4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2016, Intel Corporation.
29 * lustre/target/tgt_grant.c
31 * This file provides code related to grant space management on Lustre Targets
32 * (OSTs and MDTs). Grant is a mechanism used by client nodes to reserve disk
33 * space on a target for the data writeback cache. The Lustre client is thus
34 * assured that enough space will be available when flushing dirty pages
35 * asynchronously. Each client node is granted an initial amount of reserved
36 * space at connect time and gets additional space back from target in bulk
39 * We actually support three different cases:
40 * - The client supports the new grant parameters (i.e. OBD_CONNECT_GRANT_PARAM)
41 * which means that all grant overhead calculation happens on the client side.
42 * The server reports at connect time the backend filesystem block size, the
43 * maximum extent size as well as the extent insertion cost and it is then up
44 * to the osc layer to the track dirty extents and consume grant accordingly
45 * (see osc_cache.c). In each bulk write request, the client provides how much
46 * grant space was consumed for this RPC.
47 * - The client does not support OBD_CONNECT_GRANT_PARAM and always assumes a
48 * a backend file system block size of 4KB. We then have two cases:
49 * - If the block size is really 4KB, then the client can deal with grant
50 * allocation for partial block writes, but won't take extent insertion cost
51 * into account. For such clients, we inflate grant by 100% on the server
52 * side. It means that when 32MB of grant is hold by the client, 64MB of
53 * grant space is actually reserved on the server. All grant counters
54 * provided by such a client are inflated by 100%.
55 * - The backend filesystem block size is bigger than 4KB, which isn't
56 * supported by the client. In this case, we emulate a 4KB block size and
57 * consume one block size on the server for each 4KB of grant returned to
58 * client. With a 128KB blocksize, it means that 32MB dirty pages of 4KB
59 * on the client will actually consume 1GB of grant on the server.
60 * All grant counters provided by such a client are inflated by the block
63 * This file handles the core logic for:
64 * - grant allocation strategy
65 * - maintaining per-client as well as global grant space accounting
66 * - processing grant information packed in incoming requests
67 * - allocating server-side grant space for synchronous write RPCs which did not
68 * consume grant on the client side (OBD_BRW_FROM_GRANT flag not set). If not
69 * enough space is available, such RPCs fail with ENOSPC
71 * Author: Johann Lombardi <johann.lombardi@intel.com>
74 #define DEBUG_SUBSYSTEM S_FILTER
77 #include <obd_class.h>
79 #include "tgt_internal.h"
81 /* Clients typically hold 2x their max_rpcs_in_flight of grant space */
82 #define TGT_GRANT_SHRINK_LIMIT(exp) (2ULL * 8 * exp_max_brw_size(exp))
84 /* Helpers to inflate/deflate grants for clients that do not support the grant
86 static inline u64 tgt_grant_inflate(struct tg_grants_data *tgd, u64 val)
88 if (tgd->tgd_blockbits > COMPAT_BSIZE_SHIFT)
89 /* Client does not support such large block size, grant
90 * is thus inflated. We already significantly overestimate
91 * overhead, no need to add the extent tax in this case */
92 return val << (tgd->tgd_blockbits - COMPAT_BSIZE_SHIFT);
93 /* client can deal with the block size, but does not support per-extent
94 * grant accounting, inflate grant by 100% for such clients */
98 /* Companion of tgt_grant_inflate() */
99 static inline u64 tgt_grant_deflate(struct tg_grants_data *tgd, u64 val)
101 if (tgd->tgd_blockbits > COMPAT_BSIZE_SHIFT)
102 return val >> (tgd->tgd_blockbits - COMPAT_BSIZE_SHIFT);
106 /* Grant chunk is used as a unit for grant allocation. It should be inflated
107 * if the client does not support the grant paramaters.
108 * Check connection flag against \a data if not NULL. This is used during
109 * connection creation where exp->exp_connect_data isn't populated yet */
110 static inline u64 tgt_grant_chunk(struct obd_export *exp,
111 struct lu_target *lut,
112 struct obd_connect_data *data)
114 struct tg_grants_data *tgd = &lut->lut_tgd;
115 u64 chunk = exp_max_brw_size(exp);
118 if (exp->exp_obd->obd_self_export == exp)
119 /* Grant enough space to handle a big precreate request */
120 return OST_MAX_PRECREATE * lut->lut_dt_conf.ddp_inodespace / 2;
122 if ((data == NULL && !(exp_grant_param_supp(exp))) ||
123 (data != NULL && !OCD_HAS_FLAG(data, GRANT_PARAM)))
124 /* Try to grant enough space to send a full-size RPC */
125 return tgt_grant_inflate(tgd, chunk);
127 /* Try to return enough to send two full-size RPCs
128 * = 2 * (BRW_size + #extents_in_BRW * grant_tax) */
129 tax = 1ULL << tgd->tgd_blockbits; /* block size */
130 tax *= lut->lut_dt_conf.ddp_max_extent_blks; /* max extent size */
131 tax = (chunk + tax - 1) / tax; /* #extents in a RPC */
132 tax *= lut->lut_dt_conf.ddp_extent_tax; /* extent tax for a RPC */
133 chunk = (chunk + tax) * 2; /* we said two full RPCs */
137 static int tgt_check_export_grants(struct obd_export *exp, u64 *dirty,
138 u64 *pending, u64 *granted, u64 maxsize)
140 struct tg_export_data *ted = &exp->exp_target_data;
143 if (exp->exp_obd->obd_self_export == exp)
144 CDEBUG(D_CACHE, "%s: processing self export: %ld %ld "
145 "%ld\n", exp->exp_obd->obd_name, ted->ted_grant,
146 ted->ted_pending, ted->ted_dirty);
148 if (ted->ted_grant < 0 || ted->ted_pending < 0 || ted->ted_dirty < 0)
150 CDEBUG_LIMIT(level, "%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
151 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
152 ted->ted_dirty, ted->ted_pending, ted->ted_grant);
154 if (ted->ted_grant + ted->ted_pending > maxsize) {
155 CERROR("%s: cli %s/%p ted_grant(%ld) + ted_pending(%ld)"
156 " > maxsize(%llu)\n", exp->exp_obd->obd_name,
157 exp->exp_client_uuid.uuid, exp, ted->ted_grant,
158 ted->ted_pending, maxsize);
161 if (ted->ted_dirty > maxsize) {
162 CERROR("%s: cli %s/%p ted_dirty(%ld) > maxsize(%llu)\n",
163 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid,
164 exp, ted->ted_dirty, maxsize);
167 *granted += ted->ted_grant + ted->ted_pending;
168 *pending += ted->ted_pending;
169 *dirty += ted->ted_dirty;
174 * Perform extra sanity checks for grant accounting.
176 * This function scans the export list, sanity checks per-export grant counters
177 * and verifies accuracy of global grant accounting. If an inconsistency is
178 * found, a CERROR is printed with the function name \func that was passed as
179 * argument. LBUG is only called in case of serious counter corruption (i.e.
