projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge remote-tracking branch 'staging/staging-next'
[deliverable/linux.git] / drivers / staging / lustre / lustre / osc / osc_cache.c
CommitLineData
d7e09d03
PT		 1/*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
6a5b99a4 18 * http://www.gnu.org/licenses/gpl-2.0.html
d7e09d03 19 *
d7e09d03
PT		 20 * GPL HEADER END
21 */
22/*
23 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
25 *
1dc563a6 26 * Copyright (c) 2012, 2015, Intel Corporation.
d7e09d03
PT		 27 *
28 */
29/*
30 * This file is part of Lustre, http://www.lustre.org/
31 * Lustre is a trademark of Sun Microsystems, Inc.
32 *
33 * osc cache management.
34 *
35 * Author: Jinshan Xiong <jinshan.xiong@whamcloud.com>
36 */
37
38#define DEBUG_SUBSYSTEM S_OSC
39
40#include "osc_cl_internal.h"
41#include "osc_internal.h"
42
43static int extent_debug; /* set it to be true for more debug */
44
45static void osc_update_pending(struct osc_object *obj, int cmd, int delta);
46static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext,
47 int state);
48static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli,
49 struct osc_async_page *oap, int sent, int rc);
50static int osc_make_ready(const struct lu_env *env, struct osc_async_page *oap,
51 int cmd);
52static int osc_refresh_count(const struct lu_env *env,
53 struct osc_async_page *oap, int cmd);
54static int osc_io_unplug_async(const struct lu_env *env,
55 struct client_obd *cli, struct osc_object *osc);
56static void osc_free_grant(struct client_obd *cli, unsigned int nr_pages,
57 unsigned int lost_grant);
58
59static void osc_extent_tree_dump0(int level, struct osc_object *obj,
60 const char *func, int line);
61#define osc_extent_tree_dump(lvl, obj) \
62 osc_extent_tree_dump0(lvl, obj, __func__, __LINE__)
63
64/** \addtogroup osc
65 * @{
66 */
67
68/* ------------------ osc extent ------------------ */
69static inline char *ext_flags(struct osc_extent *ext, char *flags)
70{
71 char *buf = flags;
72 *buf++ = ext->oe_rw ? 'r' : 'w';
73 if (ext->oe_intree)
74 *buf++ = 'i';
06563b56
JX		 75 if (ext->oe_sync)
76 *buf++ = 'S';
d7e09d03
PT		 77 if (ext->oe_srvlock)
78 *buf++ = 's';
79 if (ext->oe_hp)
80 *buf++ = 'h';
81 if (ext->oe_urgent)
82 *buf++ = 'u';
83 if (ext->oe_memalloc)
84 *buf++ = 'm';
85 if (ext->oe_trunc_pending)
86 *buf++ = 't';
87 if (ext->oe_fsync_wait)
88 *buf++ = 'Y';
89 *buf = 0;
90 return flags;
91}
92
93static inline char list_empty_marker(struct list_head *list)
94{
95 return list_empty(list) ? '-' : '+';
96}
97
98#define EXTSTR "[%lu -> %lu/%lu]"
99#define EXTPARA(ext) (ext)->oe_start, (ext)->oe_end, (ext)->oe_max_end
cad6fafa
BJ		 100static const char *oes_strings[] = {
101 "inv", "active", "cache", "locking", "lockdone", "rpc", "trunc", NULL };
d7e09d03
PT		 102
103#define OSC_EXTENT_DUMP(lvl, extent, fmt, ...) do { \
104 struct osc_extent *__ext = (extent); \
d7e09d03
PT		 105 char __buf[16]; \
106 \
107 CDEBUG(lvl, \
108 "extent %p@{" EXTSTR ", " \
109 "[%d|%d|%c|%s|%s|%p], [%d|%d|%c|%c|%p|%u|%p]} " fmt, \
110 /* ----- extent part 0 ----- */ \
111 __ext, EXTPARA(__ext), \
112 /* ----- part 1 ----- */ \
29ac6840
CH		 113 atomic_read(&__ext->oe_refc), \
114 atomic_read(&__ext->oe_users), \
d7e09d03 115 list_empty_marker(&__ext->oe_link), \
cad6fafa 116 oes_strings[__ext->oe_state], ext_flags(__ext, __buf), \
d7e09d03
PT		 117 __ext->oe_obj, \
118 /* ----- part 2 ----- */ \
119 __ext->oe_grants, __ext->oe_nr_pages, \
120 list_empty_marker(&__ext->oe_pages), \
121 waitqueue_active(&__ext->oe_waitq) ? '+' : '-', \
06563b56 122 __ext->oe_dlmlock, __ext->oe_mppr, __ext->oe_owner, \
d7e09d03
PT		 123 /* ----- part 4 ----- */ \
124 ## __VA_ARGS__); \
06563b56 125 if (lvl == D_ERROR && __ext->oe_dlmlock) \
e93876dd 126 LDLM_ERROR(__ext->oe_dlmlock, "extent: %p", __ext); \
06563b56 127 else \
e93876dd 128 LDLM_DEBUG(__ext->oe_dlmlock, "extent: %p", __ext); \
d7e09d03
PT		 129} while (0)
130
131#undef EASSERTF
132#define EASSERTF(expr, ext, fmt, args...) do { \
133 if (!(expr)) { \
cad6fafa
BJ		 134 OSC_EXTENT_DUMP(D_ERROR, (ext), fmt, ##args); \
135 osc_extent_tree_dump(D_ERROR, (ext)->oe_obj); \
d7e09d03 136 LASSERT(expr); \
cad6fafa 137 } \
d7e09d03
PT		 138} while (0)
139
140#undef EASSERT
141#define EASSERT(expr, ext) EASSERTF(expr, ext, "\n")
142
143static inline struct osc_extent *rb_extent(struct rb_node *n)
144{
7f1ae4c0 145 if (!n)
d7e09d03
PT		 146 return NULL;
147
148 return container_of(n, struct osc_extent, oe_node);
149}
150
151static inline struct osc_extent *next_extent(struct osc_extent *ext)
152{
7f1ae4c0 153 if (!ext)
d7e09d03
PT		 154 return NULL;
155
156 LASSERT(ext->oe_intree);
157 return rb_extent(rb_next(&ext->oe_node));
158}
159
160static inline struct osc_extent *prev_extent(struct osc_extent *ext)
161{
7f1ae4c0 162 if (!ext)
d7e09d03
PT		 163 return NULL;
164
165 LASSERT(ext->oe_intree);
166 return rb_extent(rb_prev(&ext->oe_node));
167}
168
169static inline struct osc_extent *first_extent(struct osc_object *obj)
170{
171 return rb_extent(rb_first(&obj->oo_root));
172}
173
174/* object must be locked by caller. */
175static int osc_extent_sanity_check0(struct osc_extent *ext,
176 const char *func, const int line)
177{
178 struct osc_object *obj = ext->oe_obj;
179 struct osc_async_page *oap;
180 int page_count;
181 int rc = 0;
182
490e0e89
JL		 183 if (!osc_object_is_locked(obj)) {
184 rc = 9;
185 goto out;
186 }
d7e09d03 187
490e0e89
JL		 188 if (ext->oe_state >= OES_STATE_MAX) {
189 rc = 10;
190 goto out;
191 }
d7e09d03 192
490e0e89
JL		 193 if (atomic_read(&ext->oe_refc) <= 0) {
194 rc = 20;
195 goto out;
196 }
d7e09d03 197
490e0e89
JL		 198 if (atomic_read(&ext->oe_refc) < atomic_read(&ext->oe_users)) {
199 rc = 30;
200 goto out;
201 }
d7e09d03
PT		 202
203 switch (ext->oe_state) {
204 case OES_INV:
205 if (ext->oe_nr_pages > 0 || !list_empty(&ext->oe_pages))
490e0e89
JL		 206 rc = 35;
207 else
208 rc = 0;
209 goto out;
d7e09d03 210 case OES_ACTIVE:
490e0e89
JL		 211 if (atomic_read(&ext->oe_users) == 0) {
212 rc = 40;
213 goto out;
214 }
215 if (ext->oe_hp) {
216 rc = 50;
217 goto out;
218 }
219 if (ext->oe_fsync_wait && !ext->oe_urgent) {
220 rc = 55;
221 goto out;
222 }
d7e09d03
PT		 223 break;
224 case OES_CACHE:
490e0e89
JL		 225 if (ext->oe_grants == 0) {
226 rc = 60;
227 goto out;
228 }
229 if (ext->oe_fsync_wait && !ext->oe_urgent && !ext->oe_hp) {
230 rc = 65;
231 goto out;
232 }
d7e09d03 233 default:
490e0e89
JL		 234 if (atomic_read(&ext->oe_users) > 0) {
235 rc = 70;
236 goto out;
237 }
d7e09d03
PT		 238 }
239
490e0e89
JL		 240 if (ext->oe_max_end < ext->oe_end || ext->oe_end < ext->oe_start) {
241 rc = 80;
242 goto out;
243 }
d7e09d03 244
06563b56 245 if (ext->oe_sync && ext->oe_grants > 0) {
490e0e89
JL		 246 rc = 90;
247 goto out;
248 }
d7e09d03 249
06563b56
JX		 250 if (ext->oe_dlmlock) {
251 struct ldlm_extent *extent;
50ffcb7e 252
06563b56
JX		 253 extent = &ext->oe_dlmlock->l_policy_data.l_extent;
254 if (!(extent->start <= cl_offset(osc2cl(obj), ext->oe_start) &&
255 extent->end >= cl_offset(osc2cl(obj), ext->oe_max_end))) {
490e0e89
JL		 256 rc = 100;
257 goto out;
258 }
06563b56
JX		 259
260 if (!(ext->oe_dlmlock->l_granted_mode & (LCK_PW | LCK_GROUP))) {
261 rc = 102;
262 goto out;
263 }
d7e09d03
PT		 264 }
265
490e0e89
JL		 266 if (ext->oe_nr_pages > ext->oe_mppr) {
267 rc = 105;
268 goto out;
269 }
d7e09d03
PT		 270
271 /* Do not verify page list if extent is in RPC. This is because an
30aa9c52
OD		 272 * in-RPC extent is supposed to be exclusively accessible w/o lock.
273 */
490e0e89
JL		 274 if (ext->oe_state > OES_CACHE) {
275 rc = 0;
276 goto out;
277 }
d7e09d03 278
490e0e89
JL		 279 if (!extent_debug) {
280 rc = 0;
281 goto out;
282 }
d7e09d03
PT		 283
284 page_count = 0;
285 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
7addf402 286 pgoff_t index = osc_index(oap2osc(oap));
d7e09d03 287 ++page_count;
490e0e89
JL		 288 if (index > ext->oe_end || index < ext->oe_start) {
289 rc = 110;
290 goto out;
291 }
292 }
293 if (page_count != ext->oe_nr_pages) {
294 rc = 120;
295 goto out;
d7e09d03 296 }
d7e09d03
PT		 297
298out:
299 if (rc != 0)
300 OSC_EXTENT_DUMP(D_ERROR, ext,
301 "%s:%d sanity check %p failed with rc = %d\n",
302 func, line, ext, rc);
303 return rc;
304}
305
306#define sanity_check_nolock(ext) \
307 osc_extent_sanity_check0(ext, __func__, __LINE__)
308
29ac6840
CH		 309#define sanity_check(ext) ({ \
310 int __res; \
d7e09d03 311 osc_object_lock((ext)->oe_obj); \
29ac6840
CH		 312 __res = sanity_check_nolock(ext); \
313 osc_object_unlock((ext)->oe_obj); \
314 __res; \
d7e09d03
PT		 315})
316
d7e09d03
PT		 317/**
318 * sanity check - to make sure there is no overlapped extent in the tree.
319 */
320static int osc_extent_is_overlapped(struct osc_object *obj,
321 struct osc_extent *ext)
322{
323 struct osc_extent *tmp;
324
325 LASSERT(osc_object_is_locked(obj));
326
327 if (!extent_debug)
328 return 0;
329
7f1ae4c0 330 for (tmp = first_extent(obj); tmp; tmp = next_extent(tmp)) {
d7e09d03
PT		 331 if (tmp == ext)
332 continue;
333 if (tmp->oe_end >= ext->oe_start &&
334 tmp->oe_start <= ext->oe_end)
335 return 1;
336 }
337 return 0;
338}
339
340static void osc_extent_state_set(struct osc_extent *ext, int state)
341{
342 LASSERT(osc_object_is_locked(ext->oe_obj));
343 LASSERT(state >= OES_INV && state < OES_STATE_MAX);
344
345 /* Never try to sanity check a state changing extent :-) */
346 /* LASSERT(sanity_check_nolock(ext) == 0); */
347
348 /* TODO: validate the state machine */
349 ext->oe_state = state;
350 wake_up_all(&ext->oe_waitq);
351}
352
353static struct osc_extent *osc_extent_alloc(struct osc_object *obj)
354{
355 struct osc_extent *ext;
356
c4418dac 357 ext = kmem_cache_zalloc(osc_extent_kmem, GFP_NOFS);
7f1ae4c0 358 if (!ext)
d7e09d03
PT		 359 return NULL;
360
361 RB_CLEAR_NODE(&ext->oe_node);
362 ext->oe_obj = obj;
363 atomic_set(&ext->oe_refc, 1);
364 atomic_set(&ext->oe_users, 0);
365 INIT_LIST_HEAD(&ext->oe_link);
366 ext->oe_state = OES_INV;
367 INIT_LIST_HEAD(&ext->oe_pages);
368 init_waitqueue_head(&ext->oe_waitq);
06563b56 369 ext->oe_dlmlock = NULL;
d7e09d03
PT		 370
371 return ext;
372}
373
374static void osc_extent_free(struct osc_extent *ext)
375{
50d30362 376 kmem_cache_free(osc_extent_kmem, ext);
d7e09d03
PT		 377}
378
379static struct osc_extent *osc_extent_get(struct osc_extent *ext)
380{
381 LASSERT(atomic_read(&ext->oe_refc) >= 0);
382 atomic_inc(&ext->oe_refc);
383 return ext;
384}
385
386static void osc_extent_put(const struct lu_env *env, struct osc_extent *ext)
387{
388 LASSERT(atomic_read(&ext->oe_refc) > 0);
389 if (atomic_dec_and_test(&ext->oe_refc)) {
390 LASSERT(list_empty(&ext->oe_link));
391 LASSERT(atomic_read(&ext->oe_users) == 0);
392 LASSERT(ext->oe_state == OES_INV);
393 LASSERT(!ext->oe_intree);
394
06563b56
JX		 395 if (ext->oe_dlmlock) {
396 lu_ref_add(&ext->oe_dlmlock->l_reference,
397 "osc_extent", ext);
398 LDLM_LOCK_PUT(ext->oe_dlmlock);
399 ext->oe_dlmlock = NULL;
d7e09d03
PT		 400 }
401 osc_extent_free(ext);
402 }
403}
404
405/**
406 * osc_extent_put_trust() is a special version of osc_extent_put() when
407 * it's known that the caller is not the last user. This is to address the
408 * problem of lacking of lu_env ;-).
409 */
410static void osc_extent_put_trust(struct osc_extent *ext)
411{
412 LASSERT(atomic_read(&ext->oe_refc) > 1);
413 LASSERT(osc_object_is_locked(ext->oe_obj));
414 atomic_dec(&ext->oe_refc);
415}
416
417/**
418 * Return the extent which includes pgoff @index, or return the greatest
419 * previous extent in the tree.
420 */
421static struct osc_extent *osc_extent_search(struct osc_object *obj,
422 pgoff_t index)
423{
29ac6840 424 struct rb_node *n = obj->oo_root.rb_node;
d7e09d03
PT		 425 struct osc_extent *tmp, *p = NULL;
426
427 LASSERT(osc_object_is_locked(obj));
7f1ae4c0 428 while (n) {
d7e09d03
PT		 429 tmp = rb_extent(n);
430 if (index < tmp->oe_start) {
431 n = n->rb_left;
432 } else if (index > tmp->oe_end) {
433 p = rb_extent(n);
434 n = n->rb_right;
435 } else {
436 return tmp;
437 }
438 }
439 return p;
440}
441
442/*
443 * Return the extent covering @index, otherwise return NULL.
444 * caller must have held object lock.
