock(&mdsc->cap_dirty_lock);
4314 4315
	while (!list_empty(&s->s_cap_dirty)) {
		ci = list_first_entry(&s->s_cap_dirty, struct ceph_inode_info,
4316
				      i_dirty_item);
4317 4318
		inode = &ci->vfs_inode;
		ihold(inode);
4319
		dout("flush_dirty_caps %llx.%llx\n", ceph_vinop(inode));
4320
		spin_unlock(&mdsc->cap_dirty_lock);
4321
		ceph_check_caps(ci, CHECK_CAPS_FLUSH, NULL);
4322
		iput(inode);
4323 4324 4325
		spin_lock(&mdsc->cap_dirty_lock);
	}
	spin_unlock(&mdsc->cap_dirty_lock);
4326
	dout("flush_dirty_caps done\n");
4327 4328
}

4329 4330
void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
{
4331
	ceph_mdsc_iterate_sessions(mdsc, flush_dirty_session_caps, true);
4332 4333
}

4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345
void __ceph_touch_fmode(struct ceph_inode_info *ci,
			struct ceph_mds_client *mdsc, int fmode)
{
	unsigned long now = jiffies;
	if (fmode & CEPH_FILE_MODE_RD)
		ci->i_last_rd = now;
	if (fmode & CEPH_FILE_MODE_WR)
		ci->i_last_wr = now;
	/* queue periodic check */
	if (fmode &&
	    __ceph_is_any_real_caps(ci) &&
	    list_empty(&ci->i_cap_delay_list))
4346
		__cap_delay_requeue(mdsc, ci);
4347 4348 4349 4350
}

void ceph_get_fmode(struct ceph_inode_info *ci, int fmode, int count)
{
4351
	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->vfs_inode.i_sb);
4352
	int bits = (fmode << 1) | 1;
4353 4354 4355 4356 4357 4358
	bool is_opened = false;
	int i;

	if (count == 1)
		atomic64_inc(&mdsc->metric.opened_files);

4359 4360 4361 4362
	spin_lock(&ci->i_ceph_lock);
	for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
		if (bits & (1 << i))
			ci->i_nr_by_mode[i] += count;
4363 4364 4365 4366 4367 4368 4369 4370

		/*
		 * If any of the mode ref is larger than 1,
		 * that means it has been already opened by
		 * others. Just skip checking the PIN ref.
		 */
		if (i && ci->i_nr_by_mode[i] > 1)
			is_opened = true;
4371
	}
4372 4373 4374

	if (!is_opened)
		percpu_counter_inc(&mdsc->metric.opened_inodes);
4375 4376 4377
	spin_unlock(&ci->i_ceph_lock);
}

Sage Weil's avatar
Sage Weil committed
4378 4379 4380 4381 4382
/*
 * Drop open file reference.  If we were the last open file,
 * we may need to release capabilities to the MDS (or schedule
 * their delayed release).
 */
4383
void ceph_put_fmode(struct ceph_inode_info *ci, int fmode, int count)
Sage Weil's avatar
Sage Weil committed
4384
{
4385
	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(ci->vfs_inode.i_sb);
4386
	int bits = (fmode << 1) | 1;
4387 4388 4389 4390 4391 4392
	bool is_closed = true;
	int i;

	if (count == 1)
		atomic64_dec(&mdsc->metric.opened_files);

4393
	spin_lock(&ci->i_ceph_lock);
4394 4395
	for (i = 0; i < CEPH_FILE_MODE_BITS; i++) {
		if (bits & (1 << i)) {
4396 4397
			BUG_ON(ci->i_nr_by_mode[i] < count);
			ci->i_nr_by_mode[i] -= count;
4398
		}
4399 4400 4401 4402 4403 4404 4405 4406