180 * value larger than the device size).
181 * Those sanity checks can be pretty expensive and are disabled if the OBD
182 * device has more than 100 connected exports.
184 * \param[in] obd OBD device for which grant accounting should be
186 * \param[in] func caller's function name
188 void tgt_grant_sanity_check(struct obd_device *obd, const char *func)
190 struct lu_target *lut = obd->u.obt.obt_lut;
191 struct tg_grants_data *tgd = &lut->lut_tgd;
192 struct obd_export *exp;
202 if (list_empty(&obd->obd_exports))
205 /* We don't want to do this for large machines that do lots of
206 * mounts or unmounts. It burns... */
207 if (obd->obd_num_exports > 100)
210 maxsize = tgd->tgd_osfs.os_blocks << tgd->tgd_blockbits;
212 spin_lock(&obd->obd_dev_lock);
213 spin_lock(&tgd->tgd_grant_lock);
214 list_for_each_entry(exp, &obd->obd_exports, exp_obd_chain) {
215 error = tgt_check_export_grants(exp, &tot_dirty, &tot_pending,
216 &tot_granted, maxsize);
218 spin_unlock(&obd->obd_dev_lock);
219 spin_unlock(&tgd->tgd_grant_lock);
224 /* exports about to be unlinked should also be taken into account since
225 * they might still hold pending grant space to be released at
227 list_for_each_entry(exp, &obd->obd_unlinked_exports, exp_obd_chain) {
228 error = tgt_check_export_grants(exp, &tot_dirty, &tot_pending,
229 &tot_granted, maxsize);
231 spin_unlock(&obd->obd_dev_lock);
232 spin_unlock(&tgd->tgd_grant_lock);
237 fo_tot_granted = tgd->tgd_tot_granted;
238 fo_tot_pending = tgd->tgd_tot_pending;
239 fo_tot_dirty = tgd->tgd_tot_dirty;
240 spin_unlock(&obd->obd_dev_lock);
241 spin_unlock(&tgd->tgd_grant_lock);
243 if (tot_granted != fo_tot_granted)
244 CERROR("%s: tot_granted %llu != fo_tot_granted %llu\n",
245 func, tot_granted, fo_tot_granted);
246 if (tot_pending != fo_tot_pending)
247 CERROR("%s: tot_pending %llu != fo_tot_pending %llu\n",
248 func, tot_pending, fo_tot_pending);
249 if (tot_dirty != fo_tot_dirty)
250 CERROR("%s: tot_dirty %llu != fo_tot_dirty %llu\n",
251 func, tot_dirty, fo_tot_dirty);
252 if (tot_pending > tot_granted)
253 CERROR("%s: tot_pending %llu > tot_granted %llu\n",
254 func, tot_pending, tot_granted);
255 if (tot_granted > maxsize)
256 CERROR("%s: tot_granted %llu > maxsize %llu\n",
257 func, tot_granted, maxsize);
258 if (tot_dirty > maxsize)
259 CERROR("%s: tot_dirty %llu > maxsize %llu\n",
260 func, tot_dirty, maxsize);
262 EXPORT_SYMBOL(tgt_grant_sanity_check);
265 * Get file system statistics of target.
267 * Helper function for statfs(), also used by grant code.
268 * Implements caching for statistics to avoid calling OSD device each time.
270 * \param[in] env execution environment
271 * \param[in] lut LU target
272 * \param[out] osfs statistic data to return
273 * \param[in] max_age maximum age for cached data
274 * \param[in] from_cache show that data was get from cache or not
276 * \retval 0 if successful
277 * \retval negative value on error
279 int tgt_statfs_internal(const struct lu_env *env, struct lu_target *lut,
280 struct obd_statfs *osfs, __u64 max_age, int *from_cache)
282 struct tg_grants_data *tgd = &lut->lut_tgd;
286 spin_lock(&tgd->tgd_osfs_lock);
287 if (cfs_time_before_64(tgd->tgd_osfs_age, max_age) || max_age == 0) {
290 /* statfs data are too old, get up-to-date one.
291 * we must be cautious here since multiple threads might be
292 * willing to update statfs data concurrently and we must
293 * grant that cached statfs data are always consistent */
295 if (tgd->tgd_statfs_inflight == 0)
296 /* clear inflight counter if no users, although it would
297 * take a while to overflow this 64-bit counter ... */
298 tgd->tgd_osfs_inflight = 0;
299 /* notify tgt_grant_commit() that we want to track writes
300 * completed as of now */
301 tgd->tgd_statfs_inflight++;
302 /* record value of inflight counter before running statfs to
303 * compute the diff once statfs is completed */
304 unstable = tgd->tgd_osfs_inflight;
305 spin_unlock(&tgd->tgd_osfs_lock);
307 /* statfs can sleep ... hopefully not for too long since we can
308 * call it fairly often as space fills up */
309 rc = dt_statfs(env, lut->lut_bottom, osfs);
313 spin_lock(&tgd->tgd_grant_lock);
314 spin_lock(&tgd->tgd_osfs_lock);
315 /* calculate how much space was written while we released the
317 unstable = tgd->tgd_osfs_inflight - unstable;
318 tgd->tgd_osfs_unstable = 0;
320 /* some writes committed while we were running statfs
321 * w/o the tgd_osfs_lock. Those ones got added to
322 * the cached statfs data that we are about to crunch.
323 * Take them into account in the new statfs data */
324 osfs->os_bavail -= min_t(u64, osfs->os_bavail,
325 unstable >> tgd->tgd_blockbits);
326 /* However, we don't really know if those writes got
327 * accounted in the statfs call, so tell
328 * tgt_grant_space_left() there is some uncertainty
329 * on the accounting of those writes.
330 * The purpose is to prevent spurious error messages in
331 * tgt_grant_space_left() since those writes might be
332 * accounted twice. */
333 tgd->tgd_osfs_unstable += unstable;
335 /* similarly, there is some uncertainty on write requests
336 * between prepare & commit */
337 tgd->tgd_osfs_unstable += tgd->tgd_tot_pending;
338 spin_unlock(&tgd->tgd_grant_lock);
340 /* finally udpate cached statfs data */
341 tgd->tgd_osfs = *osfs;
342 tgd->tgd_osfs_age = cfs_time_current_64();
344 tgd->tgd_statfs_inflight--; /* stop tracking */
345 if (tgd->tgd_statfs_inflight == 0)
346 tgd->tgd_osfs_inflight = 0;
347 spin_unlock(&tgd->tgd_osfs_lock);
352 /* use cached statfs data */
353 *osfs = tgd->tgd_osfs;
354 spin_unlock(&tgd->tgd_osfs_lock);
363 EXPORT_SYMBOL(tgt_statfs_internal);
366 * Update cached statfs information from the OSD layer
368 * Refresh statfs information cached in tgd::tgd_osfs if the cache is older
369 * than 1s or if force is set. The OSD layer is in charge of estimating data &
371 * This function can sleep so it should not be called with any spinlock held.