445 */
446static struct osc_extent *osc_extent_lookup(struct osc_object *obj,
447 pgoff_t index)
448{
449 struct osc_extent *ext;
450
451 ext = osc_extent_search(obj, index);
7f1ae4c0 452 if (ext && ext->oe_start <= index && index <= ext->oe_end)
d7e09d03
PT		 453 return osc_extent_get(ext);
454 return NULL;
455}
456
457/* caller must have held object lock. */
458static void osc_extent_insert(struct osc_object *obj, struct osc_extent *ext)
459{
29ac6840
CH		 460 struct rb_node **n = &obj->oo_root.rb_node;
461 struct rb_node *parent = NULL;
d7e09d03
PT		 462 struct osc_extent *tmp;
463
464 LASSERT(ext->oe_intree == 0);
465 LASSERT(ext->oe_obj == obj);
466 LASSERT(osc_object_is_locked(obj));
7f1ae4c0 467 while (*n) {
d7e09d03
PT		 468 tmp = rb_extent(*n);
469 parent = *n;
470
471 if (ext->oe_end < tmp->oe_start)
472 n = &(*n)->rb_left;
473 else if (ext->oe_start > tmp->oe_end)
474 n = &(*n)->rb_right;
475 else
19b2056f 476 EASSERTF(0, tmp, EXTSTR"\n", EXTPARA(ext));
d7e09d03
PT		 477 }
478 rb_link_node(&ext->oe_node, parent, n);
479 rb_insert_color(&ext->oe_node, &obj->oo_root);
480 osc_extent_get(ext);
481 ext->oe_intree = 1;
482}
483
484/* caller must have held object lock. */
485static void osc_extent_erase(struct osc_extent *ext)
486{
487 struct osc_object *obj = ext->oe_obj;
50ffcb7e 488
d7e09d03
PT		 489 LASSERT(osc_object_is_locked(obj));
490 if (ext->oe_intree) {
491 rb_erase(&ext->oe_node, &obj->oo_root);
492 ext->oe_intree = 0;
493 /* rbtree held a refcount */
494 osc_extent_put_trust(ext);
495 }
496}
497
498static struct osc_extent *osc_extent_hold(struct osc_extent *ext)
499{
500 struct osc_object *obj = ext->oe_obj;
501
502 LASSERT(osc_object_is_locked(obj));
503 LASSERT(ext->oe_state == OES_ACTIVE || ext->oe_state == OES_CACHE);
504 if (ext->oe_state == OES_CACHE) {
505 osc_extent_state_set(ext, OES_ACTIVE);
506 osc_update_pending(obj, OBD_BRW_WRITE, -ext->oe_nr_pages);
507 }
508 atomic_inc(&ext->oe_users);
509 list_del_init(&ext->oe_link);
510 return osc_extent_get(ext);
511}
512
513static void __osc_extent_remove(struct osc_extent *ext)
514{
515 LASSERT(osc_object_is_locked(ext->oe_obj));
516 LASSERT(list_empty(&ext->oe_pages));
517 osc_extent_erase(ext);
518 list_del_init(&ext->oe_link);
519 osc_extent_state_set(ext, OES_INV);
520 OSC_EXTENT_DUMP(D_CACHE, ext, "destroyed.\n");
521}
522
523static void osc_extent_remove(struct osc_extent *ext)
524{
525 struct osc_object *obj = ext->oe_obj;
526
527 osc_object_lock(obj);
528 __osc_extent_remove(ext);
529 osc_object_unlock(obj);
530}
531
532/**
533 * This function is used to merge extents to get better performance. It checks
534 * if @cur and @victim are contiguous at chunk level.
535 */
536static int osc_extent_merge(const struct lu_env *env, struct osc_extent *cur,
537 struct osc_extent *victim)
538{
539 struct osc_object *obj = cur->oe_obj;
540 pgoff_t chunk_start;
541 pgoff_t chunk_end;
542 int ppc_bits;
543
544 LASSERT(cur->oe_state == OES_CACHE);
545 LASSERT(osc_object_is_locked(obj));
7f1ae4c0 546 if (!victim)
d7e09d03
PT		 547 return -EINVAL;
548
549 if (victim->oe_state != OES_CACHE || victim->oe_fsync_wait)
550 return -EBUSY;
551
552 if (cur->oe_max_end != victim->oe_max_end)
553 return -ERANGE;
554
06563b56 555 LASSERT(cur->oe_dlmlock == victim->oe_dlmlock);
09cbfeaf 556 ppc_bits = osc_cli(obj)->cl_chunkbits - PAGE_SHIFT;
d7e09d03 557 chunk_start = cur->oe_start >> ppc_bits;
29ac6840
CH		 558 chunk_end = cur->oe_end >> ppc_bits;
559 if (chunk_start != (victim->oe_end >> ppc_bits) + 1 &&
d7e09d03
PT		 560 chunk_end + 1 != victim->oe_start >> ppc_bits)
561 return -ERANGE;
562
563 OSC_EXTENT_DUMP(D_CACHE, victim, "will be merged by %p.\n", cur);
564
29ac6840
CH		 565 cur->oe_start = min(cur->oe_start, victim->oe_start);
566 cur->oe_end = max(cur->oe_end, victim->oe_end);
567 cur->oe_grants += victim->oe_grants;
d7e09d03
PT		 568 cur->oe_nr_pages += victim->oe_nr_pages;
569 /* only the following bits are needed to merge */
29ac6840 570 cur->oe_urgent |= victim->oe_urgent;
d7e09d03
PT		 571 cur->oe_memalloc |= victim->oe_memalloc;
572 list_splice_init(&victim->oe_pages, &cur->oe_pages);
573 list_del_init(&victim->oe_link);
574 victim->oe_nr_pages = 0;
575
576 osc_extent_get(victim);
577 __osc_extent_remove(victim);
578 osc_extent_put(env, victim);
579
580 OSC_EXTENT_DUMP(D_CACHE, cur, "after merging %p.\n", victim);
581 return 0;
582}
583
584/**
585 * Drop user count of osc_extent, and unplug IO asynchronously.
586 */
882e7e21 587void osc_extent_release(const struct lu_env *env, struct osc_extent *ext)
d7e09d03
PT		 588{
589 struct osc_object *obj = ext->oe_obj;
d7e09d03
PT		 590
591 LASSERT(atomic_read(&ext->oe_users) > 0);
592 LASSERT(sanity_check(ext) == 0);
593 LASSERT(ext->oe_grants > 0);
594
595 if (atomic_dec_and_lock(&ext->oe_users, &obj->oo_lock)) {
596 LASSERT(ext->oe_state == OES_ACTIVE);
597 if (ext->oe_trunc_pending) {
598 /* a truncate process is waiting for this extent.
599 * This may happen due to a race, check
30aa9c52
OD		 600 * osc_cache_truncate_start().
601 */
d7e09d03
PT		 602 osc_extent_state_set(ext, OES_TRUNC);
603 ext->oe_trunc_pending = 0;
604 } else {
605 osc_extent_state_set(ext, OES_CACHE);
606 osc_update_pending(obj, OBD_BRW_WRITE,
607 ext->oe_nr_pages);
608
609 /* try to merge the previous and next extent. */
610 osc_extent_merge(env, ext, prev_extent(ext));
611 osc_extent_merge(env, ext, next_extent(ext));
612
613 if (ext->oe_urgent)
614 list_move_tail(&ext->oe_link,
79910d7d 615 &obj->oo_urgent_exts);
d7e09d03
PT		 616 }
617 osc_object_unlock(obj);
618
619 osc_io_unplug_async(env, osc_cli(obj), obj);
620 }
621 osc_extent_put(env, ext);
d7e09d03
PT		 622}
623
624static inline int overlapped(struct osc_extent *ex1, struct osc_extent *ex2)
625{
626 return !(ex1->oe_end < ex2->oe_start || ex2->oe_end < ex1->oe_start);
627}
628
629/**
630 * Find or create an extent which includes @index, core function to manage
631 * extent tree.
632 */
6ef3f3c7
SB		 633static struct osc_extent *osc_extent_find(const struct lu_env *env,
634 struct osc_object *obj, pgoff_t index,
635 int *grants)
d7e09d03
PT		 636{
637 struct client_obd *cli = osc_cli(obj);
06563b56
JX		 638 struct osc_lock *olck;
639 struct cl_lock_descr *descr;
d7e09d03
PT		 640 struct osc_extent *cur;
641 struct osc_extent *ext;
642 struct osc_extent *conflict = NULL;
643 struct osc_extent *found = NULL;
29ac6840
CH		 644 pgoff_t chunk;
645 pgoff_t max_end;
646 int max_pages; /* max_pages_per_rpc */
647 int chunksize;
648 int ppc_bits; /* pages per chunk bits */
649 int chunk_mask;
650 int rc;
d7e09d03
PT		 651
652 cur = osc_extent_alloc(obj);
7f1ae4c0 653 if (!cur)
0a3bdb00 654 return ERR_PTR(-ENOMEM);
d7e09d03 655
06563b56
JX		 656 olck = osc_env_io(env)->oi_write_osclock;
657 LASSERTF(olck, "page %lu is not covered by lock\n", index);
658 LASSERT(olck->ols_state == OLS_GRANTED);
659
660 descr = &olck->ols_cl.cls_lock->cll_descr;
661 LASSERT(descr->cld_mode >= CLM_WRITE);
d7e09d03 662
09cbfeaf
KS		 663 LASSERT(cli->cl_chunkbits >= PAGE_SHIFT);
664 ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
d7e09d03 665 chunk_mask = ~((1 << ppc_bits) - 1);
29ac6840
CH		 666 chunksize = 1 << cli->cl_chunkbits;
667 chunk = index >> ppc_bits;
d7e09d03
PT		 668
669 /* align end to rpc edge, rpc size may not be a power 2 integer. */
670 max_pages = cli->cl_max_pages_per_rpc;
671 LASSERT((max_pages & ~chunk_mask) == 0);
672 max_end = index - (index % max_pages) + max_pages - 1;
06563b56 673 max_end = min_t(pgoff_t, max_end, descr->cld_end);
d7e09d03
PT		 674
675 /* initialize new extent by parameters so far */
676 cur->oe_max_end = max_end;
29ac6840
CH		 677 cur->oe_start = index & chunk_mask;
678 cur->oe_end = ((index + ~chunk_mask + 1) & chunk_mask) - 1;
06563b56
JX		 679 if (cur->oe_start < descr->cld_start)
680 cur->oe_start = descr->cld_start;
d7e09d03
PT		 681 if (cur->oe_end > max_end)
682 cur->oe_end = max_end;
29ac6840
CH		 683 cur->oe_grants = 0;
684 cur->oe_mppr = max_pages;
06563b56
JX		 685 if (olck->ols_dlmlock) {
686 LASSERT(olck->ols_hold);
687 cur->oe_dlmlock = LDLM_LOCK_GET(olck->ols_dlmlock);
688 lu_ref_add(&olck->ols_dlmlock->l_reference, "osc_extent", cur);
689 }
d7e09d03
PT		 690
691 /* grants has been allocated by caller */
692 LASSERTF(*grants >= chunksize + cli->cl_extent_tax,
693 "%u/%u/%u.\n", *grants, chunksize, cli->cl_extent_tax);
19b2056f
JN		 694 LASSERTF((max_end - cur->oe_start) < max_pages, EXTSTR"\n",
695 EXTPARA(cur));
d7e09d03
PT		 696
697restart:
698 osc_object_lock(obj);
699 ext = osc_extent_search(obj, cur->oe_start);
7f1ae4c0 700 if (!ext)
d7e09d03 701 ext = first_extent(obj);
7f1ae4c0 702 while (ext) {
d7e09d03 703 loff_t ext_chk_start = ext->oe_start >> ppc_bits;
29ac6840 704 loff_t ext_chk_end = ext->oe_end >> ppc_bits;
d7e09d03
PT		 705
706 LASSERT(sanity_check_nolock(ext) == 0);
707 if (chunk > ext_chk_end + 1)
708 break;
709
710 /* if covering by different locks, no chance to match */
06563b56 711 if (olck->ols_dlmlock != ext->oe_dlmlock) {
d7e09d03 712 EASSERTF(!overlapped(ext, cur), ext,
19b2056f 713 EXTSTR"\n", EXTPARA(cur));
d7e09d03
PT		 714
715 ext = next_extent(ext);
716 continue;
717 }
718
719 /* discontiguous chunks? */
720 if (chunk + 1 < ext_chk_start) {
721 ext = next_extent(ext);
722 continue;
723 }
724
725 /* ok, from now on, ext and cur have these attrs:
726 * 1. covered by the same lock
30aa9c52
OD		 727 * 2. contiguous at chunk level or overlapping.
728 */
d7e09d03
PT		 729
730 if (overlapped(ext, cur)) {
731 /* cur is the minimum unit, so overlapping means
30aa9c52
OD		 732 * full contain.
733 */
d7e09d03
PT		 734 EASSERTF((ext->oe_start <= cur->oe_start &&
735 ext->oe_end >= cur->oe_end),
19b2056f 736 ext, EXTSTR"\n", EXTPARA(cur));
d7e09d03
PT		 737
738 if (ext->oe_state > OES_CACHE || ext->oe_fsync_wait) {
739 /* for simplicity, we wait for this extent to
30aa9c52
OD		 740 * finish before going forward.
741 */
d7e09d03
PT		 742 conflict = osc_extent_get(ext);
743 break;
744 }
745
746 found = osc_extent_hold(ext);
747 break;
748 }
749
750 /* non-overlapped extent */
751 if (ext->oe_state != OES_CACHE || ext->oe_fsync_wait) {
752 /* we can't do anything for a non OES_CACHE extent, or
753 * if there is someone waiting for this extent to be
30aa9c52
OD		 754 * flushed, try next one.
755 */
d7e09d03
PT		 756 ext = next_extent(ext);
757 continue;
758 }
759
760 /* check if they belong to the same rpc slot before trying to
761 * merge. the extents are not overlapped and contiguous at
30aa9c52
OD		 762 * chunk level to get here.
763 */
d7e09d03
PT		 764 if (ext->oe_max_end != max_end) {
765 /* if they don't belong to the same RPC slot or
30aa9c52
OD		 766 * max_pages_per_rpc has ever changed, do not merge.
767 */
d7e09d03
PT		 768 ext = next_extent(ext);
769 continue;
770 }
771
772 /* it's required that an extent must be contiguous at chunk
773 * level so that we know the whole extent is covered by grant
774 * (the pages in the extent are NOT required to be contiguous).
775 * Otherwise, it will be too much difficult to know which
30aa9c52
OD		 776 * chunks have grants allocated.
777 */
d7e09d03
PT		 778
779 /* try to do front merge - extend ext's start */
780 if (chunk + 1 == ext_chk_start) {
781 /* ext must be chunk size aligned */
782 EASSERT((ext->oe_start & ~chunk_mask) == 0, ext);
783
784 /* pull ext's start back to cover cur */
29ac6840 785 ext->oe_start = cur->oe_start;
d7e09d03
PT		 786 ext->oe_grants += chunksize;
787 *grants -= chunksize;
788
789 found = osc_extent_hold(ext);
790 } else if (chunk == ext_chk_end + 1) {
791 /* rear merge */
29ac6840 792 ext->oe_end = cur->oe_end;
d7e09d03
PT		 793 ext->oe_grants += chunksize;
794 *grants -= chunksize;
795
796 /* try to merge with the next one because we just fill
30aa9c52
OD		 797 * in a gap
798 */
d7e09d03
PT		 799 if (osc_extent_merge(env, ext, next_extent(ext)) == 0)
800 /* we can save extent tax from next extent */
801 *grants += cli->cl_extent_tax;
802
803 found = osc_extent_hold(ext);
804 }
7f1ae4c0 805 if (found)
d7e09d03
PT		 806 break;
807
808 ext = next_extent(ext);
809 }
810
811 osc_extent_tree_dump(D_CACHE, obj);
7f1ae4c0
OD		 812 if (found) {
813 LASSERT(!conflict);
d7e09d03 814 if (!IS_ERR(found)) {
06563b56 815 LASSERT(found->oe_dlmlock == cur->oe_dlmlock);
d7e09d03
PT		 816 OSC_EXTENT_DUMP(D_CACHE, found,
817 "found caching ext for %lu.\n", index);
818 }
7f1ae4c0 819 } else if (!conflict) {
d7e09d03
PT		 820 /* create a new extent */
821 EASSERT(osc_extent_is_overlapped(obj, cur) == 0, cur);
822 cur->oe_grants = chunksize + cli->cl_extent_tax;
823 *grants -= cur->oe_grants;
824 LASSERT(*grants >= 0);
825
826 cur->oe_state = OES_CACHE;
827 found = osc_extent_hold(cur);
828 osc_extent_insert(obj, cur);
829 OSC_EXTENT_DUMP(D_CACHE, cur, "add into tree %lu/%lu.\n",
06563b56 830 index, descr->cld_end);
d7e09d03
PT		 831 }
832 osc_object_unlock(obj);
833
7f1ae4c0
OD		 834 if (conflict) {
835 LASSERT(!found);
d7e09d03
PT		 836
837 /* waiting for IO to finish. Please notice that it's impossible
30aa9c52
OD		 838 * to be an OES_TRUNC extent.
839 */
d7e09d03
PT		 840 rc = osc_extent_wait(env, conflict, OES_INV);
841 osc_extent_put(env, conflict);
842 conflict = NULL;
490e0e89
JL		 843 if (rc < 0) {
844 found = ERR_PTR(rc);
845 goto out;
846 }
d7e09d03
PT		 847
848 goto restart;
849 }
d7e09d03
PT		 850
851out:
852 osc_extent_put(env, cur);
853 LASSERT(*grants >= 0);
854 return found;
855}
856
857/**
858 * Called when IO is finished to an extent.