		/*
		 * If any of the mode ref is not 0 after
		 * decreased, that means it is still opened
		 * by others. Just skip checking the PIN ref.
		 */
		if (i && ci->i_nr_by_mode[i])
			is_closed = false;
4407
	}
4408 4409 4410

	if (is_closed)
		percpu_counter_dec(&mdsc->metric.opened_inodes);
4411
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
4412 4413
}

4414
/*
4415
 * For a soon-to-be unlinked file, drop the LINK caps. If it
4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438
 * looks like the link count will hit 0, drop any other caps (other
 * than PIN) we don't specifically want (due to the file still being
 * open).
 */
int ceph_drop_caps_for_unlink(struct inode *inode)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;

	spin_lock(&ci->i_ceph_lock);
	if (inode->i_nlink == 1) {
		drop |= ~(__ceph_caps_wanted(ci) | CEPH_CAP_PIN);

		if (__ceph_caps_dirty(ci)) {
			struct ceph_mds_client *mdsc =
				ceph_inode_to_client(inode)->mdsc;
			__cap_delay_requeue_front(mdsc, ci);
		}
	}
	spin_unlock(&ci->i_ceph_lock);
	return drop;
}

Sage Weil's avatar
Sage Weil committed
4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452
/*
 * Helpers for embedding cap and dentry lease releases into mds
 * requests.
 *
 * @force is used by dentry_release (below) to force inclusion of a
 * record for the directory inode, even when there aren't any caps to
 * drop.
 */
int ceph_encode_inode_release(void **p, struct inode *inode,
			      int mds, int drop, int unless, int force)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_cap *cap;
	struct ceph_mds_request_release *rel = *p;
4453
	int used, dirty;
Sage Weil's avatar
Sage Weil committed
4454 4455
	int ret = 0;

4456
	spin_lock(&ci->i_ceph_lock);
4457
	used = __ceph_caps_used(ci);
4458
	dirty = __ceph_caps_dirty(ci);
4459

4460 4461
	dout("encode_inode_release %p mds%d used|dirty %s drop %s unless %s\n",
	     inode, mds, ceph_cap_string(used|dirty), ceph_cap_string(drop),
4462 4463
	     ceph_cap_string(unless));

4464 4465
	/* only drop unused, clean caps */
	drop &= ~(used | dirty);
4466

Sage Weil's avatar
Sage Weil committed
4467 4468
	cap = __get_cap_for_mds(ci, mds);
	if (cap && __cap_is_valid(cap)) {
4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482
		unless &= cap->issued;
		if (unless) {
			if (unless & CEPH_CAP_AUTH_EXCL)
				drop &= ~CEPH_CAP_AUTH_SHARED;
			if (unless & CEPH_CAP_LINK_EXCL)
				drop &= ~CEPH_CAP_LINK_SHARED;
			if (unless & CEPH_CAP_XATTR_EXCL)
				drop &= ~CEPH_CAP_XATTR_SHARED;
			if (unless & CEPH_CAP_FILE_EXCL)
				drop &= ~CEPH_CAP_FILE_SHARED;
		}

		if (force || (cap->issued & drop)) {
			if (cap->issued & drop) {
Yan, Zheng's avatar
Yan, Zheng committed
4483 4484 4485
				int wanted = __ceph_caps_wanted(ci);
				dout("encode_inode_release %p cap %p "
				     "%s -> %s, wanted %s -> %s\n", inode, cap,
Sage Weil's avatar
Sage Weil committed
4486
				     ceph_cap_string(cap->issued),
Yan, Zheng's avatar
Yan, Zheng committed
4487 4488 4489 4490
				     ceph_cap_string(cap->issued & ~drop),
				     ceph_cap_string(cap->mds_wanted),
				     ceph_cap_string(wanted));