373 * \param[in] env LU environment passed by the caller
374 * \param[in] exp export used to print client info in debug
376 * \param[in] force force a refresh of statfs information
377 * \param[out] from_cache returns whether the statfs information are
380 static void tgt_grant_statfs(const struct lu_env *env, struct obd_export *exp,
381 int force, int *from_cache)
383 struct obd_device *obd = exp->exp_obd;
384 struct lu_target *lut = obd->u.obt.obt_lut;
385 struct tg_grants_data *tgd = &lut->lut_tgd;
386 struct tgt_thread_info *tti;
387 struct obd_statfs *osfs;
392 max_age = 0; /* get fresh statfs data */
394 max_age = cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS);
396 tti = tgt_th_info(env);
397 osfs = &tti->tti_u.osfs;
398 rc = tgt_statfs_internal(env, lut, osfs, max_age, from_cache);
405 CDEBUG(D_CACHE, "%s: cli %s/%p free: %llu avail: %llu\n",
406 obd->obd_name, exp->exp_client_uuid.uuid, exp,
407 osfs->os_bfree << tgd->tgd_blockbits,
408 osfs->os_bavail << tgd->tgd_blockbits);
412 * Figure out how much space is available on the backend filesystem after
413 * removing grant space already booked by clients.
415 * This is done by accessing cached statfs data previously populated by
416 * tgt_grant_statfs(), from which we withdraw the space already granted to
417 * clients and the reserved space.
418 * Caller must hold tgd_grant_lock spinlock.
420 * \param[in] exp export associated with the device for which the amount
421 * of available space is requested
422 * \retval amount of non-allocated space, in bytes
424 static u64 tgt_grant_space_left(struct obd_export *exp)
426 struct obd_device *obd = exp->exp_obd;
427 struct lu_target *lut = obd->u.obt.obt_lut;
428 struct tg_grants_data *tgd = &lut->lut_tgd;
435 assert_spin_locked(&tgd->tgd_grant_lock);
437 spin_lock(&tgd->tgd_osfs_lock);
438 /* get available space from cached statfs data */
439 left = tgd->tgd_osfs.os_bavail << tgd->tgd_blockbits;
440 unstable = tgd->tgd_osfs_unstable; /* those might be accounted twice */
441 spin_unlock(&tgd->tgd_osfs_lock);
443 tot_granted = tgd->tgd_tot_granted;
445 if (left < tot_granted) {
446 int mask = (left + unstable <
447 tot_granted - tgd->tgd_tot_pending) ?
450 CDEBUG_LIMIT(mask, "%s: cli %s/%p left %llu < tot_grant "
451 "%llu unstable %llu pending %llu "
453 obd->obd_name, exp->exp_client_uuid.uuid, exp,
454 left, tot_granted, unstable,
455 tgd->tgd_tot_pending,
461 /* Withdraw space already granted to clients */
464 /* Align left on block size */
465 left &= ~((1ULL << tgd->tgd_blockbits) - 1);
467 CDEBUG(D_CACHE, "%s: cli %s/%p avail %llu left %llu unstable "
468 "%llu tot_grant %llu pending %llu\n", obd->obd_name,
469 exp->exp_client_uuid.uuid, exp, avail, left, unstable,
470 tot_granted, tgd->tgd_tot_pending);
476 * Process grant information from obdo structure packed in incoming BRW
477 * and inflate grant counters if required.
479 * Grab the dirty and seen grant announcements from the incoming obdo and
480 * inflate all grant counters passed in the request if the client does not
481 * support the grant parameters.
482 * We will later calculate the client's new grant and return it.
483 * Caller must hold tgd_grant_lock spinlock.
485 * \param[in] env LU environment supplying osfs storage
486 * \param[in] exp export for which we received the request
487 * \param[in,out] oa incoming obdo sent by the client
489 static void tgt_grant_incoming(const struct lu_env *env, struct obd_export *exp,
490 struct obdo *oa, long chunk)
492 struct tg_export_data *ted = &exp->exp_target_data;
493 struct obd_device *obd = exp->exp_obd;
494 struct tg_grants_data *tgd = &obd->u.obt.obt_lut->lut_tgd;
499 assert_spin_locked(&tgd->tgd_grant_lock);
501 if ((oa->o_valid & (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) !=
502 (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) {
503 oa->o_valid &= ~OBD_MD_FLGRANT;
507 /* Add some margin, since there is a small race if other RPCs arrive
508 * out-or-order and have already consumed some grant. We want to
509 * leave this here in case there is a large error in accounting. */
511 "%s: cli %s/%p reports grant %llu dropped %u, local %lu\n",
512 obd->obd_name, exp->exp_client_uuid.uuid, exp, oa->o_grant,
513 oa->o_dropped, ted->ted_grant);
515 if ((long long)oa->o_dirty < 0)
518 /* inflate grant counters if required */
519 if (!exp_grant_param_supp(exp)) {
520 oa->o_grant = tgt_grant_inflate(tgd, oa->o_grant);
521 oa->o_dirty = tgt_grant_inflate(tgd, oa->o_dirty);
522 oa->o_dropped = tgt_grant_inflate(tgd, (u64)oa->o_dropped);
523 oa->o_undirty = tgt_grant_inflate(tgd, oa->o_undirty);
527 dropped = oa->o_dropped;
529 /* Update our accounting now so that statfs takes it into account.
530 * Note that ted_dirty is only approximate and can become incorrect
531 * if RPCs arrive out-of-order. No important calculations depend
532 * on ted_dirty however, but we must check sanity to not assert. */
533 if (dirty > ted->ted_grant + 4 * chunk)
534 dirty = ted->ted_grant + 4 * chunk;
535 tgd->tgd_tot_dirty += dirty - ted->ted_dirty;
536 if (ted->ted_grant < dropped) {
538 "%s: cli %s/%p reports %lu dropped > grant %lu\n",
539 obd->obd_name, exp->exp_client_uuid.uuid, exp, dropped,
543 if (tgd->tgd_tot_granted < dropped) {
544 CERROR("%s: cli %s/%p reports %lu dropped > tot_grant %llu\n",
545 obd->obd_name, exp->exp_client_uuid.uuid, exp,
546 dropped, tgd->tgd_tot_granted);
549 tgd->tgd_tot_granted -= dropped;
550 ted->ted_grant -= dropped;
551 ted->ted_dirty = dirty;
553 if (ted->ted_dirty < 0 || ted->ted_grant < 0 || ted->ted_pending < 0) {
554 CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
555 obd->obd_name, exp->exp_client_uuid.uuid, exp,
556 ted->ted_dirty, ted->ted_pending, ted->ted_grant);
557 spin_unlock(&tgd->tgd_grant_lock);
564 * Grant shrink request handler.