859 */
860int osc_extent_finish(const struct lu_env *env, struct osc_extent *ext,
861 int sent, int rc)
862{
863 struct client_obd *cli = osc_cli(ext->oe_obj);
864 struct osc_async_page *oap;
865 struct osc_async_page *tmp;
866 int nr_pages = ext->oe_nr_pages;
867 int lost_grant = 0;
868 int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
869 __u64 last_off = 0;
870 int last_count = -1;
d7e09d03
PT		 871
872 OSC_EXTENT_DUMP(D_CACHE, ext, "extent finished.\n");
873
874 ext->oe_rc = rc ?: ext->oe_nr_pages;
875 EASSERT(ergo(rc == 0, ext->oe_state == OES_RPC), ext);
5196e42c
JX		 876
877 osc_lru_add_batch(cli, &ext->oe_pages);
79910d7d 878 list_for_each_entry_safe(oap, tmp, &ext->oe_pages, oap_pending_item) {
d7e09d03
PT		 879 list_del_init(&oap->oap_rpc_item);
880 list_del_init(&oap->oap_pending_item);
881 if (last_off <= oap->oap_obj_off) {
882 last_off = oap->oap_obj_off;
883 last_count = oap->oap_count;
884 }
885
886 --ext->oe_nr_pages;
887 osc_ap_completion(env, cli, oap, sent, rc);
888 }
889 EASSERT(ext->oe_nr_pages == 0, ext);
890
891 if (!sent) {
892 lost_grant = ext->oe_grants;
09cbfeaf
KS		 893 } else if (blocksize < PAGE_SIZE &&
894 last_count != PAGE_SIZE) {
d7e09d03
PT		 895 /* For short writes we shouldn't count parts of pages that
896 * span a whole chunk on the OST side, or our accounting goes
30aa9c52
OD		 897 * wrong. Should match the code in filter_grant_check.
898 */
77605e41
JX		 899 int offset = last_off & ~PAGE_MASK;
900 int count = last_count + (offset & (blocksize - 1));
901 int end = (offset + last_count) & (blocksize - 1);
d7e09d03
PT		 902 if (end)
903 count += blocksize - end;
904
09cbfeaf 905 lost_grant = PAGE_SIZE - count;
d7e09d03
PT		 906 }
907 if (ext->oe_grants > 0)
908 osc_free_grant(cli, nr_pages, lost_grant);
909
910 osc_extent_remove(ext);
911 /* put the refcount for RPC */
912 osc_extent_put(env, ext);
0a3bdb00 913 return 0;
d7e09d03
PT		 914}
915
916static int extent_wait_cb(struct osc_extent *ext, int state)
917{
918 int ret;
919
920 osc_object_lock(ext->oe_obj);
921 ret = ext->oe_state == state;
922 osc_object_unlock(ext->oe_obj);
923
924 return ret;
925}
926
927/**
928 * Wait for the extent's state to become @state.
929 */
930static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext,
931 int state)
932{
933 struct osc_object *obj = ext->oe_obj;
934 struct l_wait_info lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(600), NULL,
935 LWI_ON_SIGNAL_NOOP, NULL);
936 int rc = 0;
d7e09d03
PT		 937
938 osc_object_lock(obj);
939 LASSERT(sanity_check_nolock(ext) == 0);
940 /* `Kick' this extent only if the caller is waiting for it to be
30aa9c52
OD		 941 * written out.
942 */
ce248d59
AS		 943 if (state == OES_INV && !ext->oe_urgent && !ext->oe_hp &&
944 !ext->oe_trunc_pending) {
d7e09d03
PT		 945 if (ext->oe_state == OES_ACTIVE) {
946 ext->oe_urgent = 1;
947 } else if (ext->oe_state == OES_CACHE) {
948 ext->oe_urgent = 1;
949 osc_extent_hold(ext);
950 rc = 1;
951 }
952 }
953 osc_object_unlock(obj);
954 if (rc == 1)
955 osc_extent_release(env, ext);
956
957 /* wait for the extent until its state becomes @state */
958 rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state), &lwi);
959 if (rc == -ETIMEDOUT) {
960 OSC_EXTENT_DUMP(D_ERROR, ext,
24c198e9
OD		 961 "%s: wait ext to %d timedout, recovery in progress?\n",
962 osc_export(obj)->exp_obd->obd_name, state);
d7e09d03 963
b59d812d 964 lwi = LWI_INTR(NULL, NULL);
d7e09d03
PT		 965 rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state),
966 &lwi);
967 }
968 if (rc == 0 && ext->oe_rc < 0)
969 rc = ext->oe_rc;
0a3bdb00 970 return rc;
d7e09d03
PT		 971}
972
973/**
974 * Discard pages with index greater than @size. If @ext is overlapped with
975 * @size, then partial truncate happens.
976 */
977static int osc_extent_truncate(struct osc_extent *ext, pgoff_t trunc_index,
29ac6840 978 bool partial)
d7e09d03 979{
29ac6840
CH		 980 struct cl_env_nest nest;
981 struct lu_env *env;
982 struct cl_io *io;
983 struct osc_object *obj = ext->oe_obj;
984 struct client_obd *cli = osc_cli(obj);
d7e09d03
PT		 985 struct osc_async_page *oap;
986 struct osc_async_page *tmp;
29ac6840 987 int pages_in_chunk = 0;
09cbfeaf 988 int ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
29ac6840
CH		 989 __u64 trunc_chunk = trunc_index >> ppc_bits;
990 int grants = 0;
991 int nr_pages = 0;
992 int rc = 0;
d7e09d03
PT		 993
994 LASSERT(sanity_check(ext) == 0);
ce248d59
AS		 995 EASSERT(ext->oe_state == OES_TRUNC, ext);
996 EASSERT(!ext->oe_urgent, ext);
d7e09d03
PT		 997
998 /* Request new lu_env.
999 * We can't use that env from osc_cache_truncate_start() because
30aa9c52
OD		 1000 * it's from lov_io_sub and not fully initialized.
1001 */
d7e09d03
PT		 1002 env = cl_env_nested_get(&nest);
1003 io = &osc_env_info(env)->oti_io;
1004 io->ci_obj = cl_object_top(osc2cl(obj));
1005 rc = cl_io_init(env, io, CIT_MISC, io->ci_obj);
1006 if (rc < 0)
490e0e89 1007 goto out;
d7e09d03
PT		 1008
1009 /* discard all pages with index greater then trunc_index */
79910d7d 1010 list_for_each_entry_safe(oap, tmp, &ext->oe_pages, oap_pending_item) {
7addf402
JX		 1011 pgoff_t index = osc_index(oap2osc(oap));
1012 struct cl_page *page = oap2cl_page(oap);
d7e09d03
PT		 1013
1014 LASSERT(list_empty(&oap->oap_rpc_item));
1015
1016 /* only discard the pages with their index greater than
30aa9c52
OD		 1017 * trunc_index, and ...
1018 */
7addf402
JX		 1019 if (index < trunc_index ||
1020 (index == trunc_index && partial)) {
d7e09d03
PT		 1021 /* accounting how many pages remaining in the chunk
1022 * so that we can calculate grants correctly. */
7addf402 1023 if (index >> ppc_bits == trunc_chunk)
d7e09d03
PT		 1024 ++pages_in_chunk;
1025 continue;
1026 }
1027
1028 list_del_init(&oap->oap_pending_item);
1029
1030 cl_page_get(page);
1031 lu_ref_add(&page->cp_reference, "truncate", current);
1032
1033 if (cl_page_own(env, io, page) == 0) {
d7e09d03
PT		 1034 cl_page_discard(env, io, page);
1035 cl_page_disown(env, io, page);
1036 } else {
1037 LASSERT(page->cp_state == CPS_FREEING);
1038 LASSERT(0);
1039 }
1040
1041 lu_ref_del(&page->cp_reference, "truncate", current);
1042 cl_page_put(env, page);
1043
1044 --ext->oe_nr_pages;
1045 ++nr_pages;
1046 }
1047 EASSERTF(ergo(ext->oe_start >= trunc_index + !!partial,
1048 ext->oe_nr_pages == 0),
1049 ext, "trunc_index %lu, partial %d\n", trunc_index, partial);
1050
1051 osc_object_lock(obj);
1052 if (ext->oe_nr_pages == 0) {
1053 LASSERT(pages_in_chunk == 0);
1054 grants = ext->oe_grants;
1055 ext->oe_grants = 0;
1056 } else { /* calculate how many grants we can free */
29ac6840 1057 int chunks = (ext->oe_end >> ppc_bits) - trunc_chunk;
d7e09d03
PT		 1058 pgoff_t last_index;
1059
d7e09d03 1060 /* if there is no pages in this chunk, we can also free grants
30aa9c52
OD		 1061 * for the last chunk
1062 */
d7e09d03
PT		 1063 if (pages_in_chunk == 0) {
1064 /* if this is the 1st chunk and no pages in this chunk,
1065 * ext->oe_nr_pages must be zero, so we should be in
30aa9c52
OD		 1066 * the other if-clause.
1067 */
d7e09d03
PT		 1068 LASSERT(trunc_chunk > 0);
1069 --trunc_chunk;
1070 ++chunks;
1071 }
1072
1073 /* this is what we can free from this extent */
29ac6840 1074 grants = chunks << cli->cl_chunkbits;
d7e09d03 1075 ext->oe_grants -= grants;
29ac6840
CH		 1076 last_index = ((trunc_chunk + 1) << ppc_bits) - 1;
1077 ext->oe_end = min(last_index, ext->oe_max_end);
d7e09d03
PT		 1078 LASSERT(ext->oe_end >= ext->oe_start);
1079 LASSERT(ext->oe_grants > 0);
1080 }
1081 osc_object_unlock(obj);
1082
1083 if (grants > 0 || nr_pages > 0)
1084 osc_free_grant(cli, nr_pages, grants);
1085
1086out:
1087 cl_io_fini(env, io);
1088 cl_env_nested_put(&nest, env);
0a3bdb00 1089 return rc;
d7e09d03
PT		 1090}
1091
1092/**
1093 * This function is used to make the extent prepared for transfer.
34ca8748 1094 * A race with flushing page - ll_writepage() has to be handled cautiously.
d7e09d03
PT		 1095 */
1096static int osc_extent_make_ready(const struct lu_env *env,
1097 struct osc_extent *ext)
1098{
1099 struct osc_async_page *oap;
1100 struct osc_async_page *last = NULL;
1101 struct osc_object *obj = ext->oe_obj;
1102 int page_count = 0;
1103 int rc;
d7e09d03
PT		 1104
1105 /* we're going to grab page lock, so object lock must not be taken. */
1106 LASSERT(sanity_check(ext) == 0);
1107 /* in locking state, any process should not touch this extent. */
1108 EASSERT(ext->oe_state == OES_LOCKING, ext);
7f1ae4c0 1109 EASSERT(ext->oe_owner, ext);
d7e09d03
PT		 1110
1111 OSC_EXTENT_DUMP(D_CACHE, ext, "make ready\n");
1112
1113 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
1114 ++page_count;
7f1ae4c0 1115 if (!last || last->oap_obj_off < oap->oap_obj_off)
d7e09d03
PT		 1116 last = oap;
1117
1118 /* checking ASYNC_READY is race safe */
1119 if ((oap->oap_async_flags & ASYNC_READY) != 0)
1120 continue;
1121
1122 rc = osc_make_ready(env, oap, OBD_BRW_WRITE);
1123 switch (rc) {
1124 case 0:
1125 spin_lock(&oap->oap_lock);
1126 oap->oap_async_flags |= ASYNC_READY;
1127 spin_unlock(&oap->oap_lock);
1128 break;
1129 case -EALREADY:
1130 LASSERT((oap->oap_async_flags & ASYNC_READY) != 0);
1131 break;
1132 default:
1133 LASSERTF(0, "unknown return code: %d\n", rc);
1134 }
1135 }
1136
1137 LASSERT(page_count == ext->oe_nr_pages);
7f1ae4c0 1138 LASSERT(last);
d7e09d03 1139 /* the last page is the only one we need to refresh its count by
30aa9c52
OD		 1140 * the size of file.
1141 */
d7e09d03
PT		 1142 if (!(last->oap_async_flags & ASYNC_COUNT_STABLE)) {
1143 last->oap_count = osc_refresh_count(env, last, OBD_BRW_WRITE);
1144 LASSERT(last->oap_count > 0);
09cbfeaf 1145 LASSERT(last->oap_page_off + last->oap_count <= PAGE_SIZE);
82281bc0 1146 spin_lock(&last->oap_lock);
d7e09d03 1147 last->oap_async_flags |= ASYNC_COUNT_STABLE;
82281bc0 1148 spin_unlock(&last->oap_lock);
d7e09d03
PT		 1149 }
1150
1151 /* for the rest of pages, we don't need to call osf_refresh_count()
30aa9c52
OD		 1152 * because it's known they are not the last page
1153 */
d7e09d03
PT		 1154 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
1155 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
09cbfeaf 1156 oap->oap_count = PAGE_SIZE - oap->oap_page_off;
82281bc0 1157 spin_lock(&last->oap_lock);
d7e09d03 1158 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
82281bc0 1159 spin_unlock(&last->oap_lock);
d7e09d03
PT		 1160 }
1161 }
1162
1163 osc_object_lock(obj);
1164 osc_extent_state_set(ext, OES_RPC);
1165 osc_object_unlock(obj);
1166 /* get a refcount for RPC. */
1167 osc_extent_get(ext);
1168
0a3bdb00 1169 return 0;
d7e09d03
PT		 1170}
1171
1172/**
1173 * Quick and simple version of osc_extent_find(). This function is frequently
1174 * called to expand the extent for the same IO. To expand the extent, the
1175 * page index must be in the same or next chunk of ext->oe_end.
1176 */
1177static int osc_extent_expand(struct osc_extent *ext, pgoff_t index, int *grants)
1178{
1179 struct osc_object *obj = ext->oe_obj;
1180 struct client_obd *cli = osc_cli(obj);
1181 struct osc_extent *next;
09cbfeaf 1182 int ppc_bits = cli->cl_chunkbits - PAGE_SHIFT;
d7e09d03
PT		 1183 pgoff_t chunk = index >> ppc_bits;
1184 pgoff_t end_chunk;
1185 pgoff_t end_index;
1186 int chunksize = 1 << cli->cl_chunkbits;
1187 int rc = 0;
d7e09d03
PT		 1188
1189 LASSERT(ext->oe_max_end >= index && ext->oe_start <= index);
1190 osc_object_lock(obj);
1191 LASSERT(sanity_check_nolock(ext) == 0);
1192 end_chunk = ext->oe_end >> ppc_bits;
490e0e89
JL		 1193 if (chunk > end_chunk + 1) {
1194 rc = -ERANGE;
1195 goto out;
1196 }
d7e09d03 1197
490e0e89
JL		 1198 if (end_chunk >= chunk) {
1199 rc = 0;
1200 goto out;
1201 }
d7e09d03
PT		 1202
1203 LASSERT(end_chunk + 1 == chunk);
1204 /* try to expand this extent to cover @index */
1205 end_index = min(ext->oe_max_end, ((chunk + 1) << ppc_bits) - 1);
1206
1207 next = next_extent(ext);
7f1ae4c0 1208 if (next && next->oe_start <= end_index) {
d7e09d03 1209 /* complex mode - overlapped with the next extent,
30aa9c52
OD		 1210 * this case will be handled by osc_extent_find()
1211 */
490e0e89
JL		 1212 rc = -EAGAIN;
1213 goto out;
1214 }
d7e09d03
PT		 1215
1216 ext->oe_end = end_index;
1217 ext->oe_grants += chunksize;
1218 *grants -= chunksize;
1219 LASSERT(*grants >= 0);
1220 EASSERTF(osc_extent_is_overlapped(obj, ext) == 0, ext,
1221 "overlapped after expanding for %lu.\n", index);
d7e09d03
PT		 1222
1223out:
1224 osc_object_unlock(obj);
0a3bdb00 1225 return rc;
d7e09d03
PT		 1226}
1227
1228static void osc_extent_tree_dump0(int level, struct osc_object *obj,
1229 const char *func, int line)
1230{
1231 struct osc_extent *ext;
1232 int cnt;
1233
1234 CDEBUG(level, "Dump object %p extents at %s:%d, mppr: %u.\n",
1235 obj, func, line, osc_cli(obj)->cl_max_pages_per_rpc);
1236
1237 /* osc_object_lock(obj); */
1238 cnt = 1;
7f1ae4c0 1239 for (ext = first_extent(obj); ext; ext = next_extent(ext))
d7e09d03
PT		 1240 OSC_EXTENT_DUMP(level, ext, "in tree %d.\n", cnt++);
1241
1242 cnt = 1;
1243 list_for_each_entry(ext, &obj->oo_hp_exts, oe_link)
1244 OSC_EXTENT_DUMP(level, ext, "hp %d.\n", cnt++);
1245
1246 cnt = 1;
1247 list_for_each_entry(ext, &obj->oo_urgent_exts, oe_link)
1248 OSC_EXTENT_DUMP(level, ext, "urgent %d.\n", cnt++);
1249
1250 cnt = 1;
1251 list_for_each_entry(ext, &obj->oo_reading_exts, oe_link)
1252 OSC_EXTENT_DUMP(level, ext, "reading %d.\n", cnt++);
1253 /* osc_object_unlock(obj); */
1254}
1255
1256/* ------------------ osc extent end ------------------ */
1257
1258static inline int osc_is_ready(struct osc_object *osc)
1259{
1260 return !list_empty(&osc->oo_ready_item) ||
1261 !list_empty(&osc->oo_hp_ready_item);
1262}
1263
1264#define OSC_IO_DEBUG(OSC, STR, args...) \
1265 CDEBUG(D_CACHE, "obj %p ready %d|%c|%c wr %d|%c|%c rd %d|%c " STR, \
1266 (OSC), osc_is_ready(OSC), \
1267 list_empty_marker(&(OSC)->oo_hp_ready_item), \
1268 list_empty_marker(&(OSC)->oo_ready_item), \
1269 atomic_read(&(OSC)->oo_nr_writes), \
1270 list_empty_marker(&(OSC)->oo_hp_exts), \
1271 list_empty_marker(&(OSC)->oo_urgent_exts), \
1272 atomic_read(&(OSC)->oo_nr_reads), \
1273 list_empty_marker(&(OSC)->oo_reading_exts), \
1274 ##args)
1275
1276static int osc_make_ready(const struct lu_env *env, struct osc_async_page *oap,
1277 int cmd)
1278{
29ac6840 1279 struct osc_page *opg = oap2osc_page(oap);
7addf402 1280 struct cl_page *page = oap2cl_page(oap);
d7e09d03
PT		 1281 int result;
1282
1283 LASSERT(cmd == OBD_BRW_WRITE); /* no cached reads */
1284
d7e09d03
PT		 1285 result = cl_page_make_ready(env, page, CRT_WRITE);
1286 if (result == 0)
1287 opg->ops_submit_time = cfs_time_current();
0a3bdb00 1288 return result;
d7e09d03
PT		 1289}
1290
1291static int osc_refresh_count(const struct lu_env *env,
1292 struct osc_async_page *oap, int cmd)
1293{
29ac6840 1294 struct osc_page *opg = oap2osc_page(oap);
7addf402 1295 pgoff_t index = osc_index(oap2osc(oap));
d7e09d03 1296 struct cl_object *obj;
29ac6840 1297 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
d7e09d03
PT		 1298
1299 int result;
1300 loff_t kms;
1301
1302 /* readpage queues with _COUNT_STABLE, shouldn't get here. */
1303 LASSERT(!(cmd & OBD_BRW_READ));
d7e09d03
PT		 1304 obj = opg->ops_cl.cpl_obj;
1305
1306 cl_object_attr_lock(obj);
1307 result = cl_object_attr_get(env, obj, attr);
1308 cl_object_attr_unlock(obj);
1309 if (result < 0)
1310 return result;
1311 kms = attr->cat_kms;
7addf402 1312 if (cl_offset(obj, index) >= kms)
d7e09d03
PT		 1313 /* catch race with truncate */
1314 return 0;
7addf402 1315 else if (cl_offset(obj, index + 1) > kms)
d7e09d03 1316 /* catch sub-page write at end of file */
09cbfeaf 1317 return kms % PAGE_SIZE;
d7e09d03 1318 else
09cbfeaf 1319 return PAGE_SIZE;
d7e09d03
PT		 1320}
1321
1322static int osc_completion(const struct lu_env *env, struct osc_async_page *oap,
1323 int cmd, int rc)
1324{
29ac6840 1325 struct osc_page *opg = oap2osc_page(oap);
7addf402 1326 struct cl_page *page = oap2cl_page(oap);
29ac6840
CH		 1327 struct osc_object *obj = cl2osc(opg->ops_cl.cpl_obj);
1328 enum cl_req_type crt;
d7e09d03
PT		 1329 int srvlock;
1330
d7e09d03 1331 cmd &= ~OBD_BRW_NOQUOTA;
72a50e4f
NY		 1332 LASSERTF(equi(page->cp_state == CPS_PAGEIN, cmd == OBD_BRW_READ),
1333 "cp_state:%u, cmd:%d\n", page->cp_state, cmd);
1334 LASSERTF(equi(page->cp_state == CPS_PAGEOUT, cmd == OBD_BRW_WRITE),
1335 "cp_state:%u, cmd:%d\n", page->cp_state, cmd);
d7e09d03
PT		 1336 LASSERT(opg->ops_transfer_pinned);
1337
1338 /*
1339 * page->cp_req can be NULL if io submission failed before
1340 * cl_req was allocated.