Sage Weil's avatar
Sage Weil committed
4491 4492
				cap->issued &= ~drop;
				cap->implemented &= ~drop;
Yan, Zheng's avatar
Yan, Zheng committed
4493
				cap->mds_wanted = wanted;
4494 4495 4496
				if (cap == ci->i_auth_cap &&
				    !(wanted & CEPH_CAP_ANY_FILE_WR))
					ci->i_requested_max_size = 0;
Sage Weil's avatar
Sage Weil committed
4497 4498 4499 4500 4501 4502 4503 4504 4505
			} else {
				dout("encode_inode_release %p cap %p %s"
				     " (force)\n", inode, cap,
				     ceph_cap_string(cap->issued));
			}

			rel->ino = cpu_to_le64(ceph_ino(inode));
			rel->cap_id = cpu_to_le64(cap->cap_id);
			rel->seq = cpu_to_le32(cap->seq);
4506
			rel->issue_seq = cpu_to_le32(cap->issue_seq);
Sage Weil's avatar
Sage Weil committed
4507
			rel->mseq = cpu_to_le32(cap->mseq);
4508
			rel->caps = cpu_to_le32(cap->implemented);
Sage Weil's avatar
Sage Weil committed
4509 4510 4511 4512 4513 4514
			rel->wanted = cpu_to_le32(cap->mds_wanted);
			rel->dname_len = 0;
			rel->dname_seq = 0;
			*p += sizeof(*rel);
			ret = 1;
		} else {
4515
			dout("encode_inode_release %p cap %p %s (noop)\n",
Sage Weil's avatar
Sage Weil committed
4516 4517 4518
			     inode, cap, ceph_cap_string(cap->issued));
		}
	}
4519
	spin_unlock(&ci->i_ceph_lock);
Sage Weil's avatar
Sage Weil committed
4520 4521 4522 4523
	return ret;
}

int ceph_encode_dentry_release(void **p, struct dentry *dentry,
4524
			       struct inode *dir,
Sage Weil's avatar
Sage Weil committed
4525 4526
			       int mds, int drop, int unless)
{
4527
	struct dentry *parent = NULL;
Sage Weil's avatar
Sage Weil committed
4528 4529 4530 4531 4532 4533 4534
	struct ceph_mds_request_release *rel = *p;
	struct ceph_dentry_info *di = ceph_dentry(dentry);
	int force = 0;
	int ret;

	/*
	 * force an record for the directory caps if we have a dentry lease.
4535
	 * this is racy (can't take i_ceph_lock and d_lock together), but it
Sage Weil's avatar
Sage Weil committed
4536 4537 4538 4539 4540 4541
	 * doesn't have to be perfect; the mds will revoke anything we don't
	 * release.
	 */
	spin_lock(&dentry->d_lock);
	if (di->lease_session && di->lease_session->s_mds == mds)
		force = 1;
4542 4543 4544 4545
	if (!dir) {
		parent = dget(dentry->d_parent);
		dir = d_inode(parent);
	}
Sage Weil's avatar
Sage Weil committed
4546 4547
	spin_unlock(&dentry->d_lock);

4548
	ret = ceph_encode_inode_release(p, dir, mds, drop, unless, force);
4549
	dput(parent);
Sage Weil's avatar
Sage Weil committed
4550 4551 4552 4553 4554 4555 4556 4557 4558

	spin_lock(&dentry->d_lock);
	if (ret && di->lease_session && di->lease_session->s_mds == mds) {
		dout("encode_dentry_release %p mds%d seq %d\n",
		     dentry, mds, (int)di->lease_seq);
		rel->dname_len = cpu_to_le32(dentry->d_name.len);
		memcpy(*p, dentry->d_name.name, dentry->d_name.len);
		*p += dentry->d_name.len;
		rel->dname_seq = cpu_to_le32(di->lease_seq);
Sage Weil's avatar
Sage Weil committed
4559
		__ceph_mdsc_drop_dentry_lease(dentry);
Sage Weil's avatar
Sage Weil committed
4560 4561 4562 4563
	}
	spin_unlock(&dentry->d_lock);
	return ret;
}
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679

static int remove_capsnaps(struct ceph_mds_client *mdsc, struct inode *inode)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_cap_snap *capsnap;
	int capsnap_release = 0;