566 * Client nodes can explicitly release grant space (i.e. process called grant
567 * shrinking). This function proceeds with the shrink request when there is
568 * less ungranted space remaining than the amount all of the connected clients
569 * would consume if they used their full grant.
570 * Caller must hold tgd_grant_lock spinlock.
572 * \param[in] exp export releasing grant space
573 * \param[in,out] oa incoming obdo sent by the client
574 * \param[in] left_space remaining free space with space already granted
577 static void tgt_grant_shrink(struct obd_export *exp, struct obdo *oa,
580 struct tg_export_data *ted = &exp->exp_target_data;
581 struct obd_device *obd = exp->exp_obd;
582 struct tg_grants_data *tgd = &obd->u.obt.obt_lut->lut_tgd;
585 assert_spin_locked(&tgd->tgd_grant_lock);
587 if (left_space >= tgd->tgd_tot_granted_clients *
588 TGT_GRANT_SHRINK_LIMIT(exp))
591 grant_shrink = oa->o_grant;
593 ted->ted_grant -= grant_shrink;
594 tgd->tgd_tot_granted -= grant_shrink;
596 CDEBUG(D_CACHE, "%s: cli %s/%p shrink %ld ted_grant %ld total %llu\n",
597 obd->obd_name, exp->exp_client_uuid.uuid, exp, grant_shrink,
598 ted->ted_grant, tgd->tgd_tot_granted);
600 /* client has just released some grant, don't grant any space back */
605 * Calculate how much space is required to write a given network buffer
607 * This function takes block alignment into account to estimate how much on-disk
608 * space will be required to successfully write the whole niobuf.
609 * Estimated space is inflated if the export does not support
610 * OBD_CONNECT_GRANT_PARAM and if the backend filesystem has a block size
611 * larger than the minimal supported page size (i.e. 4KB).
613 * \param[in] exp export associated which the write request
614 * if NULL, then size estimate is done for server-side
616 * \param[in] lut LU target handling the request
617 * \param[in] rnb network buffer to estimate size of
619 * \retval space (in bytes) that will be consumed to write the
622 static inline u64 tgt_grant_rnb_size(struct obd_export *exp,
623 struct lu_target *lut,
624 struct niobuf_remote *rnb)
626 struct tg_grants_data *tgd = &lut->lut_tgd;
631 if (exp && !exp_grant_param_supp(exp) &&
632 tgd->tgd_blockbits > COMPAT_BSIZE_SHIFT)
633 blksize = 1ULL << COMPAT_BSIZE_SHIFT;
635 blksize = 1ULL << tgd->tgd_blockbits;
637 /* The network buffer might span several blocks, align it on block
639 bytes = rnb->rnb_offset & (blksize - 1);
640 bytes += rnb->rnb_len;
641 end = bytes & (blksize - 1);
643 bytes += blksize - end;
645 if (exp == NULL || exp_grant_param_supp(exp)) {
646 /* add per-extent insertion cost */
650 max_ext = blksize * lut->lut_dt_conf.ddp_max_extent_blks;
651 nr_ext = (bytes + max_ext - 1) / max_ext;
652 bytes += nr_ext * lut->lut_dt_conf.ddp_extent_tax;
654 /* Inflate grant space if client does not support extent-based
655 * grant allocation */
656 bytes = tgt_grant_inflate(tgd, (u64)bytes);
663 * Validate grant accounting for each incoming remote network buffer.
665 * When clients have dirtied as much space as they've been granted they
666 * fall through to sync writes. These sync writes haven't been expressed
667 * in grants and need to error with ENOSPC when there isn't room in the
668 * filesystem for them after grants are taken into account. However,
669 * writeback of the dirty data that was already granted space can write
671 * The OBD_BRW_GRANTED flag will be set in the rnb_flags of each network
672 * buffer which has been granted enough space to proceed. Buffers without
673 * this flag will fail to be written with -ENOSPC (see tgt_preprw_write().
674 * Caller must hold tgd_grant_lock spinlock.
676 * \param[in] env LU environment passed by the caller
677 * \param[in] exp export identifying the client which sent the RPC
678 * \param[in] oa incoming obdo in which we should return the pack the
680 * \param[in,out] rnb the list of network buffers
681 * \param[in] niocount the number of network buffers in the list
682 * \param[in] left the remaining free space with space already granted
685 static void tgt_grant_check(const struct lu_env *env, struct obd_export *exp,
686 struct obdo *oa, struct niobuf_remote *rnb,
687 int niocount, u64 *left)
689 struct tg_export_data *ted = &exp->exp_target_data;
690 struct obd_device *obd = exp->exp_obd;
691 struct lu_target *lut = obd->u.obt.obt_lut;
692 struct tg_grants_data *tgd = &lut->lut_tgd;
693 unsigned long ungranted = 0;
694 unsigned long granted = 0;
700 assert_spin_locked(&tgd->tgd_grant_lock);
702 if (obd->obd_recovering) {
703 /* Replaying write. Grant info have been processed already so no
704 * need to do any enforcement here. It is worth noting that only
705 * bulk writes with all rnbs having OBD_BRW_FROM_GRANT can be
706 * replayed. If one page hasn't OBD_BRW_FROM_GRANT set, then
707 * the whole bulk is written synchronously */
709 CDEBUG(D_CACHE, "Replaying write, skipping accounting\n");
710 } else if ((oa->o_valid & OBD_MD_FLFLAGS) &&
711 (oa->o_flags & OBD_FL_RECOV_RESEND)) {
712 /* Recoverable resend, grant info have already been processed as
715 CDEBUG(D_CACHE, "Recoverable resend arrived, skipping "
717 } else if (exp_grant_param_supp(exp) && oa->o_grant_used > 0) {
718 /* Client supports the new grant parameters and is telling us
719 * how much grant space it consumed for this bulk write.
720 * Although all rnbs are supposed to have the OBD_BRW_FROM_GRANT
721 * flag set, we will scan the rnb list and looks for non-cache
722 * I/O in case it changes in the future */
723 if (ted->ted_grant >= oa->o_grant_used) {
724 /* skip grant accounting for rnbs with
725 * OBD_BRW_FROM_GRANT and just used grant consumption
726 * claimed in the request */
727 granted = oa->o_grant_used;
730 /* client has used more grants for this request that
732 CERROR("%s: cli %s claims %lu GRANT, real grant %lu\n",
733 exp->exp_obd->obd_name,
734 exp->exp_client_uuid.uuid,
735 (unsigned long)oa->o_grant_used, ted->ted_grant);
737 /* check whether we can fill the gap with unallocated
739 if (*left > (oa->o_grant_used - ted->ted_grant)) {
740 /* ouf .. we are safe for now */
741 granted = ted->ted_grant;
742 ungranted = oa->o_grant_used - granted;
746 /* too bad, but we cannot afford to blow up our grant
747 * accounting. The loop below will handle each rnb in
752 for (i = 0; i < niocount; i++) {
755 if ((rnb[i].rnb_flags & OBD_BRW_FROM_GRANT)) {
757 rnb[i].rnb_flags |= OBD_BRW_GRANTED;
761 /* compute how much grant space is actually needed for
762 * this rnb, inflate grant if required */
763 bytes = tgt_grant_rnb_size(exp, lut, &rnb[i]);
764 if (ted->ted_grant >= granted + bytes) {
766 rnb[i].rnb_flags |= OBD_BRW_GRANTED;
770 CDEBUG(D_CACHE, "%s: cli %s/%p claims %ld+%d GRANT, "
771 "real grant %lu idx %d\n", obd->obd_name,
772 exp->exp_client_uuid.uuid, exp, granted, bytes,
776 if (obd->obd_recovering)
777 CERROR("%s: cli %s is replaying OST_WRITE while one rnb"
778 " hasn't OBD_BRW_FROM_GRANT set (0x%x)\n",
779 obd->obd_name, exp->exp_client_uuid.uuid,
782 /* Consume grant space on the server.