1341 */
7f1ae4c0 1342 if (page->cp_req)
d7e09d03 1343 cl_req_page_done(env, page);
7f1ae4c0 1344 LASSERT(!page->cp_req);
d7e09d03
PT		 1345
1346 crt = cmd == OBD_BRW_READ ? CRT_READ : CRT_WRITE;
1347 /* Clear opg->ops_transfer_pinned before VM lock is released. */
1348 opg->ops_transfer_pinned = 0;
1349
1350 spin_lock(&obj->oo_seatbelt);
7f1ae4c0 1351 LASSERT(opg->ops_submitter);
d7e09d03
PT		 1352 LASSERT(!list_empty(&opg->ops_inflight));
1353 list_del_init(&opg->ops_inflight);
1354 opg->ops_submitter = NULL;
1355 spin_unlock(&obj->oo_seatbelt);
1356
1357 opg->ops_submit_time = 0;
1358 srvlock = oap->oap_brw_flags & OBD_BRW_SRVLOCK;
1359
1360 /* statistic */
1361 if (rc == 0 && srvlock) {
29ac6840 1362 struct lu_device *ld = opg->ops_cl.cpl_obj->co_lu.lo_dev;
d7e09d03
PT		 1363 struct osc_stats *stats = &lu2osc_dev(ld)->od_stats;
1364 int bytes = oap->oap_count;
1365
1366 if (crt == CRT_READ)
1367 stats->os_lockless_reads += bytes;
1368 else
1369 stats->os_lockless_writes += bytes;
1370 }
1371
1372 /*
1373 * This has to be the last operation with the page, as locks are
1374 * released in cl_page_completion() and nothing except for the
1375 * reference counter protects page from concurrent reclaim.
1376 */
1377 lu_ref_del(&page->cp_reference, "transfer", page);
1378
1379 cl_page_completion(env, page, crt, rc);
1380
0a3bdb00 1381 return 0;
d7e09d03
PT		 1382}
1383
c3558553 1384#define OSC_DUMP_GRANT(lvl, cli, fmt, args...) do { \
d7e09d03 1385 struct client_obd *__tmp = (cli); \
c3558553 1386 CDEBUG(lvl, "%s: grant { dirty: %ld/%ld dirty_pages: %d/%d " \
d806f30e
JX		 1387 "dropped: %ld avail: %ld, reserved: %ld, flight: %d }" \
1388 "lru {in list: %d, left: %d, waiters: %d }" fmt, \
d7e09d03 1389 __tmp->cl_import->imp_obd->obd_name, \
3147b268 1390 __tmp->cl_dirty_pages, __tmp->cl_dirty_max_pages, \
d7e09d03 1391 atomic_read(&obd_dirty_pages), obd_max_dirty_pages, \
d7e09d03 1392 __tmp->cl_lost_grant, __tmp->cl_avail_grant, \
c3558553
JX		 1393 __tmp->cl_reserved_grant, __tmp->cl_w_in_flight, \
1394 atomic_read(&__tmp->cl_lru_in_list), \
1395 atomic_read(&__tmp->cl_lru_busy), \
1396 atomic_read(&__tmp->cl_lru_shrinkers), ##args); \
d7e09d03
PT		 1397} while (0)
1398
1399/* caller must hold loi_list_lock */
1400static void osc_consume_write_grant(struct client_obd *cli,
1401 struct brw_page *pga)
1402{
7d53d8f4 1403 assert_spin_locked(&cli->cl_loi_list_lock);
d7e09d03
PT		 1404 LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
1405 atomic_inc(&obd_dirty_pages);
3147b268 1406 cli->cl_dirty_pages++;
d7e09d03
PT		 1407 pga->flag |= OBD_BRW_FROM_GRANT;
1408 CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
09cbfeaf 1409 PAGE_SIZE, pga, pga->pg);
d7e09d03
PT		 1410 osc_update_next_shrink(cli);
1411}
1412
1413/* the companion to osc_consume_write_grant, called when a brw has completed.
30aa9c52
OD		 1414 * must be called with the loi lock held.
1415 */
d7e09d03
PT		 1416static void osc_release_write_grant(struct client_obd *cli,
1417 struct brw_page *pga)
1418{
7d53d8f4 1419 assert_spin_locked(&cli->cl_loi_list_lock);
d7e09d03 1420 if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
d7e09d03
PT		 1421 return;
1422 }
1423
1424 pga->flag &= ~OBD_BRW_FROM_GRANT;
1425 atomic_dec(&obd_dirty_pages);
3147b268 1426 cli->cl_dirty_pages--;
d7e09d03
PT		 1427 if (pga->flag & OBD_BRW_NOCACHE) {
1428 pga->flag &= ~OBD_BRW_NOCACHE;
1429 atomic_dec(&obd_dirty_transit_pages);
3147b268 1430 cli->cl_dirty_transit--;
d7e09d03 1431 }
d7e09d03
PT		 1432}
1433
1434/**
1435 * To avoid sleeping with object lock held, it's good for us allocate enough
1436 * grants before entering into critical section.
1437 *
7d53d8f4 1438 * spin_lock held by caller
d7e09d03
PT		 1439 */
1440static int osc_reserve_grant(struct client_obd *cli, unsigned int bytes)
1441{
1442 int rc = -EDQUOT;
1443
1444 if (cli->cl_avail_grant >= bytes) {
29ac6840 1445 cli->cl_avail_grant -= bytes;
d7e09d03
PT		 1446 cli->cl_reserved_grant += bytes;
1447 rc = 0;
1448 }
1449 return rc;
1450}
1451
1452static void __osc_unreserve_grant(struct client_obd *cli,
1453 unsigned int reserved, unsigned int unused)
1454{
1455 /* it's quite normal for us to get more grant than reserved.
1456 * Thinking about a case that two extents merged by adding a new
1457 * chunk, we can save one extent tax. If extent tax is greater than
30aa9c52
OD		 1458 * one chunk, we can save more grant by adding a new chunk
1459 */
d7e09d03
PT		 1460 cli->cl_reserved_grant -= reserved;
1461 if (unused > reserved) {
1462 cli->cl_avail_grant += reserved;
1463 cli->cl_lost_grant += unused - reserved;
1464 } else {
1465 cli->cl_avail_grant += unused;
1466 }
1467}
1468
6ef3f3c7
SB		 1469static void osc_unreserve_grant(struct client_obd *cli,
1470 unsigned int reserved, unsigned int unused)
d7e09d03 1471{
7d53d8f4 1472 spin_lock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1473 __osc_unreserve_grant(cli, reserved, unused);
1474 if (unused > 0)
1475 osc_wake_cache_waiters(cli);
7d53d8f4 1476 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1477}
1478
1479/**
1480 * Free grant after IO is finished or canceled.
1481 *
1482 * @lost_grant is used to remember how many grants we have allocated but not
1483 * used, we should return these grants to OST. There're two cases where grants
1484 * can be lost:
1485 * 1. truncate;
ea1754a0 1486 * 2. blocksize at OST is less than PAGE_SIZE and a partial page was
d7e09d03
PT		 1487 * written. In this case OST may use less chunks to serve this partial
1488 * write. OSTs don't actually know the page size on the client side. so
1489 * clients have to calculate lost grant by the blocksize on the OST.
1490 * See filter_grant_check() for details.
1491 */
1492static void osc_free_grant(struct client_obd *cli, unsigned int nr_pages,
1493 unsigned int lost_grant)
1494{
1495 int grant = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
1496
7d53d8f4 1497 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 1498 atomic_sub(nr_pages, &obd_dirty_pages);
3147b268 1499 cli->cl_dirty_pages -= nr_pages;
d7e09d03
PT		 1500 cli->cl_lost_grant += lost_grant;
1501 if (cli->cl_avail_grant < grant && cli->cl_lost_grant >= grant) {
1502 /* borrow some grant from truncate to avoid the case that
30aa9c52
OD		 1503 * truncate uses up all avail grant
1504 */
d7e09d03
PT		 1505 cli->cl_lost_grant -= grant;
1506 cli->cl_avail_grant += grant;
1507 }
1508 osc_wake_cache_waiters(cli);
7d53d8f4 1509 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1510 CDEBUG(D_CACHE, "lost %u grant: %lu avail: %lu dirty: %lu\n",
1511 lost_grant, cli->cl_lost_grant,
3147b268 1512 cli->cl_avail_grant, cli->cl_dirty_pages << PAGE_SHIFT);
d7e09d03
PT		 1513}
1514
1515/**
1516 * The companion to osc_enter_cache(), called when @oap is no longer part of
1517 * the dirty accounting due to error.
1518 */
1519static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap)
1520{
7d53d8f4 1521 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 1522 osc_release_write_grant(cli, &oap->oap_brw_page);
7d53d8f4 1523 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1524}
1525
1526/**
1527 * Non-blocking version of osc_enter_cache() that consumes grant only when it
1528 * is available.
1529 */
1530static int osc_enter_cache_try(struct client_obd *cli,
1531 struct osc_async_page *oap,
1532 int bytes, int transient)
1533{
1534 int rc;
1535
c3558553 1536 OSC_DUMP_GRANT(D_CACHE, cli, "need:%d.\n", bytes);
d7e09d03
PT		 1537
1538 rc = osc_reserve_grant(cli, bytes);
1539 if (rc < 0)
1540 return 0;
1541
3147b268 1542 if (cli->cl_dirty_pages <= cli->cl_dirty_max_pages &&
d806f30e 1543 atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages) {
d7e09d03
PT		 1544 osc_consume_write_grant(cli, &oap->oap_brw_page);
1545 if (transient) {
3147b268 1546 cli->cl_dirty_transit++;
d7e09d03
PT		 1547 atomic_inc(&obd_dirty_transit_pages);
1548 oap->oap_brw_flags |= OBD_BRW_NOCACHE;
1549 }
1550 rc = 1;
1551 } else {
1552 __osc_unreserve_grant(cli, bytes, bytes);
1553 rc = 0;
1554 }
1555 return rc;
1556}
1557
1558static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
1559{
1560 int rc;
50ffcb7e 1561
7d53d8f4 1562 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 1563 rc = list_empty(&ocw->ocw_entry);
7d53d8f4 1564 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1565 return rc;
1566}
1567
1568/**
1569 * The main entry to reserve dirty page accounting. Usually the grant reserved
1570 * in this function will be freed in bulk in osc_free_grant() unless it fails
1571 * to add osc cache, in that case, it will be freed in osc_exit_cache().
1572 *
1573 * The process will be put into sleep if it's already run out of grant.
1574 */
1575static int osc_enter_cache(const struct lu_env *env, struct client_obd *cli,
1576 struct osc_async_page *oap, int bytes)
1577{
1578 struct osc_object *osc = oap->oap_obj;
29ac6840 1579 struct lov_oinfo *loi = osc->oo_oinfo;
d7e09d03 1580 struct osc_cache_waiter ocw;
c3558553
JX		 1581 struct l_wait_info lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(600), NULL,
1582 LWI_ON_SIGNAL_NOOP, NULL);
d7e09d03 1583 int rc = -EDQUOT;
d7e09d03 1584
c3558553 1585 OSC_DUMP_GRANT(D_CACHE, cli, "need:%d.\n", bytes);
d7e09d03 1586
7d53d8f4 1587 spin_lock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1588
1589 /* force the caller to try sync io. this can jump the list
30aa9c52
OD		 1590 * of queued writes and create a discontiguous rpc stream
1591 */
d7e09d03 1592 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_NO_GRANT) ||
3147b268
HZ		 1593 !cli->cl_dirty_max_pages || cli->cl_ar.ar_force_sync ||
1594 loi->loi_ar.ar_force_sync) {
490e0e89
JL		 1595 rc = -EDQUOT;
1596 goto out;
1597 }
d7e09d03
PT		 1598
1599 /* Hopefully normal case - cache space and write credits available */
490e0e89
JL		 1600 if (osc_enter_cache_try(cli, oap, bytes, 0)) {
1601 rc = 0;
1602 goto out;
1603 }
d7e09d03
PT		 1604
1605 /* We can get here for two reasons: too many dirty pages in cache, or
1606 * run out of grants. In both cases we should write dirty pages out.
1607 * Adding a cache waiter will trigger urgent write-out no matter what
1608 * RPC size will be.
1609 * The exiting condition is no avail grants and no dirty pages caching,
30aa9c52
OD		 1610 * that really means there is no space on the OST.