	lockdep_assert_held(&ci->i_ceph_lock);

	dout("removing capsnaps, ci is %p, inode is %p\n", ci, inode);

	while (!list_empty(&ci->i_cap_snaps)) {
		capsnap = list_first_entry(&ci->i_cap_snaps,
					   struct ceph_cap_snap, ci_item);
		__ceph_remove_capsnap(inode, capsnap, NULL, NULL);
		ceph_put_snap_context(capsnap->context);
		ceph_put_cap_snap(capsnap);
		capsnap_release++;
	}
	wake_up_all(&ci->i_cap_wq);
	wake_up_all(&mdsc->cap_flushing_wq);
	return capsnap_release;
}

int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate)
{
	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
	struct ceph_mds_client *mdsc = fsc->mdsc;
	struct ceph_inode_info *ci = ceph_inode(inode);
	bool is_auth;
	bool dirty_dropped = false;
	int iputs = 0;

	lockdep_assert_held(&ci->i_ceph_lock);

	dout("removing cap %p, ci is %p, inode is %p\n",
	     cap, ci, &ci->vfs_inode);

	is_auth = (cap == ci->i_auth_cap);
	__ceph_remove_cap(cap, false);
	if (is_auth) {
		struct ceph_cap_flush *cf;

		if (READ_ONCE(fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN) {
			if (inode->i_data.nrpages > 0)
				*invalidate = true;
			if (ci->i_wrbuffer_ref > 0)
				mapping_set_error(&inode->i_data, -EIO);
		}

		spin_lock(&mdsc->cap_dirty_lock);

		/* trash all of the cap flushes for this inode */
		while (!list_empty(&ci->i_cap_flush_list)) {
			cf = list_first_entry(&ci->i_cap_flush_list,
					      struct ceph_cap_flush, i_list);
			list_del_init(&cf->g_list);
			list_del_init(&cf->i_list);
			if (!cf->is_capsnap)
				ceph_free_cap_flush(cf);
		}

		if (!list_empty(&ci->i_dirty_item)) {
			pr_warn_ratelimited(
				" dropping dirty %s state for %p %lld\n",
				ceph_cap_string(ci->i_dirty_caps),
				inode, ceph_ino(inode));
			ci->i_dirty_caps = 0;
			list_del_init(&ci->i_dirty_item);
			dirty_dropped = true;
		}
		if (!list_empty(&ci->i_flushing_item)) {
			pr_warn_ratelimited(
				" dropping dirty+flushing %s state for %p %lld\n",
				ceph_cap_string(ci->i_flushing_caps),
				inode, ceph_ino(inode));
			ci->i_flushing_caps = 0;
			list_del_init(&ci->i_flushing_item);
			mdsc->num_cap_flushing--;
			dirty_dropped = true;
		}
		spin_unlock(&mdsc->cap_dirty_lock);

		if (dirty_dropped) {
			mapping_set_error(inode->i_mapping, -EIO);

			if (ci->i_wrbuffer_ref_head == 0 &&
			    ci->i_wr_ref == 0 &&
			    ci->i_dirty_caps == 0 &&
			    ci->i_flushing_caps == 0) {
				ceph_put_snap_context(ci->i_head_snapc);
				ci->i_head_snapc = NULL;
			}
		}

		if (atomic_read(&ci->i_filelock_ref) > 0) {
			/* make further file lock syscall return -EIO */
			ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
			pr_warn_ratelimited(" dropping file locks for %p %lld\n",
					    inode, ceph_ino(inode));
		}

		if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
			cf = ci->i_prealloc_cap_flush;
			ci->i_prealloc_cap_flush = NULL;
			if (!cf->is_capsnap)
				ceph_free_cap_flush(cf);
		}

		if (!list_empty(&ci->i_cap_snaps))
			iputs = remove_capsnaps(mdsc, inode);
	}
	if (dirty_dropped)
		++iputs;
	return iputs;
}