783 * Unlike above, tgt_grant_rnb_size() is called with exp = NULL
784 * so that the required grant space isn't inflated. This is
785 * done on purpose since the server can deal with large block
786 * size, unlike some clients */
787 bytes = tgt_grant_rnb_size(NULL, lut, &rnb[i]);
789 /* if enough space, pretend it was granted */
792 rnb[i].rnb_flags |= OBD_BRW_GRANTED;
796 /* We can't check for already-mapped blocks here (make sense
797 * when backend filesystem does not use COW) as it requires
798 * dropping the grant lock.
799 * Instead, we clear OBD_BRW_GRANTED and in that case we need
800 * to go through and verify if all of the blocks not marked
801 * BRW_GRANTED are already mapped and we can ignore this error.
803 rnb[i].rnb_flags &= ~OBD_BRW_GRANTED;
804 CDEBUG(D_CACHE, "%s: cli %s/%p idx %d no space for %d\n",
805 obd->obd_name, exp->exp_client_uuid.uuid, exp, i, bytes);
808 /* record in o_grant_used the actual space reserved for the I/O, will be
809 * used later in tgt_grant_commmit() */
810 oa->o_grant_used = granted + ungranted;
812 /* record space used for the I/O, will be used in tgt_grant_commmit() */
813 /* Now substract what the clients has used already. We don't subtract
814 * this from the tot_granted yet, so that other client's can't grab
815 * that space before we have actually allocated our blocks. That
816 * happens in tgt_grant_commit() after the writes are done. */
817 ted->ted_grant -= granted;
818 ted->ted_pending += oa->o_grant_used;
819 tgd->tgd_tot_granted += ungranted;
820 tgd->tgd_tot_pending += oa->o_grant_used;
823 "%s: cli %s/%p granted: %lu ungranted: %lu grant: %lu dirty: %lu"
824 "\n", obd->obd_name, exp->exp_client_uuid.uuid, exp,
825 granted, ungranted, ted->ted_grant, ted->ted_dirty);
827 if (obd->obd_recovering || (oa->o_valid & OBD_MD_FLGRANT) == 0)
828 /* don't update dirty accounting during recovery or
829 * if grant information got discarded (e.g. during resend) */
832 if (ted->ted_dirty < granted) {
833 CWARN("%s: cli %s/%p claims granted %lu > ted_dirty %lu\n",
834 obd->obd_name, exp->exp_client_uuid.uuid, exp,
835 granted, ted->ted_dirty);
836 granted = ted->ted_dirty;
838 tgd->tgd_tot_dirty -= granted;
839 ted->ted_dirty -= granted;
841 if (ted->ted_dirty < 0 || ted->ted_grant < 0 || ted->ted_pending < 0) {
842 CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
843 obd->obd_name, exp->exp_client_uuid.uuid, exp,
844 ted->ted_dirty, ted->ted_pending, ted->ted_grant);
845 spin_unlock(&tgd->tgd_grant_lock);
852 * Allocate additional grant space to a client
854 * Calculate how much grant space to return to client, based on how much space
855 * is currently free and how much of that is already granted.
856 * Caller must hold tgd_grant_lock spinlock.
858 * \param[in] exp export of the client which sent the request
859 * \param[in] curgrant current grant claimed by the client
860 * \param[in] want how much grant space the client would like to
862 * \param[in] left remaining free space with granted space taken
864 * \param[in] conservative if set to true, the server should be cautious
865 * and limit how much space is granted back to the
866 * client. Otherwise, the server should try hard to
867 * satisfy the client request.
869 * \retval amount of grant space allocated
871 static long tgt_grant_alloc(struct obd_export *exp, u64 curgrant,
872 u64 want, u64 left, long chunk,
875 struct obd_device *obd = exp->exp_obd;
876 struct tg_grants_data *tgd = &obd->u.obt.obt_lut->lut_tgd;
877 struct tg_export_data *ted = &exp->exp_target_data;
882 /* When tgd_grant_compat_disable is set, we don't grant any space to
883 * clients not supporting OBD_CONNECT_GRANT_PARAM.
884 * Otherwise, space granted to such a client is inflated since it
885 * consumes PAGE_SIZE of grant space per block */
886 if ((obd->obd_self_export != exp && !exp_grant_param_supp(exp) &&
887 tgd->tgd_grant_compat_disable) || left == 0 || exp->exp_failed)
890 if (want > 0x7fffffff) {
891 CERROR("%s: client %s/%p requesting > 2GB grant %llu\n",
892 obd->obd_name, exp->exp_client_uuid.uuid, exp, want);
896 /* Grant some fraction of the client's requested grant space so that
897 * they are not always waiting for write credits (not all of it to
898 * avoid overgranting in face of multiple RPCs in flight). This
899 * essentially will be able to control the OSC_MAX_RIF for a client.
901 * If we do have a large disparity between what the client thinks it
902 * has and what we think it has, don't grant very much and let the
903 * client consume its grant first. Either it just has lots of RPCs
904 * in flight, or it was evicted and its grants will soon be used up. */
905 if (curgrant >= want || curgrant >= ted->ted_grant + chunk)
908 if (obd->obd_recovering)
909 conservative = false;
912 /* don't grant more than 1/8th of the remaining free space in
915 grant = min(want - curgrant, left);
916 /* round grant up to the next block size */
917 grant = (grant + (1 << tgd->tgd_blockbits) - 1) &
918 ~((1ULL << tgd->tgd_blockbits) - 1);
923 /* Limit to grant_chunk if not reconnect/recovery */
924 if ((grant > chunk) && conservative)
927 tgd->tgd_tot_granted += grant;
928 ted->ted_grant += grant;
930 if (ted->ted_grant < 0) {
931 CERROR("%s: cli %s/%p grant %ld want %llu current %llu\n",
932 obd->obd_name, exp->exp_client_uuid.uuid, exp,
933 ted->ted_grant, want, curgrant);
934 spin_unlock(&tgd->tgd_grant_lock);
939 "%s: cli %s/%p wants: %llu current grant %llu"
940 " granting: %llu\n", obd->obd_name, exp->exp_client_uuid.uuid,
941 exp, want, curgrant, grant);
943 "%s: cli %s/%p tot cached:%llu granted:%llu"
944 " num_exports: %d\n", obd->obd_name, exp->exp_client_uuid.uuid,
945 exp, tgd->tgd_tot_dirty, tgd->tgd_tot_granted,
946 obd->obd_num_exports);
952 * Handle grant space allocation on client connection & reconnection.