1611 */
d7e09d03
PT		 1612 init_waitqueue_head(&ocw.ocw_waitq);
1613 ocw.ocw_oap = oap;
1614 ocw.ocw_grant = bytes;
3147b268 1615 while (cli->cl_dirty_pages > 0 || cli->cl_w_in_flight > 0) {
d7e09d03
PT		 1616 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
1617 ocw.ocw_rc = 0;
7d53d8f4 1618 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1619
1620 osc_io_unplug_async(env, cli, NULL);
1621
1622 CDEBUG(D_CACHE, "%s: sleeping for cache space @ %p for %p\n",
1623 cli->cl_import->imp_obd->obd_name, &ocw, oap);
1624
1625 rc = l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
1626
7d53d8f4 1627 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 1628
c3558553 1629 /* l_wait_event is interrupted by signal, or timed out */
d7e09d03 1630 if (rc < 0) {
c3558553
JX		 1631 if (rc == -ETIMEDOUT) {
1632 OSC_DUMP_GRANT(D_ERROR, cli,
1633 "try to reserve %d.\n", bytes);
1634 osc_extent_tree_dump(D_ERROR, osc);
1635 rc = -EDQUOT;
1636 }
1637
d7e09d03 1638 list_del_init(&ocw.ocw_entry);
490e0e89 1639 goto out;
d7e09d03
PT		 1640 }
1641
1642 LASSERT(list_empty(&ocw.ocw_entry));
1643 rc = ocw.ocw_rc;
1644
1645 if (rc != -EDQUOT)
490e0e89
JL		 1646 goto out;
1647 if (osc_enter_cache_try(cli, oap, bytes, 0)) {
1648 rc = 0;
1649 goto out;
1650 }
d7e09d03 1651 }
d7e09d03 1652out:
7d53d8f4 1653 spin_unlock(&cli->cl_loi_list_lock);
c3558553 1654 OSC_DUMP_GRANT(D_CACHE, cli, "returned %d.\n", rc);
0a3bdb00 1655 return rc;
d7e09d03
PT		 1656}
1657
1658/* caller must hold loi_list_lock */
1659void osc_wake_cache_waiters(struct client_obd *cli)
1660{
1661 struct list_head *l, *tmp;
1662 struct osc_cache_waiter *ocw;
1663
d7e09d03
PT		 1664 list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
1665 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
1666 list_del_init(&ocw->ocw_entry);
1667
1668 ocw->ocw_rc = -EDQUOT;
1669 /* we can't dirty more */
3147b268 1670 if ((cli->cl_dirty_pages > cli->cl_dirty_max_pages) ||
d806f30e 1671 (atomic_read(&obd_dirty_pages) + 1 > obd_max_dirty_pages)) {
2d00bd17 1672 CDEBUG(D_CACHE, "no dirty room: dirty: %ld osc max %ld, sys max %d\n",
3147b268
HZ		 1673 cli->cl_dirty_pages, cli->cl_dirty_max_pages,
1674 obd_max_dirty_pages);
d7e09d03
PT		 1675 goto wakeup;
1676 }
1677
1678 ocw->ocw_rc = 0;
1679 if (!osc_enter_cache_try(cli, ocw->ocw_oap, ocw->ocw_grant, 0))
1680 ocw->ocw_rc = -EDQUOT;
1681
1682wakeup:
1683 CDEBUG(D_CACHE, "wake up %p for oap %p, avail grant %ld, %d\n",
1684 ocw, ocw->ocw_oap, cli->cl_avail_grant, ocw->ocw_rc);
1685
1686 wake_up(&ocw->ocw_waitq);
1687 }
d7e09d03
PT		 1688}
1689
1690static int osc_max_rpc_in_flight(struct client_obd *cli, struct osc_object *osc)
1691{
1692 int hprpc = !!list_empty(&osc->oo_hp_exts);
50ffcb7e 1693
d7e09d03
PT		 1694 return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
1695}
1696
1697/* This maintains the lists of pending pages to read/write for a given object
1698 * (lop). This is used by osc_check_rpcs->osc_next_obj() and osc_list_maint()
30aa9c52
OD		 1699 * to quickly find objects that are ready to send an RPC.
1700 */
d7e09d03
PT		 1701static int osc_makes_rpc(struct client_obd *cli, struct osc_object *osc,
1702 int cmd)
1703{
1704 int invalid_import = 0;
d7e09d03
PT		 1705
1706 /* if we have an invalid import we want to drain the queued pages
1707 * by forcing them through rpcs that immediately fail and complete
1708 * the pages. recovery relies on this to empty the queued pages
30aa9c52
OD		 1709 * before canceling the locks and evicting down the llite pages
1710 */
7f1ae4c0 1711 if (!cli->cl_import || cli->cl_import->imp_invalid)
d7e09d03
PT		 1712 invalid_import = 1;
1713
1714 if (cmd & OBD_BRW_WRITE) {
1715 if (atomic_read(&osc->oo_nr_writes) == 0)
0a3bdb00 1716 return 0;
d7e09d03
PT		 1717 if (invalid_import) {
1718 CDEBUG(D_CACHE, "invalid import forcing RPC\n");
0a3bdb00 1719 return 1;
d7e09d03
PT		 1720 }
1721 if (!list_empty(&osc->oo_hp_exts)) {
1722 CDEBUG(D_CACHE, "high prio request forcing RPC\n");
0a3bdb00 1723 return 1;
d7e09d03
PT		 1724 }
1725 if (!list_empty(&osc->oo_urgent_exts)) {
1726 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
0a3bdb00 1727 return 1;
d7e09d03
PT		 1728 }
1729 /* trigger a write rpc stream as long as there are dirtiers
1730 * waiting for space. as they're waiting, they're not going to
30aa9c52
OD		 1731 * create more pages to coalesce with what's waiting..
1732 */
d7e09d03
PT		 1733 if (!list_empty(&cli->cl_cache_waiters)) {
1734 CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
0a3bdb00 1735 return 1;
d7e09d03
PT		 1736 }
1737 if (atomic_read(&osc->oo_nr_writes) >=
1738 cli->cl_max_pages_per_rpc)
0a3bdb00 1739 return 1;
d7e09d03
PT		 1740 } else {
1741 if (atomic_read(&osc->oo_nr_reads) == 0)
0a3bdb00 1742 return 0;
d7e09d03
PT		 1743 if (invalid_import) {
1744 CDEBUG(D_CACHE, "invalid import forcing RPC\n");
0a3bdb00 1745 return 1;
d7e09d03
PT		 1746 }
1747 /* all read are urgent. */
1748 if (!list_empty(&osc->oo_reading_exts))
0a3bdb00 1749 return 1;
d7e09d03
PT		 1750 }
1751
0a3bdb00 1752 return 0;
d7e09d03
PT		 1753}
1754
1755static void osc_update_pending(struct osc_object *obj, int cmd, int delta)
1756{
1757 struct client_obd *cli = osc_cli(obj);
50ffcb7e 1758
d7e09d03
PT		 1759 if (cmd & OBD_BRW_WRITE) {
1760 atomic_add(delta, &obj->oo_nr_writes);
1761 atomic_add(delta, &cli->cl_pending_w_pages);
1762 LASSERT(atomic_read(&obj->oo_nr_writes) >= 0);
1763 } else {
1764 atomic_add(delta, &obj->oo_nr_reads);
1765 atomic_add(delta, &cli->cl_pending_r_pages);
1766 LASSERT(atomic_read(&obj->oo_nr_reads) >= 0);
1767 }
1768 OSC_IO_DEBUG(obj, "update pending cmd %d delta %d.\n", cmd, delta);
1769}
1770
1771static int osc_makes_hprpc(struct osc_object *obj)
1772{
1773 return !list_empty(&obj->oo_hp_exts);
1774}
1775
1776static void on_list(struct list_head *item, struct list_head *list, int should_be_on)
1777{
1778 if (list_empty(item) && should_be_on)
1779 list_add_tail(item, list);
1780 else if (!list_empty(item) && !should_be_on)
1781 list_del_init(item);
1782}
1783
1784/* maintain the osc's cli list membership invariants so that osc_send_oap_rpc
30aa9c52
OD		 1785 * can find pages to build into rpcs quickly
1786 */
d7e09d03
PT		 1787static int __osc_list_maint(struct client_obd *cli, struct osc_object *osc)
1788{
1789 if (osc_makes_hprpc(osc)) {
1790 /* HP rpc */
1791 on_list(&osc->oo_ready_item, &cli->cl_loi_ready_list, 0);
1792 on_list(&osc->oo_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
1793 } else {
1794 on_list(&osc->oo_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
1795 on_list(&osc->oo_ready_item, &cli->cl_loi_ready_list,
1796 osc_makes_rpc(cli, osc, OBD_BRW_WRITE) ||
1797 osc_makes_rpc(cli, osc, OBD_BRW_READ));
1798 }
1799
1800 on_list(&osc->oo_write_item, &cli->cl_loi_write_list,
1801 atomic_read(&osc->oo_nr_writes) > 0);
1802
1803 on_list(&osc->oo_read_item, &cli->cl_loi_read_list,
1804 atomic_read(&osc->oo_nr_reads) > 0);
1805
1806 return osc_is_ready(osc);
1807}
1808
1809static int osc_list_maint(struct client_obd *cli, struct osc_object *osc)
1810{
1811 int is_ready;
1812
7d53d8f4 1813 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 1814 is_ready = __osc_list_maint(cli, osc);
7d53d8f4 1815 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1816
1817 return is_ready;
1818}
1819
11d66e89 1820/* this is trying to propagate async writeback errors back up to the
d7e09d03
PT		 1821 * application. As an async write fails we record the error code for later if
1822 * the app does an fsync. As long as errors persist we force future rpcs to be
1823 * sync so that the app can get a sync error and break the cycle of queueing
30aa9c52
OD		 1824 * pages for which writeback will fail.
1825 */
d7e09d03
PT		 1826static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
1827 int rc)
1828{
1829 if (rc) {
1830 if (!ar->ar_rc)
1831 ar->ar_rc = rc;
1832
1833 ar->ar_force_sync = 1;
1834 ar->ar_min_xid = ptlrpc_sample_next_xid();
1835 return;
d7e09d03
PT		 1836 }
1837
1838 if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
1839 ar->ar_force_sync = 0;
1840}
1841
d7e09d03 1842/* this must be called holding the loi list lock to give coverage to exit_cache,
30aa9c52
OD		 1843 * async_flag maintenance, and oap_request
1844 */
d7e09d03
PT		 1845static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli,
1846 struct osc_async_page *oap, int sent, int rc)
1847{
1848 struct osc_object *osc = oap->oap_obj;
29ac6840 1849 struct lov_oinfo *loi = osc->oo_oinfo;
d7e09d03
PT		 1850 __u64 xid = 0;
1851
7f1ae4c0 1852 if (oap->oap_request) {
d7e09d03
PT		 1853 xid = ptlrpc_req_xid(oap->oap_request);
1854 ptlrpc_req_finished(oap->oap_request);
1855 oap->oap_request = NULL;
1856 }
1857
1858 /* As the transfer for this page is being done, clear the flags */
1859 spin_lock(&oap->oap_lock);
1860 oap->oap_async_flags = 0;
1861 spin_unlock(&oap->oap_lock);
1862 oap->oap_interrupted = 0;
1863
1864 if (oap->oap_cmd & OBD_BRW_WRITE && xid > 0) {
7d53d8f4 1865 spin_lock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1866 osc_process_ar(&cli->cl_ar, xid, rc);
1867 osc_process_ar(&loi->loi_ar, xid, rc);
7d53d8f4 1868 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 1869 }
1870
1871 rc = osc_completion(env, oap, oap->oap_cmd, rc);
1872 if (rc)
1873 CERROR("completion on oap %p obj %p returns %d.\n",
1874 oap, osc, rc);
d7e09d03
PT		 1875}
1876
1877/**
1878 * Try to add extent to one RPC. We need to think about the following things:
1879 * - # of pages must not be over max_pages_per_rpc
1880 * - extent must be compatible with previous ones
1881 */
1882static int try_to_add_extent_for_io(struct client_obd *cli,
1883 struct osc_extent *ext, struct list_head *rpclist,
1884 int *pc, unsigned int *max_pages)
1885{
1886 struct osc_extent *tmp;
c00266e3
AB		 1887 struct osc_async_page *oap = list_first_entry(&ext->oe_pages,
1888 struct osc_async_page,
1889 oap_pending_item);
d7e09d03
PT		 1890
1891 EASSERT((ext->oe_state == OES_CACHE || ext->oe_state == OES_LOCK_DONE),
1892 ext);
1893
1894 *max_pages = max(ext->oe_mppr, *max_pages);
1895 if (*pc + ext->oe_nr_pages > *max_pages)
0a3bdb00 1896 return 0;
d7e09d03
PT		 1897
1898 list_for_each_entry(tmp, rpclist, oe_link) {
c00266e3
AB		 1899 struct osc_async_page *oap2;
1900
1901 oap2 = list_first_entry(&tmp->oe_pages, struct osc_async_page,
1902 oap_pending_item);
d7e09d03 1903 EASSERT(tmp->oe_owner == current, tmp);
c00266e3
AB		 1904 if (oap2cl_page(oap)->cp_type != oap2cl_page(oap2)->cp_type) {
1905 CDEBUG(D_CACHE, "Do not permit different type of IO"
1906 " for a same RPC\n");
1907 return 0;
1908 }
d7e09d03
PT		 1909
1910 if (tmp->oe_srvlock != ext->oe_srvlock ||
1911 !tmp->oe_grants != !ext->oe_grants)
0a3bdb00 1912 return 0;
d7e09d03
PT		 1913
1914 /* remove break for strict check */
1915 break;
1916 }
1917
1918 *pc += ext->oe_nr_pages;
1919 list_move_tail(&ext->oe_link, rpclist);
1920 ext->oe_owner = current;
0a3bdb00 1921 return 1;
d7e09d03
PT		 1922}
1923
1924/**
1925 * In order to prevent multiple ptlrpcd from breaking contiguous extents,
1926 * get_write_extent() takes all appropriate extents in atomic.
1927 *
1928 * The following policy is used to collect extents for IO:
1929 * 1. Add as many HP extents as possible;
1930 * 2. Add the first urgent extent in urgent extent list and take it out of
1931 * urgent list;
1932 * 3. Add subsequent extents of this urgent extent;
1933 * 4. If urgent list is not empty, goto 2;
1934 * 5. Traverse the extent tree from the 1st extent;
1935 * 6. Above steps exit if there is no space in this RPC.
1936 */
1937static int get_write_extents(struct osc_object *obj, struct list_head *rpclist)
1938{
1939 struct client_obd *cli = osc_cli(obj);
1940 struct osc_extent *ext;
4a81ce53 1941 struct osc_extent *temp;
d7e09d03
PT		 1942 int page_count = 0;
1943 unsigned int max_pages = cli->cl_max_pages_per_rpc;
1944
1945 LASSERT(osc_object_is_locked(obj));
4a81ce53 1946 list_for_each_entry_safe(ext, temp, &obj->oo_hp_exts, oe_link) {
d7e09d03
PT		 1947 LASSERT(ext->oe_state == OES_CACHE);
1948 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1949 &max_pages))
1950 return page_count;
1951 EASSERT(ext->oe_nr_pages <= max_pages, ext);
1952 }
1953 if (page_count == max_pages)
1954 return page_count;
1955
1956 while (!list_empty(&obj->oo_urgent_exts)) {
1957 ext = list_entry(obj->oo_urgent_exts.next,
79910d7d 1958 struct osc_extent, oe_link);
d7e09d03
PT		 1959 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1960 &max_pages))
1961 return page_count;
1962
1963 if (!ext->oe_intree)
1964 continue;
1965
1966 while ((ext = next_extent(ext)) != NULL) {
1967 if ((ext->oe_state != OES_CACHE) ||
1968 (!list_empty(&ext->oe_link) &&
7f1ae4c0 1969 ext->oe_owner))
d7e09d03
PT		 1970 continue;
1971
1972 if (!try_to_add_extent_for_io(cli, ext, rpclist,
1973 &page_count, &max_pages))
1974 return page_count;
1975 }
1976 }
1977 if (page_count == max_pages)
1978 return page_count;
1979
1980 ext = first_extent(obj);
7f1ae4c0 1981 while (ext) {
d7e09d03
PT		 1982 if ((ext->oe_state != OES_CACHE) ||
1983 /* this extent may be already in current rpclist */
7f1ae4c0 1984 (!list_empty(&ext->oe_link) && ext->oe_owner)) {
d7e09d03
PT		 1985 ext = next_extent(ext);
1986 continue;
1987 }
1988
1989 if (!try_to_add_extent_for_io(cli, ext, rpclist, &page_count,
1990 &max_pages))
1991 return page_count;
1992
1993 ext = next_extent(ext);
1994 }
1995 return page_count;
1996}
1997
1998static int
1999osc_send_write_rpc(const struct lu_env *env, struct client_obd *cli,
c5c4c6fa 2000 struct osc_object *osc)
a161de86 2001 __must_hold(osc)
d7e09d03
PT		 2002{
2003 LIST_HEAD(rpclist);
2004 struct osc_extent *ext;
2005 struct osc_extent *tmp;
2006 struct osc_extent *first = NULL;
21aef7d9 2007 u32 page_count = 0;
d7e09d03
PT		 2008 int srvlock = 0;
2009 int rc = 0;
d7e09d03
PT		 2010
2011 LASSERT(osc_object_is_locked(osc));
2012
2013 page_count = get_write_extents(osc, &rpclist);
2014 LASSERT(equi(page_count == 0, list_empty(&rpclist)));
2015
2016 if (list_empty(&rpclist))
0a3bdb00 2017 return 0;
d7e09d03
PT		 2018
2019 osc_update_pending(osc, OBD_BRW_WRITE, -page_count);
2020
2021 list_for_each_entry(ext, &rpclist, oe_link) {
2022 LASSERT(ext->oe_state == OES_CACHE ||
2023 ext->oe_state == OES_LOCK_DONE);
2024 if (ext->oe_state == OES_CACHE)
2025 osc_extent_state_set(ext, OES_LOCKING);
2026 else
2027 osc_extent_state_set(ext, OES_RPC);
2028 }
2029
2030 /* we're going to grab page lock, so release object lock because
30aa9c52
OD		 2031 * lock order is page lock -> object lock.
2032 */
d7e09d03
PT		 2033 osc_object_unlock(osc);
2034
2035 list_for_each_entry_safe(ext, tmp, &rpclist, oe_link) {
2036 if (ext->oe_state == OES_LOCKING) {
2037 rc = osc_extent_make_ready(env, ext);
2038 if (unlikely(rc < 0)) {
2039 list_del_init(&ext->oe_link);
2040 osc_extent_finish(env, ext, 0, rc);
2041 continue;
2042 }
2043 }
7f1ae4c0 2044 if (!first) {
d7e09d03
PT		 2045 first = ext;
2046 srvlock = ext->oe_srvlock;
2047 } else {
2048 LASSERT(srvlock == ext->oe_srvlock);
2049 }
2050 }
2051
2052 if (!list_empty(&rpclist)) {
2053 LASSERT(page_count > 0);
c5c4c6fa 2054 rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_WRITE);
d7e09d03
PT		 2055 LASSERT(list_empty(&rpclist));
2056 }
2057
2058 osc_object_lock(osc);
0a3bdb00 2059 return rc;
d7e09d03
PT		 2060}
2061
2062/**
2063 * prepare pages for ASYNC io and put pages in send queue.
2064 *
2065 * \param cmd OBD_BRW_* macroses
2066 * \param lop pending pages
2067 *
2068 * \return zero if no page added to send queue.
2069 * \return 1 if pages successfully added to send queue.
2070 * \return negative on errors.
2071 */
2072static int
2073osc_send_read_rpc(const struct lu_env *env, struct client_obd *cli,
c5c4c6fa 2074 struct osc_object *osc)
a161de86 2075 __must_hold(osc)
d7e09d03
PT		 2076{
2077 struct osc_extent *ext;
2078 struct osc_extent *next;
2079 LIST_HEAD(rpclist);
2080 int page_count = 0;
2081 unsigned int max_pages = cli->cl_max_pages_per_rpc;
2082 int rc = 0;
d7e09d03
PT		 2083
2084 LASSERT(osc_object_is_locked(osc));
79910d7d 2085 list_for_each_entry_safe(ext, next, &osc->oo_reading_exts, oe_link) {
d7e09d03
PT		 2086 EASSERT(ext->oe_state == OES_LOCK_DONE, ext);
2087 if (!try_to_add_extent_for_io(cli, ext, &rpclist, &page_count,
2088 &max_pages))
2089 break;
2090 osc_extent_state_set(ext, OES_RPC);
2091 EASSERT(ext->oe_nr_pages <= max_pages, ext);
2092 }
2093 LASSERT(page_count <= max_pages);
2094
2095 osc_update_pending(osc, OBD_BRW_READ, -page_count);
2096
2097 if (!list_empty(&rpclist)) {
2098 osc_object_unlock(osc);
2099
2100 LASSERT(page_count > 0);
c5c4c6fa 2101 rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_READ);
d7e09d03
PT		 2102 LASSERT(list_empty(&rpclist));
2103
2104 osc_object_lock(osc);
2105 }
0a3bdb00 2106 return rc;
d7e09d03
PT		 2107}
2108
2109#define list_to_obj(list, item) ({ \
2110 struct list_head *__tmp = (list)->next; \
29ac6840 2111 list_del_init(__tmp); \
d7e09d03
PT		 2112 list_entry(__tmp, struct osc_object, oo_##item); \
2113})
2114
2115/* This is called by osc_check_rpcs() to find which objects have pages that
30aa9c52
OD		 2116 * we could be sending. These lists are maintained by osc_makes_rpc().
2117 */
d7e09d03
PT		 2118static struct osc_object *osc_next_obj(struct client_obd *cli)
2119{
d7e09d03
PT		 2120 /* First return objects that have blocked locks so that they
2121 * will be flushed quickly and other clients can get the lock,
30aa9c52
OD		 2122 * then objects which have pages ready to be stuffed into RPCs
2123 */
d7e09d03 2124 if (!list_empty(&cli->cl_loi_hp_ready_list))
0a3bdb00 2125 return list_to_obj(&cli->cl_loi_hp_ready_list, hp_ready_item);
d7e09d03 2126 if (!list_empty(&cli->cl_loi_ready_list))
0a3bdb00 2127 return list_to_obj(&cli->cl_loi_ready_list, ready_item);
d7e09d03
PT		 2128
2129 /* then if we have cache waiters, return all objects with queued
2130 * writes. This is especially important when many small files
2131 * have filled up the cache and not been fired into rpcs because
30aa9c52
OD		 2132 * they don't pass the nr_pending/object threshold
2133 */
d7e09d03
PT		 2134 if (!list_empty(&cli->cl_cache_waiters) &&
2135 !list_empty(&cli->cl_loi_write_list))
0a3bdb00 2136 return list_to_obj(&cli->cl_loi_write_list, write_item);
d7e09d03
PT		 2137
2138 /* then return all queued objects when we have an invalid import
30aa9c52
OD		 2139 * so that they get flushed
2140 */
7f1ae4c0 2141 if (!cli->cl_import || cli->cl_import->imp_invalid) {
d7e09d03 2142 if (!list_empty(&cli->cl_loi_write_list))
0a3bdb00 2143 return list_to_obj(&cli->cl_loi_write_list, write_item);
d7e09d03 2144 if (!list_empty(&cli->cl_loi_read_list))
0a3bdb00 2145 return list_to_obj(&cli->cl_loi_read_list, read_item);
d7e09d03 2146 }
0a3bdb00 2147 return NULL;
d7e09d03
PT		 2148}
2149
2150/* called with the loi list lock held */
c5c4c6fa 2151static void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
a161de86 2152 __must_hold(&cli->cl_loi_list_lock)
d7e09d03
PT		 2153{
2154 struct osc_object *osc;
2155 int rc = 0;
d7e09d03
PT		 2156
2157 while ((osc = osc_next_obj(cli)) != NULL) {
2158 struct cl_object *obj = osc2cl(osc);
631abc6e 2159 struct lu_ref_link link;
d7e09d03
PT		 2160
2161 OSC_IO_DEBUG(osc, "%lu in flight\n", rpcs_in_flight(cli));
2162
2163 if (osc_max_rpc_in_flight(cli, osc)) {
2164 __osc_list_maint(cli, osc);
2165 break;
2166 }
2167
2168 cl_object_get(obj);
7d53d8f4 2169 spin_unlock(&cli->cl_loi_list_lock);
d9d47901 2170 lu_object_ref_add_at(&obj->co_lu, &link, "check", current);
d7e09d03
PT		 2171
2172 /* attempt some read/write balancing by alternating between
2173 * reads and writes in an object. The makes_rpc checks here
2174 * would be redundant if we were getting read/write work items
2175 * instead of objects. we don't want send_oap_rpc to drain a
2176 * partial read pending queue when we're given this object to
30aa9c52
OD		 2177 * do io on writes while there are cache waiters
2178 */
d7e09d03
PT		 2179 osc_object_lock(osc);
2180 if (osc_makes_rpc(cli, osc, OBD_BRW_WRITE)) {
c5c4c6fa 2181 rc = osc_send_write_rpc(env, cli, osc);
d7e09d03
PT		 2182 if (rc < 0) {
2183 CERROR("Write request failed with %d\n", rc);
2184
2185 /* osc_send_write_rpc failed, mostly because of
2186 * memory pressure.
2187 *
2188 * It can't break here, because if:
2189 * - a page was submitted by osc_io_submit, so
2190 * page locked;
2191 * - no request in flight
2192 * - no subsequent request
2193 * The system will be in live-lock state,
2194 * because there is no chance to call
2195 * osc_io_unplug() and osc_check_rpcs() any
2196 * more. pdflush can't help in this case,
2197 * because it might be blocked at grabbing
2198 * the page lock as we mentioned.
2199 *
30aa9c52
OD		 2200 * Anyway, continue to drain pages.
2201 */
d7e09d03
PT		 2202 /* break; */
2203 }
2204 }
2205 if (osc_makes_rpc(cli, osc, OBD_BRW_READ)) {
c5c4c6fa 2206 rc = osc_send_read_rpc(env, cli, osc);
d7e09d03
PT		 2207 if (rc < 0)
2208 CERROR("Read request failed with %d\n", rc);
2209 }
2210 osc_object_unlock(osc);
2211
2212 osc_list_maint(cli, osc);
d9d47901 2213 lu_object_ref_del_at(&obj->co_lu, &link, "check", current);
d7e09d03
PT		 2214 cl_object_put(env, obj);
2215
7d53d8f4 2216 spin_lock(&cli->cl_loi_list_lock);
d7e09d03
PT		 2217 }
2218}
2219
2220static int osc_io_unplug0(const struct lu_env *env, struct client_obd *cli,
c5c4c6fa 2221 struct osc_object *osc, int async)
d7e09d03 2222{
d7e09d03
PT		 2223 int rc = 0;
2224
7f1ae4c0 2225 if (osc && osc_list_maint(cli, osc) == 0)
cad6fafa
BJ		 2226 return 0;
2227
2228 if (!async) {
2229 /* disable osc_lru_shrink() temporarily to avoid
30aa9c52
OD		 2230 * potential stack overrun problem. LU-2859
2231 */
cad6fafa 2232 atomic_inc(&cli->cl_lru_shrinkers);
7d53d8f4 2233 spin_lock(&cli->cl_loi_list_lock);
c5c4c6fa 2234 osc_check_rpcs(env, cli);
7d53d8f4 2235 spin_unlock(&cli->cl_loi_list_lock);
cad6fafa
BJ		 2236 atomic_dec(&cli->cl_lru_shrinkers);
2237 } else {
2238 CDEBUG(D_CACHE, "Queue writeback work for client %p.\n", cli);
7f1ae4c0 2239 LASSERT(cli->cl_writeback_work);
cad6fafa 2240 rc = ptlrpcd_queue_work(cli->cl_writeback_work);
d7e09d03 2241 }
d7e09d03
PT		 2242 return rc;
2243}
2244
2245static int osc_io_unplug_async(const struct lu_env *env,
29ac6840 2246 struct client_obd *cli, struct osc_object *osc)
d7e09d03 2247{
c5c4c6fa 2248 return osc_io_unplug0(env, cli, osc, 1);
d7e09d03
PT		 2249}
2250
2251void osc_io_unplug(const struct lu_env *env, struct client_obd *cli,
c5c4c6fa 2252 struct osc_object *osc)
d7e09d03 2253{
c5c4c6fa 2254 (void)osc_io_unplug0(env, cli, osc, 0);
d7e09d03
PT		 2255}
2256
2257int osc_prep_async_page(struct osc_object *osc, struct osc_page *ops,
2258 struct page *page, loff_t offset)
2259{
29ac6840 2260 struct obd_export *exp = osc_export(osc);
d7e09d03 2261 struct osc_async_page *oap = &ops->ops_oap;
d7e09d03
PT		 2262
2263 if (!page)
2264 return cfs_size_round(sizeof(*oap));
2265
2266 oap->oap_magic = OAP_MAGIC;
2267 oap->oap_cli = &exp->exp_obd->u.cli;
2268 oap->oap_obj = osc;
2269
2270 oap->oap_page = page;
2271 oap->oap_obj_off = offset;
616387e8 2272 LASSERT(!(offset & ~PAGE_MASK));
d7e09d03 2273
341f1f0a 2274 if (capable(CFS_CAP_SYS_RESOURCE))
d7e09d03
PT		 2275 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2276
2277 INIT_LIST_HEAD(&oap->oap_pending_item);
2278 INIT_LIST_HEAD(&oap->oap_rpc_item);
2279
2280 spin_lock_init(&oap->oap_lock);
b0f5aad5 2281 CDEBUG(D_INFO, "oap %p page %p obj off %llu\n",
d7e09d03 2282 oap, page, oap->oap_obj_off);
0a3bdb00 2283 return 0;
d7e09d03
PT		 2284}
2285
2286int osc_queue_async_io(const struct lu_env *env, struct cl_io *io,
2287 struct osc_page *ops)
2288{
2289 struct osc_io *oio = osc_env_io(env);
29ac6840 2290 struct osc_extent *ext = NULL;
d7e09d03 2291 struct osc_async_page *oap = &ops->ops_oap;
29ac6840
CH		 2292 struct client_obd *cli = oap->oap_cli;
2293 struct osc_object *osc = oap->oap_obj;
d7e09d03 2294 pgoff_t index;
29ac6840
CH		 2295 int grants = 0;
2296 int brw_flags = OBD_BRW_ASYNC;
2297 int cmd = OBD_BRW_WRITE;
2298 int need_release = 0;
2299 int rc = 0;
d7e09d03
PT		 2300
2301 if (oap->oap_magic != OAP_MAGIC)
0a3bdb00 2302 return -EINVAL;
d7e09d03 2303
7f1ae4c0 2304 if (!cli->cl_import || cli->cl_import->imp_invalid)
0a3bdb00 2305 return -EIO;
d7e09d03
PT		 2306
2307 if (!list_empty(&oap->oap_pending_item) ||
2308 !list_empty(&oap->oap_rpc_item))
0a3bdb00 2309 return -EBUSY;
d7e09d03
PT		 2310
2311 /* Set the OBD_BRW_SRVLOCK before the page is queued. */
2312 brw_flags |= ops->ops_srvlock ? OBD_BRW_SRVLOCK : 0;
341f1f0a 2313 if (capable(CFS_CAP_SYS_RESOURCE)) {
d7e09d03
PT		 2314 brw_flags |= OBD_BRW_NOQUOTA;
2315 cmd |= OBD_BRW_NOQUOTA;
2316 }
2317
2318 /* check if the file's owner/group is over quota */
2319 if (!(cmd & OBD_BRW_NOQUOTA)) {
2320 struct cl_object *obj;
29ac6840 2321 struct cl_attr *attr;
d7e09d03
PT		 2322 unsigned int qid[MAXQUOTAS];
2323
2324 obj = cl_object_top(&osc->oo_cl);
2325 attr = &osc_env_info(env)->oti_attr;
2326
2327 cl_object_attr_lock(obj);
2328 rc = cl_object_attr_get(env, obj, attr);
2329 cl_object_attr_unlock(obj);
2330
2331 qid[USRQUOTA] = attr->cat_uid;
2332 qid[GRPQUOTA] = attr->cat_gid;
2333 if (rc == 0 && osc_quota_chkdq(cli, qid) == NO_QUOTA)
2334 rc = -EDQUOT;
2335 if (rc)
0a3bdb00 2336 return rc;
d7e09d03
PT		 2337 }
2338
2339 oap->oap_cmd = cmd;
2340 oap->oap_page_off = ops->ops_from;
2341 oap->oap_count = ops->ops_to - ops->ops_from;
82281bc0
SB		 2342 /*
2343 * No need to hold a lock here,
2344 * since this page is not in any list yet.
2345 */
d7e09d03
PT		 2346 oap->oap_async_flags = 0;
2347 oap->oap_brw_flags = brw_flags;
2348
2349 OSC_IO_DEBUG(osc, "oap %p page %p added for cmd %d\n",
2350 oap, oap->oap_page, oap->oap_cmd & OBD_BRW_RWMASK);
2351
7addf402 2352 index = osc_index(oap2osc(oap));
d7e09d03
PT		 2353
2354 /* Add this page into extent by the following steps:
2355 * 1. if there exists an active extent for this IO, mostly this page
2356 * can be added to the active extent and sometimes we need to
11d66e89 2357 * expand extent to accommodate this page;
30aa9c52
OD		 2358 * 2. otherwise, a new extent will be allocated.
2359 */
d7e09d03
PT		 2360
2361 ext = oio->oi_active;
7f1ae4c0 2362 if (ext && ext->oe_start <= index && ext->oe_max_end >= index) {
d7e09d03 2363 /* one chunk plus extent overhead must be enough to write this
30aa9c52
OD		 2364 * page
2365 */
d7e09d03
PT		 2366 grants = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
2367 if (ext->oe_end >= index)
2368 grants = 0;
2369
2370 /* it doesn't need any grant to dirty this page */
7d53d8f4 2371 spin_lock(&cli->cl_loi_list_lock);
d7e09d03 2372 rc = osc_enter_cache_try(cli, oap, grants, 0);
7d53d8f4 2373 spin_unlock(&cli->cl_loi_list_lock);
d7e09d03
PT		 2374 if (rc == 0) { /* try failed */
2375 grants = 0;
2376 need_release = 1;
2377 } else if (ext->oe_end < index) {
2378 int tmp = grants;
2379 /* try to expand this extent */
2380 rc = osc_extent_expand(ext, index, &tmp);
2381 if (rc < 0) {
2382 need_release = 1;
2383 /* don't free reserved grant */
2384 } else {
2385 OSC_EXTENT_DUMP(D_CACHE, ext,
2386 "expanded for %lu.\n", index);
2387 osc_unreserve_grant(cli, grants, tmp);
2388 grants = 0;
2389 }
2390 }
2391 rc = 0;
7f1ae4c0 2392 } else if (ext) {
d7e09d03
PT		 2393 /* index is located outside of active extent */
2394 need_release = 1;
2395 }
2396 if (need_release) {
2397 osc_extent_release(env, ext);
2398 oio->oi_active = NULL;
2399 ext = NULL;
2400 }
2401
7f1ae4c0 2402 if (!ext) {
d7e09d03
PT		 2403 int tmp = (1 << cli->cl_chunkbits) + cli->cl_extent_tax;
2404
2405 /* try to find new extent to cover this page */
7f1ae4c0 2406 LASSERT(!oio->oi_active);
d7e09d03 2407 /* we may have allocated grant for this page if we failed
30aa9c52
OD		 2408 * to expand the previous active extent.