954 * A new non-readonly connection gets an initial grant allocation equals to
955 * tgt_grant_chunk() (i.e. twice the max BRW size in most of the cases).
956 * On reconnection, grant counters between client & target are resynchronized
957 * and additional space might be granted back if possible.
959 * \param[in] env LU environment provided by the caller
960 * \param[in] exp client's export which is (re)connecting
961 * \param[in,out] data obd_connect_data structure sent by the client in the
963 * \param[in] new_conn must set to true if this is a new connection and false
966 void tgt_grant_connect(const struct lu_env *env, struct obd_export *exp,
967 struct obd_connect_data *data, bool new_conn)
969 struct lu_target *lut = exp->exp_obd->u.obt.obt_lut;
970 struct tg_grants_data *tgd = &lut->lut_tgd;
971 struct tg_export_data *ted = &exp->exp_target_data;
976 int force = 0; /* can use cached data */
978 /* don't grant space to client with read-only access */
979 if (OCD_HAS_FLAG(data, RDONLY) ||
980 (!OCD_HAS_FLAG(data, GRANT_PARAM) &&
981 tgd->tgd_grant_compat_disable)) {
983 data->ocd_connect_flags &= ~(OBD_CONNECT_GRANT |
984 OBD_CONNECT_GRANT_PARAM);
988 if (OCD_HAS_FLAG(data, GRANT_PARAM))
989 want = data->ocd_grant;
991 want = tgt_grant_inflate(tgd, data->ocd_grant);
992 chunk = tgt_grant_chunk(exp, lut, data);
994 tgt_grant_statfs(env, exp, force, &from_cache);
996 spin_lock(&tgd->tgd_grant_lock);
998 /* Grab free space from cached info and take out space already granted
999 * to clients as well as reserved space */
1000 left = tgt_grant_space_left(exp);
1002 /* get fresh statfs data if we are short in ungranted space */
1003 if (from_cache && left < 32 * chunk) {
1004 spin_unlock(&tgd->tgd_grant_lock);
1005 CDEBUG(D_CACHE, "fs has no space left and statfs too old\n");
1010 tgt_grant_alloc(exp, (u64)ted->ted_grant, want, left, chunk, new_conn);
1012 /* return to client its current grant */
1013 if (OCD_HAS_FLAG(data, GRANT_PARAM))
1014 data->ocd_grant = ted->ted_grant;
1017 data->ocd_grant = tgt_grant_deflate(tgd, (u64)ted->ted_grant);
1019 /* reset dirty accounting */
1020 tgd->tgd_tot_dirty -= ted->ted_dirty;
1023 if (new_conn && OCD_HAS_FLAG(data, GRANT))
1024 tgd->tgd_tot_granted_clients++;
1026 spin_unlock(&tgd->tgd_grant_lock);
1028 CDEBUG(D_CACHE, "%s: cli %s/%p ocd_grant: %d want: %llu left: %llu\n",
1029 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid,
1030 exp, data->ocd_grant, want, left);
1034 EXPORT_SYMBOL(tgt_grant_connect);
1037 * Release all grant space attached to a given export.
1039 * Remove a client from the grant accounting totals. We also remove
1040 * the export from the obd device under the osfs and dev locks to ensure
1041 * that the tgt_grant_sanity_check() calculations are always valid.
1042 * The client should do something similar when it invalidates its import.
1044 * \param[in] exp client's export to remove from grant accounting
1046 void tgt_grant_discard(struct obd_export *exp)
1048 struct obd_device *obd = exp->exp_obd;
1049 struct tg_grants_data *tgd = &obd->u.obt.obt_lut->lut_tgd;
1050 struct tg_export_data *ted = &exp->exp_target_data;
1052 spin_lock(&tgd->tgd_grant_lock);
1053 LASSERTF(tgd->tgd_tot_granted >= ted->ted_grant,
1054 "%s: tot_granted %llu cli %s/%p ted_grant %ld\n",
1055 obd->obd_name, tgd->tgd_tot_granted,
1056 exp->exp_client_uuid.uuid, exp, ted->ted_grant);
1057 tgd->tgd_tot_granted -= ted->ted_grant;
1059 LASSERTF(tgd->tgd_tot_pending >= ted->ted_pending,
1060 "%s: tot_pending %llu cli %s/%p ted_pending %ld\n",
1061 obd->obd_name, tgd->tgd_tot_pending,
1062 exp->exp_client_uuid.uuid, exp, ted->ted_pending);
1063 /* tgd_tot_pending is handled in tgt_grant_commit as bulk
1065 LASSERTF(tgd->tgd_tot_dirty >= ted->ted_dirty,
1066 "%s: tot_dirty %llu cli %s/%p ted_dirty %ld\n",
1067 obd->obd_name, tgd->tgd_tot_dirty,
1068 exp->exp_client_uuid.uuid, exp, ted->ted_dirty);
1069 tgd->tgd_tot_dirty -= ted->ted_dirty;
1071 spin_unlock(&tgd->tgd_grant_lock);
1073 EXPORT_SYMBOL(tgt_grant_discard);
1076 * Process grant information from incoming bulk read request.
1078 * Extract grant information packed in obdo structure (OBD_MD_FLGRANT set in
1079 * o_valid). Bulk reads usually comes with grant announcements (number of dirty
1080 * blocks, remaining amount of grant space, ...) and could also include a grant
1081 * shrink request. Unlike bulk write, no additional grant space is returned on
1082 * bulk read request.
1084 * \param[in] env is the lu environment provided by the caller
1085 * \param[in] exp is the export of the client which sent the request
1086 * \param[in,out] oa is the incoming obdo sent by the client
1088 void tgt_grant_prepare_read(const struct lu_env *env,
1089 struct obd_export *exp, struct obdo *oa)
1091 struct lu_target *lut = exp->exp_obd->u.obt.obt_lut;
1092 struct tg_grants_data *tgd = &lut->lut_tgd;
1101 if ((oa->o_valid & OBD_MD_FLGRANT) == 0)
1102 /* The read request does not contain any grant
1106 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
1107 (oa->o_flags & OBD_FL_SHRINK_GRANT)) {
1108 /* To process grant shrink request, we need to know how much
1109 * available space remains on the backend filesystem.