2409 */
d7e09d03
PT		 2410 LASSERT(ergo(grants > 0, grants >= tmp));
2411
2412 rc = 0;
2413 if (grants == 0) {
2414 /* we haven't allocated grant for this page. */
2415 rc = osc_enter_cache(env, cli, oap, tmp);
2416 if (rc == 0)
2417 grants = tmp;
2418 }
2419
2420 tmp = grants;
2421 if (rc == 0) {
2422 ext = osc_extent_find(env, osc, index, &tmp);
2423 if (IS_ERR(ext)) {
2424 LASSERT(tmp == grants);
2425 osc_exit_cache(cli, oap);
2426 rc = PTR_ERR(ext);
2427 ext = NULL;
2428 } else {
2429 oio->oi_active = ext;
2430 }
2431 }
2432 if (grants > 0)
2433 osc_unreserve_grant(cli, grants, tmp);
2434 }
2435
7f1ae4c0
OD		 2436 LASSERT(ergo(rc == 0, ext));
2437 if (ext) {
d7e09d03
PT		 2438 EASSERTF(ext->oe_end >= index && ext->oe_start <= index,
2439 ext, "index = %lu.\n", index);
2440 LASSERT((oap->oap_brw_flags & OBD_BRW_FROM_GRANT) != 0);
2441
2442 osc_object_lock(osc);
2443 if (ext->oe_nr_pages == 0)
2444 ext->oe_srvlock = ops->ops_srvlock;
2445 else
2446 LASSERT(ext->oe_srvlock == ops->ops_srvlock);
2447 ++ext->oe_nr_pages;
2448 list_add_tail(&oap->oap_pending_item, &ext->oe_pages);
2449 osc_object_unlock(osc);
2450 }
0a3bdb00 2451 return rc;
d7e09d03
PT		 2452}
2453
2454int osc_teardown_async_page(const struct lu_env *env,
2455 struct osc_object *obj, struct osc_page *ops)
2456{
2457 struct osc_async_page *oap = &ops->ops_oap;
29ac6840 2458 struct osc_extent *ext = NULL;
d7e09d03 2459 int rc = 0;
d7e09d03
PT		 2460
2461 LASSERT(oap->oap_magic == OAP_MAGIC);
2462
2463 CDEBUG(D_INFO, "teardown oap %p page %p at index %lu.\n",
7addf402 2464 oap, ops, osc_index(oap2osc(oap)));
d7e09d03
PT		 2465
2466 osc_object_lock(obj);
2467 if (!list_empty(&oap->oap_rpc_item)) {
2468 CDEBUG(D_CACHE, "oap %p is not in cache.\n", oap);
2469 rc = -EBUSY;
2470 } else if (!list_empty(&oap->oap_pending_item)) {
7addf402 2471 ext = osc_extent_lookup(obj, osc_index(oap2osc(oap)));
d7e09d03
PT		 2472 /* only truncated pages are allowed to be taken out.
2473 * See osc_extent_truncate() and osc_cache_truncate_start()
30aa9c52
OD		 2474 * for details.
2475 */
7f1ae4c0 2476 if (ext && ext->oe_state != OES_TRUNC) {
d7e09d03 2477 OSC_EXTENT_DUMP(D_ERROR, ext, "trunc at %lu.\n",
7addf402 2478 osc_index(oap2osc(oap)));
d7e09d03
PT		 2479 rc = -EBUSY;
2480 }
2481 }
2482 osc_object_unlock(obj);
7f1ae4c0 2483 if (ext)
d7e09d03 2484 osc_extent_put(env, ext);
0a3bdb00 2485 return rc;
d7e09d03
PT		 2486}
2487
2488/**
2489 * This is called when a page is picked up by kernel to write out.
2490 *
2491 * We should find out the corresponding extent and add the whole extent
2492 * into urgent list. The extent may be being truncated or used, handle it
2493 * carefully.
2494 */
2495int osc_flush_async_page(const struct lu_env *env, struct cl_io *io,
2496 struct osc_page *ops)
2497{
29ac6840
CH		 2498 struct osc_extent *ext = NULL;
2499 struct osc_object *obj = cl2osc(ops->ops_cl.cpl_obj);
2500 struct cl_page *cp = ops->ops_cl.cpl_page;
7addf402 2501 pgoff_t index = osc_index(ops);
d7e09d03
PT		 2502 struct osc_async_page *oap = &ops->ops_oap;
2503 bool unplug = false;
2504 int rc = 0;
d7e09d03
PT		 2505
2506 osc_object_lock(obj);
2507 ext = osc_extent_lookup(obj, index);
7f1ae4c0 2508 if (!ext) {
d7e09d03
PT		 2509 osc_extent_tree_dump(D_ERROR, obj);
2510 LASSERTF(0, "page index %lu is NOT covered.\n", index);
2511 }
2512
2513 switch (ext->oe_state) {
2514 case OES_RPC:
2515 case OES_LOCK_DONE:
7addf402 2516 CL_PAGE_DEBUG(D_ERROR, env, cp, "flush an in-rpc page?\n");
d7e09d03
PT		 2517 LASSERT(0);
2518 break;
2519 case OES_LOCKING:
2520 /* If we know this extent is being written out, we should abort
2521 * so that the writer can make this page ready. Otherwise, there
2522 * exists a deadlock problem because other process can wait for
2523 * page writeback bit holding page lock; and meanwhile in
2524 * vvp_page_make_ready(), we need to grab page lock before
30aa9c52
OD		 2525 * really sending the RPC.
2526 */
d7e09d03
PT		 2527 case OES_TRUNC:
2528 /* race with truncate, page will be redirtied */
15f13cde
AK		 2529 case OES_ACTIVE:
2530 /* The extent is active so we need to abort and let the caller
2531 * re-dirty the page. If we continued on here, and we were the
2532 * one making the extent active, we could deadlock waiting for
2533 * the page writeback to clear but it won't because the extent
30aa9c52
OD		 2534 * is active and won't be written out.
2535 */
490e0e89
JL		 2536 rc = -EAGAIN;
2537 goto out;
d7e09d03
PT		 2538 default:
2539 break;
2540 }
2541
7addf402 2542 rc = cl_page_prep(env, io, cp, CRT_WRITE);
d7e09d03 2543 if (rc)
490e0e89 2544 goto out;
d7e09d03
PT		 2545
2546 spin_lock(&oap->oap_lock);
cd94f231 2547 oap->oap_async_flags |= ASYNC_READY | ASYNC_URGENT;
d7e09d03
PT		 2548 spin_unlock(&oap->oap_lock);
2549
2550 if (memory_pressure_get())
2551 ext->oe_memalloc = 1;
2552
2553 ext->oe_urgent = 1;
2554 if (ext->oe_state == OES_CACHE) {
2555 OSC_EXTENT_DUMP(D_CACHE, ext,
2556 "flush page %p make it urgent.\n", oap);
2557 if (list_empty(&ext->oe_link))
2558 list_add_tail(&ext->oe_link, &obj->oo_urgent_exts);
2559 unplug = true;
2560 }
2561 rc = 0;
d7e09d03
PT		 2562
2563out:
2564 osc_object_unlock(obj);
2565 osc_extent_put(env, ext);
2566 if (unplug)
2567 osc_io_unplug_async(env, osc_cli(obj), obj);
2568 return rc;
2569}
2570
2571/**
2572 * this is called when a sync waiter receives an interruption. Its job is to
2573 * get the caller woken as soon as possible. If its page hasn't been put in an
2574 * rpc yet it can dequeue immediately. Otherwise it has to mark the rpc as
2575 * desiring interruption which will forcefully complete the rpc once the rpc
2576 * has timed out.
2577 */
2578int osc_cancel_async_page(const struct lu_env *env, struct osc_page *ops)
2579{
2580 struct osc_async_page *oap = &ops->ops_oap;
29ac6840
CH		 2581 struct osc_object *obj = oap->oap_obj;
2582 struct client_obd *cli = osc_cli(obj);
2583 struct osc_extent *ext;
2584 struct osc_extent *found = NULL;
2585 struct list_head *plist;
7addf402 2586 pgoff_t index = osc_index(ops);
29ac6840
CH		 2587 int rc = -EBUSY;
2588 int cmd;
d7e09d03
PT		 2589
2590 LASSERT(!oap->oap_interrupted);
2591 oap->oap_interrupted = 1;
2592
2593 /* Find out the caching extent */
2594 osc_object_lock(obj);
2595 if (oap->oap_cmd & OBD_BRW_WRITE) {
2596 plist = &obj->oo_urgent_exts;
29ac6840 2597 cmd = OBD_BRW_WRITE;
d7e09d03
PT		 2598 } else {
2599 plist = &obj->oo_reading_exts;
29ac6840 2600 cmd = OBD_BRW_READ;
d7e09d03
PT		 2601 }
2602 list_for_each_entry(ext, plist, oe_link) {
2603 if (ext->oe_start <= index && ext->oe_end >= index) {
2604 LASSERT(ext->oe_state == OES_LOCK_DONE);
2605 /* For OES_LOCK_DONE state extent, it has already held
30aa9c52
OD		 2606 * a refcount for RPC.
2607 */
d7e09d03
PT		 2608 found = osc_extent_get(ext);
2609 break;
2610 }
2611 }
7f1ae4c0 2612 if (found) {
d7e09d03
PT		 2613 list_del_init(&found->oe_link);
2614 osc_update_pending(obj, cmd, -found->oe_nr_pages);
2615 osc_object_unlock(obj);
2616
2617 osc_extent_finish(env, found, 0, -EINTR);
2618 osc_extent_put(env, found);
2619 rc = 0;
2620 } else {
2621 osc_object_unlock(obj);
2622 /* ok, it's been put in an rpc. only one oap gets a request
30aa9c52
OD		 2623 * reference
2624 */
7f1ae4c0 2625 if (oap->oap_request) {
d7e09d03
PT		 2626 ptlrpc_mark_interrupted(oap->oap_request);
2627 ptlrpcd_wake(oap->oap_request);
2628 ptlrpc_req_finished(oap->oap_request);
2629 oap->oap_request = NULL;
2630 }
2631 }
2632
2633 osc_list_maint(cli, obj);
0a3bdb00 2634 return rc;
d7e09d03
PT		 2635}
2636
2637int osc_queue_sync_pages(const struct lu_env *env, struct osc_object *obj,
2638 struct list_head *list, int cmd, int brw_flags)
2639{
29ac6840
CH		 2640 struct client_obd *cli = osc_cli(obj);
2641 struct osc_extent *ext;
f13ab92e 2642 struct osc_async_page *oap, *tmp;
29ac6840
CH		 2643 int page_count = 0;
2644 int mppr = cli->cl_max_pages_per_rpc;
2645 pgoff_t start = CL_PAGE_EOF;
2646 pgoff_t end = 0;
d7e09d03
PT		 2647
2648 list_for_each_entry(oap, list, oap_pending_item) {
7addf402 2649 pgoff_t index = osc_index(oap2osc(oap));
50ffcb7e 2650
7addf402
JX		 2651 if (index > end)
2652 end = index;
2653 if (index < start)
2654 start = index;
d7e09d03
PT		 2655 ++page_count;
2656 mppr <<= (page_count > mppr);
2657 }
2658
2659 ext = osc_extent_alloc(obj);
7f1ae4c0 2660 if (!ext) {
f13ab92e 2661 list_for_each_entry_safe(oap, tmp, list, oap_pending_item) {
d7e09d03
PT		 2662 list_del_init(&oap->oap_pending_item);
2663 osc_ap_completion(env, cli, oap, 0, -ENOMEM);
2664 }
0a3bdb00 2665 return -ENOMEM;
d7e09d03
PT		 2666 }
2667
2668 ext->oe_rw = !!(cmd & OBD_BRW_READ);
06563b56 2669 ext->oe_sync = 1;
d7e09d03
PT		 2670 ext->oe_urgent = 1;
2671 ext->oe_start = start;
04a6284f
NC		 2672 ext->oe_end = end;
2673 ext->oe_max_end = end;
d7e09d03
PT		 2674 ext->oe_obj = obj;
2675 ext->oe_srvlock = !!(brw_flags & OBD_BRW_SRVLOCK);
2676 ext->oe_nr_pages = page_count;
2677 ext->oe_mppr = mppr;
2678 list_splice_init(list, &ext->oe_pages);
2679
2680 osc_object_lock(obj);
2681 /* Reuse the initial refcount for RPC, don't drop it */
2682 osc_extent_state_set(ext, OES_LOCK_DONE);
2683 if (cmd & OBD_BRW_WRITE) {
2684 list_add_tail(&ext->oe_link, &obj->oo_urgent_exts);
2685 osc_update_pending(obj, OBD_BRW_WRITE, page_count);
2686 } else {
2687 list_add_tail(&ext->oe_link, &obj->oo_reading_exts);
2688 osc_update_pending(obj, OBD_BRW_READ, page_count);
2689 }
2690 osc_object_unlock(obj);
2691
c61ac979 2692 osc_io_unplug_async(env, cli, obj);
0a3bdb00 2693 return 0;
d7e09d03
PT		 2694}
2695
2696/**
2697 * Called by osc_io_setattr_start() to freeze and destroy covering extents.
2698 */
2699int osc_cache_truncate_start(const struct lu_env *env, struct osc_io *oio,
2700 struct osc_object *obj, __u64 size)
2701{
2702 struct client_obd *cli = osc_cli(obj);
2703 struct osc_extent *ext;
4a81ce53 2704 struct osc_extent *temp;
d7e09d03
PT		 2705 struct osc_extent *waiting = NULL;
2706 pgoff_t index;
2707 LIST_HEAD(list);
2708 int result = 0;
2709 bool partial;
d7e09d03
PT		 2710
2711 /* pages with index greater or equal to index will be truncated. */
2712 index = cl_index(osc2cl(obj), size);
2713 partial = size > cl_offset(osc2cl(obj), index);
2714
2715again:
2716 osc_object_lock(obj);
2717 ext = osc_extent_search(obj, index);
7f1ae4c0 2718 if (!ext)
d7e09d03
PT		 2719 ext = first_extent(obj);
2720 else if (ext->oe_end < index)
2721 ext = next_extent(ext);
7f1ae4c0 2722 while (ext) {
d7e09d03
PT		 2723 EASSERT(ext->oe_state != OES_TRUNC, ext);
2724
2725 if (ext->oe_state > OES_CACHE || ext->oe_urgent) {
2726 /* if ext is in urgent state, it means there must exist
2727 * a page already having been flushed by write_page().
2728 * We have to wait for this extent because we can't
30aa9c52
OD		 2729 * truncate that page.
2730 */
d7e09d03
PT		 2731 LASSERT(!ext->oe_hp);
2732 OSC_EXTENT_DUMP(D_CACHE, ext,
2733 "waiting for busy extent\n");
2734 waiting = osc_extent_get(ext);
2735 break;
2736 }
2737
b0f5aad5 2738 OSC_EXTENT_DUMP(D_CACHE, ext, "try to trunc:%llu.\n", size);
d7e09d03
PT		 2739
2740 osc_extent_get(ext);
2741 if (ext->oe_state == OES_ACTIVE) {
2742 /* though we grab inode mutex for write path, but we
2743 * release it before releasing extent(in osc_io_end()),
2744 * so there is a race window that an extent is still
30aa9c52
OD		 2745 * in OES_ACTIVE when truncate starts.
2746 */
d7e09d03
PT		 2747 LASSERT(!ext->oe_trunc_pending);
2748 ext->oe_trunc_pending = 1;
2749 } else {
2750 EASSERT(ext->oe_state == OES_CACHE, ext);
2751 osc_extent_state_set(ext, OES_TRUNC);
2752 osc_update_pending(obj, OBD_BRW_WRITE,
2753 -ext->oe_nr_pages);
2754 }
2755 EASSERT(list_empty(&ext->oe_link), ext);
2756 list_add_tail(&ext->oe_link, &list);
2757
2758 ext = next_extent(ext);
2759 }
2760 osc_object_unlock(obj);
2761
2762 osc_list_maint(cli, obj);
2763
4a81ce53 2764 list_for_each_entry_safe(ext, temp, &list, oe_link) {
d7e09d03
PT		 2765 int rc;
2766
d7e09d03
PT		 2767 list_del_init(&ext->oe_link);
2768
2769 /* extent may be in OES_ACTIVE state because inode mutex
30aa9c52
OD		 2770 * is released before osc_io_end() in file write case
2771 */
d7e09d03
PT		 2772 if (ext->oe_state != OES_TRUNC)
2773 osc_extent_wait(env, ext, OES_TRUNC);
2774
2775 rc = osc_extent_truncate(ext, index, partial);
2776 if (rc < 0) {
2777 if (result == 0)
2778 result = rc;
2779
2780 OSC_EXTENT_DUMP(D_ERROR, ext,
2781 "truncate error %d\n", rc);
2782 } else if (ext->oe_nr_pages == 0) {
2783 osc_extent_remove(ext);
2784 } else {
2785 /* this must be an overlapped extent which means only
2786 * part of pages in this extent have been truncated.
2787 */
2788 EASSERTF(ext->oe_start <= index, ext,
2789 "trunc index = %lu/%d.\n", index, partial);
2790 /* fix index to skip this partially truncated extent */
2791 index = ext->oe_end + 1;
2792 partial = false;
2793
2794 /* we need to hold this extent in OES_TRUNC state so
2795 * that no writeback will happen. This is to avoid
30aa9c52
OD		 2796 * BUG 17397.
2797 */
7f1ae4c0 2798 LASSERT(!oio->oi_trunc);
d7e09d03
PT		 2799 oio->oi_trunc = osc_extent_get(ext);
2800 OSC_EXTENT_DUMP(D_CACHE, ext,
b0f5aad5 2801 "trunc at %llu\n", size);
d7e09d03
PT		 2802 }
2803 osc_extent_put(env, ext);
2804 }
7f1ae4c0 2805 if (waiting) {
d7e09d03
PT		 2806 int rc;
2807
2808 /* ignore the result of osc_extent_wait the write initiator
30aa9c52
OD		 2809 * should take care of it.
2810 */
d7e09d03
PT		 2811 rc = osc_extent_wait(env, waiting, OES_INV);
2812 if (rc < 0)
451721cc 2813 OSC_EXTENT_DUMP(D_CACHE, waiting, "error: %d.\n", rc);
d7e09d03
PT		 2814
2815 osc_extent_put(env, waiting);
2816 waiting = NULL;
2817 goto again;
2818 }
0a3bdb00 2819 return result;
d7e09d03
PT		 2820}
2821
2822/**
2823 * Called after osc_io_setattr_end to add oio->oi_trunc back to cache.
2824 */
2825void osc_cache_truncate_end(const struct lu_env *env, struct osc_io *oio,
2826 struct osc_object *obj)
2827{
2828 struct osc_extent *ext = oio->oi_trunc;
2829
2830 oio->oi_trunc = NULL;
7f1ae4c0 2831 if (ext) {
d7e09d03
PT		 2832 bool unplug = false;
2833
2834 EASSERT(ext->oe_nr_pages > 0, ext);
2835 EASSERT(ext->oe_state == OES_TRUNC, ext);
2836 EASSERT(!ext->oe_urgent, ext);
2837
2838 OSC_EXTENT_DUMP(D_CACHE, ext, "trunc -> cache.\n");
2839 osc_object_lock(obj);
2840 osc_extent_state_set(ext, OES_CACHE);
2841 if (ext->oe_fsync_wait && !ext->oe_urgent) {
2842 ext->oe_urgent = 1;
2843 list_move_tail(&ext->oe_link, &obj->oo_urgent_exts);
2844 unplug = true;
2845 }
2846 osc_update_pending(obj, OBD_BRW_WRITE, ext->oe_nr_pages);
2847 osc_object_unlock(obj);
2848 osc_extent_put(env, ext);
2849
2850 if (unplug)
2851 osc_io_unplug_async(env, osc_cli(obj), obj);
2852 }
2853}
2854
2855/**
2856 * Wait for extents in a specific range to be written out.
2857 * The caller must have called osc_cache_writeback_range() to issue IO
2858 * otherwise it will take a long time for this function to finish.
2859 *
2860 * Caller must hold inode_mutex , or cancel exclusive dlm lock so that
2861 * nobody else can dirty this range of file while we're waiting for
2862 * extents to be written.
2863 */
2864int osc_cache_wait_range(const struct lu_env *env, struct osc_object *obj,
2865 pgoff_t start, pgoff_t end)
2866{
2867 struct osc_extent *ext;
2868 pgoff_t index = start;
29ac6840 2869 int result = 0;
d7e09d03
PT		 2870
2871again:
2872 osc_object_lock(obj);
2873 ext = osc_extent_search(obj, index);
7f1ae4c0 2874 if (!ext)
d7e09d03
PT		 2875 ext = first_extent(obj);
2876 else if (ext->oe_end < index)
2877 ext = next_extent(ext);
7f1ae4c0 2878 while (ext) {
d7e09d03
PT		 2879 int rc;
2880
2881 if (ext->oe_start > end)
2882 break;
2883
2884 if (!ext->oe_fsync_wait) {
2885 ext = next_extent(ext);
2886 continue;
2887 }
2888
2889 EASSERT(ergo(ext->oe_state == OES_CACHE,
2890 ext->oe_hp || ext->oe_urgent), ext);
2891 EASSERT(ergo(ext->oe_state == OES_ACTIVE,
2892 !ext->oe_hp && ext->oe_urgent), ext);
2893
2894 index = ext->oe_end + 1;
2895 osc_extent_get(ext);
2896 osc_object_unlock(obj);
2897
2898 rc = osc_extent_wait(env, ext, OES_INV);
2899 if (result == 0)
2900 result = rc;
2901 osc_extent_put(env, ext);
2902 goto again;
2903 }
2904 osc_object_unlock(obj);
2905
2906 OSC_IO_DEBUG(obj, "sync file range.\n");
0a3bdb00 2907 return result;
d7e09d03
PT		 2908}
2909
2910/**
2911 * Called to write out a range of osc object.
2912 *
2913 * @hp : should be set this is caused by lock cancel;
2914 * @discard: is set if dirty pages should be dropped - file will be deleted or
2915 * truncated, this implies there is no partially discarding extents.
2916 *
2917 * Return how many pages will be issued, or error code if error occurred.
2918 */
2919int osc_cache_writeback_range(const struct lu_env *env, struct osc_object *obj,
2920 pgoff_t start, pgoff_t end, int hp, int discard)
2921{
2922 struct osc_extent *ext;
2923 LIST_HEAD(discard_list);
2924 bool unplug = false;
2925 int result = 0;
d7e09d03
PT		 2926
2927 osc_object_lock(obj);
2928 ext = osc_extent_search(obj, start);
7f1ae4c0 2929 if (!ext)
d7e09d03
PT		 2930 ext = first_extent(obj);
2931 else if (ext->oe_end < start)
2932 ext = next_extent(ext);
7f1ae4c0 2933 while (ext) {
d7e09d03
PT		 2934 if (ext->oe_start > end)
2935 break;
2936
2937 ext->oe_fsync_wait = 1;
2938 switch (ext->oe_state) {
2939 case OES_CACHE:
2940 result += ext->oe_nr_pages;
2941 if (!discard) {
2942 struct list_head *list = NULL;
50ffcb7e 2943
d7e09d03
PT		 2944 if (hp) {
2945 EASSERT(!ext->oe_hp, ext);
2946 ext->oe_hp = 1;
2947 list = &obj->oo_hp_exts;
2948 } else if (!ext->oe_urgent) {
2949 ext->oe_urgent = 1;
2950 list = &obj->oo_urgent_exts;
2951 }
7f1ae4c0 2952 if (list)
d7e09d03
PT		 2953 list_move_tail(&ext->oe_link, list);
2954 unplug = true;
2955 } else {
2956 /* the only discarder is lock cancelling, so
30aa9c52
OD		 2957 * [start, end] must contain this extent
2958 */
d7e09d03
PT		 2959 EASSERT(ext->oe_start >= start &&
2960 ext->oe_max_end <= end, ext);
2961 osc_extent_state_set(ext, OES_LOCKING);
2962 ext->oe_owner = current;
79910d7d 2963 list_move_tail(&ext->oe_link, &discard_list);
d7e09d03
PT		 2964 osc_update_pending(obj, OBD_BRW_WRITE,
2965 -ext->oe_nr_pages);
2966 }
2967 break;
2968 case OES_ACTIVE:
2969 /* It's pretty bad to wait for ACTIVE extents, because
2970 * we don't know how long we will wait for it to be
2971 * flushed since it may be blocked at awaiting more
30aa9c52
OD		 2972 * grants. We do this for the correctness of fsync.
2973 */
d7e09d03
PT		 2974 LASSERT(hp == 0 && discard == 0);
2975 ext->oe_urgent = 1;
2976 break;
2977 case OES_TRUNC:
2978 /* this extent is being truncated, can't do anything
2979 * for it now. it will be set to urgent after truncate
30aa9c52
OD		 2980 * is finished in osc_cache_truncate_end().
2981 */
d7e09d03
PT		 2982 default:
2983 break;
2984 }
2985 ext = next_extent(ext);
2986 }
2987 osc_object_unlock(obj);
2988
2989 LASSERT(ergo(!discard, list_empty(&discard_list)));
2990 if (!list_empty(&discard_list)) {
2991 struct osc_extent *tmp;
2992 int rc;
2993
2994 osc_list_maint(osc_cli(obj), obj);
2995 list_for_each_entry_safe(ext, tmp, &discard_list, oe_link) {
2996 list_del_init(&ext->oe_link);
2997 EASSERT(ext->oe_state == OES_LOCKING, ext);
2998
2999 /* Discard caching pages. We don't actually write this
30aa9c52
OD		 3000 * extent out but we complete it as if we did.
3001 */
d7e09d03
PT		 3002 rc = osc_extent_make_ready(env, ext);
3003 if (unlikely(rc < 0)) {
3004 OSC_EXTENT_DUMP(D_ERROR, ext,
3005 "make_ready returned %d\n", rc);
3006 if (result >= 0)
3007 result = rc;
3008 }
3009
3010 /* finish the extent as if the pages were sent */
3011 osc_extent_finish(env, ext, 0, 0);
3012 }
3013 }
3014
3015 if (unplug)
c5c4c6fa 3016 osc_io_unplug(env, osc_cli(obj), obj);
d7e09d03
PT		 3017
3018 if (hp || discard) {
3019 int rc;
50ffcb7e 3020
d7e09d03
PT		 3021 rc = osc_cache_wait_range(env, obj, start, end);
3022 if (result >= 0 && rc < 0)
3023 result = rc;
3024 }
3025
c3558553 3026 OSC_IO_DEBUG(obj, "pageout [%lu, %lu], %d.\n", start, end, result);
0a3bdb00 3027 return result;
d7e09d03
PT		 3028}
3029
d9d47901
JX		 3030/**
3031 * Returns a list of pages by a given [start, end] of \a obj.
3032 *
3033 * \param resched If not NULL, then we give up before hogging CPU for too
3034 * long and set *resched = 1, in that case caller should implement a retry
3035 * logic.
3036 *
3037 * Gang tree lookup (radix_tree_gang_lookup()) optimization is absolutely
3038 * crucial in the face of [offset, EOF] locks.
3039 *
3040 * Return at least one page in @queue unless there is no covered page.
3041 */
3042int osc_page_gang_lookup(const struct lu_env *env, struct cl_io *io,
3043 struct osc_object *osc, pgoff_t start, pgoff_t end,
3044 osc_page_gang_cbt cb, void *cbdata)
3045{
3046 struct osc_page *ops;
3047 void **pvec;
3048 pgoff_t idx;
3049 unsigned int nr;
3050 unsigned int i;
3051 unsigned int j;
3052 int res = CLP_GANG_OKAY;
3053 bool tree_lock = true;
3054
3055 idx = start;
3056 pvec = osc_env_info(env)->oti_pvec;
3057 spin_lock(&osc->oo_tree_lock);
3058 while ((nr = radix_tree_gang_lookup(&osc->oo_tree, pvec,
3059 idx, OTI_PVEC_SIZE)) > 0) {
3060 struct cl_page *page;
3061 bool end_of_region = false;
3062
3063 for (i = 0, j = 0; i < nr; ++i) {
3064 ops = pvec[i];
3065 pvec[i] = NULL;
3066
3067 idx = osc_index(ops);
3068 if (idx > end) {
3069 end_of_region = true;
3070 break;
3071 }
3072
7addf402 3073 page = ops->ops_cl.cpl_page;
d9d47901
JX		 3074 LASSERT(page->cp_type == CPT_CACHEABLE);
3075 if (page->cp_state == CPS_FREEING)
3076 continue;
3077
3078 cl_page_get(page);
3079 lu_ref_add_atomic(&page->cp_reference,
3080 "gang_lookup", current);
3081 pvec[j++] = ops;
3082 }
3083 ++idx;
3084
3085 /*
3086 * Here a delicate locking dance is performed. Current thread
3087 * holds a reference to a page, but has to own it before it
3088 * can be placed into queue. Owning implies waiting, so
3089 * radix-tree lock is to be released. After a wait one has to
3090 * check that pages weren't truncated (cl_page_own() returns
3091 * error in the latter case).
3092 */
3093 spin_unlock(&osc->oo_tree_lock);
3094 tree_lock = false;
3095
3096 for (i = 0; i < j; ++i) {
3097 ops = pvec[i];
3098 if (res == CLP_GANG_OKAY)
3099 res = (*cb)(env, io, ops, cbdata);
3100
7addf402 3101 page = ops->ops_cl.cpl_page;
d9d47901
JX		 3102 lu_ref_del(&page->cp_reference, "gang_lookup", current);
3103 cl_page_put(env, page);
3104 }
3105 if (nr < OTI_PVEC_SIZE || end_of_region)
3106 break;
3107
3108 if (res == CLP_GANG_OKAY && need_resched())
3109 res = CLP_GANG_RESCHED;
3110 if (res != CLP_GANG_OKAY)
3111 break;
3112
3113 spin_lock(&osc->oo_tree_lock);
3114 tree_lock = true;
3115 }
3116 if (tree_lock)
3117 spin_unlock(&osc->oo_tree_lock);
3118 return res;
3119}
3120
3121/**
3122 * Check if page @page is covered by an extra lock or discard it.
3123 */
3124static int check_and_discard_cb(const struct lu_env *env, struct cl_io *io,
3125 struct osc_page *ops, void *cbdata)
3126{
3127 struct osc_thread_info *info = osc_env_info(env);
06563b56 3128 struct osc_object *osc = cbdata;
d9d47901
JX		 3129 pgoff_t index;
3130
3131 index = osc_index(ops);
3132 if (index >= info->oti_fn_index) {
06563b56 3133 struct ldlm_lock *tmp;
7addf402 3134 struct cl_page *page = ops->ops_cl.cpl_page;
d9d47901
JX		 3135
3136 /* refresh non-overlapped index */
06563b56 3137 tmp = osc_dlmlock_at_pgoff(env, osc, index, 0, 0);
d9d47901 3138 if (tmp) {
06563b56 3139 __u64 end = tmp->l_policy_data.l_extent.end;
d9d47901 3140 /* Cache the first-non-overlapped index so as to skip
06563b56
JX		 3141 * all pages within [index, oti_fn_index). This is safe
3142 * because if tmp lock is canceled, it will discard
3143 * these pages.
d9d47901 3144 */
06563b56
JX		 3145 info->oti_fn_index = cl_index(osc2cl(osc), end + 1);
3146 if (end == OBD_OBJECT_EOF)
d9d47901 3147 info->oti_fn_index = CL_PAGE_EOF;
06563b56 3148 LDLM_LOCK_PUT(tmp);
d9d47901
JX		 3149 } else if (cl_page_own(env, io, page) == 0) {
3150 /* discard the page */
3151 cl_page_discard(env, io, page);
3152 cl_page_disown(env, io, page);
3153 } else {
3154 LASSERT(page->cp_state == CPS_FREEING);
3155 }
3156 }
3157
3158 info->oti_next_index = index + 1;
3159 return CLP_GANG_OKAY;
3160}
3161
3162static int discard_cb(const struct lu_env *env, struct cl_io *io,
3163 struct osc_page *ops, void *cbdata)
3164{
3165 struct osc_thread_info *info = osc_env_info(env);
7addf402 3166 struct cl_page *page = ops->ops_cl.cpl_page;
d9d47901 3167
d9d47901
JX		 3168 /* page is top page. */
3169 info->oti_next_index = osc_index(ops) + 1;
3170 if (cl_page_own(env, io, page) == 0) {
77605e41 3171 KLASSERT(ergo(page->cp_type == CPT_CACHEABLE,
7addf402 3172 !PageDirty(cl_page_vmpage(page))));
77605e41 3173
d9d47901
JX		 3174 /* discard the page */
3175 cl_page_discard(env, io, page);
3176 cl_page_disown(env, io, page);
3177 } else {
3178 LASSERT(page->cp_state == CPS_FREEING);
3179 }
3180
3181 return CLP_GANG_OKAY;
3182}
3183
3184/**
3185 * Discard pages protected by the given lock. This function traverses radix
3186 * tree to find all covering pages and discard them. If a page is being covered
3187 * by other locks, it should remain in cache.
3188 *
3189 * If error happens on any step, the process continues anyway (the reasoning
3190 * behind this being that lock cancellation cannot be delayed indefinitely).
3191 */
06563b56
JX		 3192int osc_lock_discard_pages(const struct lu_env *env, struct osc_object *osc,
3193 pgoff_t start, pgoff_t end, enum cl_lock_mode mode)
d9d47901
JX		 3194{
3195 struct osc_thread_info *info = osc_env_info(env);
3196 struct cl_io *io = &info->oti_io;
d9d47901
JX		 3197 osc_page_gang_cbt cb;
3198 int res;
3199 int result;
3200
06563b56 3201 io->ci_obj = cl_object_top(osc2cl(osc));
d9d47901
JX		 3202 io->ci_ignore_layout = 1;
3203 result = cl_io_init(env, io, CIT_MISC, io->ci_obj);
3204 if (result != 0)
3205 goto out;
3206
06563b56 3207 cb = mode == CLM_READ ? check_and_discard_cb : discard_cb;
04a6284f
NC		 3208 info->oti_fn_index = start;
3209 info->oti_next_index = start;
d9d47901 3210 do {
06563b56
JX		 3211 res = osc_page_gang_lookup(env, io, osc,
3212 info->oti_next_index, end, cb, osc);
3213 if (info->oti_next_index > end)
d9d47901
JX		 3214 break;
3215
3216 if (res == CLP_GANG_RESCHED)
3217 cond_resched();
3218 } while (res != CLP_GANG_OKAY);
3219out:
3220 cl_io_fini(env, io);
3221 return result;
3222}
3223
d7e09d03 3224/** @} osc */
Linux kernel - Deliverables
This page took 0.697985 seconds and 5 git commands to generate.