1110 * Shrink requests are not so common, we always get fresh
1111 * statfs information. */
1112 tgt_grant_statfs(env, exp, 1, NULL);
1114 /* protect all grant counters */
1115 spin_lock(&tgd->tgd_grant_lock);
1117 /* Grab free space from cached statfs data and take out space
1118 * already granted to clients as well as reserved space */
1119 left = tgt_grant_space_left(exp);
1121 /* all set now to proceed with shrinking */
1124 /* no grant shrinking request packed in the obdo and
1125 * since we don't grant space back on reads, no point
1126 * in running statfs, so just skip it and process
1127 * incoming grant data directly. */
1128 spin_lock(&tgd->tgd_grant_lock);
1132 /* extract incoming grant information provided by the client and
1133 * inflate grant counters if required */
1134 tgt_grant_incoming(env, exp, oa, tgt_grant_chunk(exp, lut, NULL));
1136 /* unlike writes, we don't return grants back on reads unless a grant
1137 * shrink request was packed and we decided to turn it down. */
1139 tgt_grant_shrink(exp, oa, left);
1143 if (!exp_grant_param_supp(exp))
1144 oa->o_grant = tgt_grant_deflate(tgd, oa->o_grant);
1145 spin_unlock(&tgd->tgd_grant_lock);
1148 EXPORT_SYMBOL(tgt_grant_prepare_read);
1151 * Process grant information from incoming bulk write request.
1153 * This function extracts client's grant announcements from incoming bulk write
1154 * request and attempts to allocate grant space for network buffers that need it
1155 * (i.e. OBD_BRW_FROM_GRANT not set in rnb_fags).
1156 * Network buffers which aren't granted the OBD_BRW_GRANTED flag should not
1157 * proceed further and should fail with -ENOSPC.
1158 * Whenever possible, additional grant space will be returned to the client
1159 * in the bulk write reply.
1160 * tgt_grant_prepare_write() must be called before writting any buffers to
1161 * the backend storage. This function works in pair with tgt_grant_commit()
1162 * which must be invoked once all buffers have been written to disk in order
1163 * to release space from the pending grant counter.
1165 * \param[in] env LU environment provided by the caller
1166 * \param[in] exp export of the client which sent the request
1167 * \param[in] oa incoming obdo sent by the client
1168 * \param[in] rnb list of network buffers
1169 * \param[in] niocount number of network buffers in the list
1171 void tgt_grant_prepare_write(const struct lu_env *env,
1172 struct obd_export *exp, struct obdo *oa,
1173 struct niobuf_remote *rnb, int niocount)
1175 struct obd_device *obd = exp->exp_obd;
1176 struct lu_target *lut = obd->u.obt.obt_lut;
1177 struct tg_grants_data *tgd = &lut->lut_tgd;
1180 int force = 0; /* can use cached data intially */
1181 long chunk = tgt_grant_chunk(exp, lut, NULL);
1186 /* get statfs information from OSD layer */
1187 tgt_grant_statfs(env, exp, force, &from_cache);
1189 spin_lock(&tgd->tgd_grant_lock); /* protect all grant counters */
1191 /* Grab free space from cached statfs data and take out space already
1192 * granted to clients as well as reserved space */
1193 left = tgt_grant_space_left(exp);
1195 /* Get fresh statfs data if we are short in ungranted space */
1196 if (from_cache && left < 32 * chunk) {
1197 spin_unlock(&tgd->tgd_grant_lock);
1198 CDEBUG(D_CACHE, "%s: fs has no space left and statfs too old\n",
1204 /* When close to free space exhaustion, trigger a sync to force
1205 * writeback cache to consume required space immediately and release as
1206 * much space as possible. */
1207 if (!obd->obd_recovering && force != 2 && left < chunk) {
1208 bool from_grant = true;
1211 /* That said, it is worth running a sync only if some pages did
1212 * not consume grant space on the client and could thus fail
1213 * with ENOSPC later in tgt_grant_check() */
1214 for (i = 0; i < niocount; i++)
1215 if (!(rnb[i].rnb_flags & OBD_BRW_FROM_GRANT))
1219 /* at least one network buffer requires acquiring grant
1220 * space on the server */
1221 spin_unlock(&tgd->tgd_grant_lock);
1222 /* discard errors, at least we tried ... */
1223 dt_sync(env, lut->lut_bottom);
1229 /* extract incoming grant information provided by the client,
1230 * and inflate grant counters if required */
1231 tgt_grant_incoming(env, exp, oa, chunk);
1234 tgt_grant_check(env, exp, oa, rnb, niocount, &left);
1236 if (!(oa->o_valid & OBD_MD_FLGRANT)) {
1237 spin_unlock(&tgd->tgd_grant_lock);
1241 /* if OBD_FL_SHRINK_GRANT is set, the client is willing to release some
1243 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
1244 (oa->o_flags & OBD_FL_SHRINK_GRANT))
1245 tgt_grant_shrink(exp, oa, left);
1247 /* grant more space back to the client if possible */
1248 oa->o_grant = tgt_grant_alloc(exp, oa->o_grant, oa->o_undirty,
1251 if (!exp_grant_param_supp(exp))
1252 oa->o_grant = tgt_grant_deflate(tgd, oa->o_grant);
1253 spin_unlock(&tgd->tgd_grant_lock);
1256 EXPORT_SYMBOL(tgt_grant_prepare_write);
1259 * Consume grant space reserved for object creation.
1261 * Grant space is allocated to the local self export for object precreation.
1262 * This is required to prevent object precreation from consuming grant space
1263 * allocated to client nodes for the data writeback cache.
1264 * This function consumes enough space to create \a nr objects and allocates
1265 * more grant space to the self export for future precreation requests, if
1268 * \param[in] env LU environment provided by the caller
1269 * \param[in] exp export holding the grant space for precreation (= self
1271 * \param[in] nr number of objects to be created
1273 * \retval >= 0 amount of grant space allocated to the precreate request
1274 * \retval -ENOSPC on failure
1276 long tgt_grant_create(const struct lu_env *env, struct obd_export *exp, int *nr)
1278 struct lu_target *lut = exp->exp_obd->u.obt.obt_lut;
1279 struct tg_grants_data *tgd = &lut->lut_tgd;
1280 struct tg_export_data *ted = &exp->exp_target_data;
1282 unsigned long wanted;
1283 unsigned long granted;
1286 if (exp->exp_obd->obd_recovering ||
1287 lut->lut_dt_conf.ddp_inodespace == 0)
1288 /* don't enforce grant during recovery */
1291 /* Update statfs data if required */
1292 tgt_grant_statfs(env, exp, 1, NULL);
1294 /* protect all grant counters */
1295 spin_lock(&tgd->tgd_grant_lock);
1297 /* fail precreate request if there is not enough blocks available for
1299 if (tgd->tgd_osfs.os_bavail - (ted->ted_grant >> tgd->tgd_blockbits) <
1300 (tgd->tgd_osfs.os_blocks >> 10)) {
1301 spin_unlock(&tgd->tgd_grant_lock);
1302 CDEBUG(D_RPCTRACE, "%s: not enough space for create %llu\n",
1303 exp->exp_obd->obd_name,
1304 tgd->tgd_osfs.os_bavail * tgd->tgd_osfs.os_blocks);
1308 /* Grab free space from cached statfs data and take out space
1309 * already granted to clients as well as reserved space */
1310 left = tgt_grant_space_left(exp);
1312 /* compute how much space is required to handle the precreation
1314 wanted = *nr * lut->lut_dt_conf.ddp_inodespace;
1315 if (wanted > ted->ted_grant + left) {
1316 /* that's beyond what remains, adjust the number of objects that
1317 * can be safely precreated */
1318 wanted = ted->ted_grant + left;
1319 *nr = wanted / lut->lut_dt_conf.ddp_inodespace;
1321 /* we really have no space any more for precreation,
1322 * fail the precreate request with ENOSPC */
1323 spin_unlock(&tgd->tgd_grant_lock);
1326 /* compute space needed for the new number of creations */
1327 wanted = *nr * lut->lut_dt_conf.ddp_inodespace;
1329 LASSERT(wanted <= ted->ted_grant + left);
1331 if (wanted <= ted->ted_grant) {
1332 /* we've enough grant space to handle this precreate request */
1333 ted->ted_grant -= wanted;
1335 /* we need to take some space from the ungranted pool */
1336 tgd->tgd_tot_granted += wanted - ted->ted_grant;
1337 left -= wanted - ted->ted_grant;
1341 ted->ted_pending += granted;
1342 tgd->tgd_tot_pending += granted;
1344 /* grant more space for precreate purpose if possible. */
1345 wanted = OST_MAX_PRECREATE * lut->lut_dt_conf.ddp_inodespace / 2;
1346 if (wanted > ted->ted_grant) {
1349 /* always try to book enough space to handle a large precreate
1351 chunk = tgt_grant_chunk(exp, lut, NULL);
1352 wanted -= ted->ted_grant;
1353 tgt_grant_alloc(exp, ted->ted_grant, wanted, left, chunk,
1356 spin_unlock(&tgd->tgd_grant_lock);
1359 EXPORT_SYMBOL(tgt_grant_create);
1362 * Release grant space added to the pending counter by tgt_grant_prepare_write()
1364 * Update pending grant counter once buffers have been written to the disk.
1366 * \param[in] exp export of the client which sent the request
1367 * \param[in] pending amount of reserved space to be released
1368 * \param[in] rc return code of pre-commit operations
1370 void tgt_grant_commit(struct obd_export *exp, unsigned long pending,
1373 struct tg_grants_data *tgd = &exp->exp_obd->u.obt.obt_lut->lut_tgd;
1377 /* get space accounted in tot_pending for the I/O, set in
1378 * tgt_grant_check() */
1382 spin_lock(&tgd->tgd_grant_lock);
1383 /* Don't update statfs data for errors raised before commit (e.g.
1384 * bulk transfer failed, ...) since we know those writes have not been
1385 * processed. For other errors hit during commit, we cannot really tell
1386 * whether or not something was written, so we update statfs data.
1387 * In any case, this should not be fatal since we always get fresh
1388 * statfs data before failing a request with ENOSPC */
1390 spin_lock(&tgd->tgd_osfs_lock);
1391 /* Take pending out of cached statfs data */
1392 tgd->tgd_osfs.os_bavail -= min_t(u64,
1393 tgd->tgd_osfs.os_bavail,
1394 pending >> tgd->tgd_blockbits);
1395 if (tgd->tgd_statfs_inflight)
1396 /* someone is running statfs and want to be notified of
1397 * writes happening meanwhile */
1398 tgd->tgd_osfs_inflight += pending;
1399 spin_unlock(&tgd->tgd_osfs_lock);
1402 if (exp->exp_target_data.ted_pending < pending) {
1403 CERROR("%s: cli %s/%p ted_pending(%lu) < grant_used(%lu)\n",
1404 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
1405 exp->exp_target_data.ted_pending, pending);
1406 spin_unlock(&tgd->tgd_grant_lock);
1409 exp->exp_target_data.ted_pending -= pending;
1411 if (tgd->tgd_tot_granted < pending) {
1412 CERROR("%s: cli %s/%p tot_granted(%llu) < grant_used(%lu)\n",
1413 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
1414 tgd->tgd_tot_granted, pending);
1415 spin_unlock(&tgd->tgd_grant_lock);
1418 tgd->tgd_tot_granted -= pending;
1420 if (tgd->tgd_tot_pending < pending) {
1421 CERROR("%s: cli %s/%p tot_pending(%llu) < grant_used(%lu)\n",
1422 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
1423 tgd->tgd_tot_pending, pending);
1424 spin_unlock(&tgd->tgd_grant_lock);
1427 tgd->tgd_tot_pending -= pending;
1428 spin_unlock(&tgd->tgd_grant_lock);
1431 EXPORT_SYMBOL(tgt_grant_commit);
1433 struct tgt_grant_cb {
1434 /* commit callback structure */
1435 struct dt_txn_commit_cb tgc_cb;
1436 /* export associated with the bulk write */
1437 struct obd_export *tgc_exp;
1438 /* pending grant to be released */
1439 unsigned long tgc_granted;
1443 * Callback function for grant releasing
1445 * Release grant space reserved by the client node.
1447 * \param[in] env execution environment
1448 * \param[in] th transaction handle
1449 * \param[in] cb callback data
1450 * \param[in] err error code
1452 static void tgt_grant_commit_cb(struct lu_env *env, struct thandle *th,
1453 struct dt_txn_commit_cb *cb, int err)
1455 struct tgt_grant_cb *tgc;
1457 tgc = container_of(cb, struct tgt_grant_cb, tgc_cb);
1459 tgt_grant_commit(tgc->tgc_exp, tgc->tgc_granted, err);
1460 class_export_cb_put(tgc->tgc_exp);
1465 * Add callback for grant releasing
1467 * Register a commit callback to release grant space.
1469 * \param[in] th transaction handle
1470 * \param[in] exp OBD export of client
1471 * \param[in] granted amount of grant space to be released upon commit
1473 * \retval 0 on successful callback adding
1474 * \retval negative value on error
1476 int tgt_grant_commit_cb_add(struct thandle *th, struct obd_export *exp,
1477 unsigned long granted)
1479 struct tgt_grant_cb *tgc;
1480 struct dt_txn_commit_cb *dcb;
1488 tgc->tgc_exp = class_export_cb_get(exp);
1489 tgc->tgc_granted = granted;
1492 dcb->dcb_func = tgt_grant_commit_cb;
1493 INIT_LIST_HEAD(&dcb->dcb_linkage);
1494 strlcpy(dcb->dcb_name, "tgt_grant_commit_cb", sizeof(dcb->dcb_name));
1496 rc = dt_trans_cb_add(th, dcb);
1498 class_export_cb_put(tgc->tgc_exp);
1504 EXPORT_SYMBOL(tgt_grant_commit_cb_add);