drbd_req.c 32.9 KB
Newer Older
P
Philipp Reisner 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
/*
   drbd_req.c

   This file is part of DRBD by Philipp Reisner and Lars Ellenberg.

   Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
   Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
   Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.

   drbd is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.

   drbd is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with drbd; see the file COPYING.  If not, write to
   the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.

 */

#include <linux/module.h>

#include <linux/slab.h>
#include <linux/drbd.h>
#include "drbd_int.h"
#include "drbd_req.h"


/* Update disk stats at start of I/O request */
static void _drbd_start_io_acct(struct drbd_conf *mdev, struct drbd_request *req, struct bio *bio)
{
	const int rw = bio_data_dir(bio);
	int cpu;
	cpu = part_stat_lock();
	part_stat_inc(cpu, &mdev->vdisk->part0, ios[rw]);
	part_stat_add(cpu, &mdev->vdisk->part0, sectors[rw], bio_sectors(bio));
42
	part_inc_in_flight(&mdev->vdisk->part0, rw);
P
Philipp Reisner 已提交
43 44 45 46 47 48 49 50 51 52 53 54
	part_stat_unlock();
}

/* Update disk stats when completing request upwards */
static void _drbd_end_io_acct(struct drbd_conf *mdev, struct drbd_request *req)
{
	int rw = bio_data_dir(req->master_bio);
	unsigned long duration = jiffies - req->start_time;
	int cpu;
	cpu = part_stat_lock();
	part_stat_add(cpu, &mdev->vdisk->part0, ticks[rw], duration);
	part_round_stats(cpu, &mdev->vdisk->part0);
55
	part_dec_in_flight(&mdev->vdisk->part0, rw);
P
Philipp Reisner 已提交
56 57 58
	part_stat_unlock();
}

59 60 61 62 63 64 65 66 67 68 69
static struct drbd_request *drbd_req_new(struct drbd_conf *mdev,
					       struct bio *bio_src)
{
	struct drbd_request *req;

	req = mempool_alloc(drbd_request_mempool, GFP_NOIO);
	if (!req)
		return NULL;

	drbd_req_make_private_bio(req, bio_src);
	req->rq_state    = bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0;
70
	req->w.mdev      = mdev;
71 72
	req->master_bio  = bio_src;
	req->epoch       = 0;
73

74 75 76
	drbd_clear_interval(&req->i);
	req->i.sector     = bio_src->bi_sector;
	req->i.size      = bio_src->bi_size;
77
	req->i.local = true;
78 79
	req->i.waiting = false;

80 81 82 83 84 85 86 87 88 89 90 91
	INIT_LIST_HEAD(&req->tl_requests);
	INIT_LIST_HEAD(&req->w.list);

	return req;
}

static void drbd_req_free(struct drbd_request *req)
{
	mempool_free(req, drbd_request_mempool);
}

/* rw is bio_data_dir(), only READ or WRITE */
P
Philipp Reisner 已提交
92 93 94
static void _req_is_done(struct drbd_conf *mdev, struct drbd_request *req, const int rw)
{
	const unsigned long s = req->rq_state;
95 96 97 98 99 100 101 102 103

	/* remove it from the transfer log.
	 * well, only if it had been there in the first
	 * place... if it had not (local only or conflicting
	 * and never sent), it should still be "empty" as
	 * initialized in drbd_req_new(), so we can list_del() it
	 * here unconditionally */
	list_del(&req->tl_requests);

P
Philipp Reisner 已提交
104 105 106 107 108 109 110 111 112
	/* if it was a write, we may have to set the corresponding
	 * bit(s) out-of-sync first. If it had a local part, we need to
	 * release the reference to the activity log. */
	if (rw == WRITE) {
		/* Set out-of-sync unless both OK flags are set
		 * (local only or remote failed).
		 * Other places where we set out-of-sync:
		 * READ with local io-error */
		if (!(s & RQ_NET_OK) || !(s & RQ_LOCAL_OK))
113
			drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
P
Philipp Reisner 已提交
114 115

		if ((s & RQ_NET_OK) && (s & RQ_LOCAL_OK) && (s & RQ_NET_SIS))
116
			drbd_set_in_sync(mdev, req->i.sector, req->i.size);
P
Philipp Reisner 已提交
117 118

		/* one might be tempted to move the drbd_al_complete_io
119
		 * to the local io completion callback drbd_request_endio.
P
Philipp Reisner 已提交
120 121 122 123 124 125 126 127 128 129
		 * but, if this was a mirror write, we may only
		 * drbd_al_complete_io after this is RQ_NET_DONE,
		 * otherwise the extent could be dropped from the al
		 * before it has actually been written on the peer.
		 * if we crash before our peer knows about the request,
		 * but after the extent has been dropped from the al,
		 * we would forget to resync the corresponding extent.
		 */
		if (s & RQ_LOCAL_MASK) {
			if (get_ldev_if_state(mdev, D_FAILED)) {
130
				if (s & RQ_IN_ACT_LOG)
131
					drbd_al_complete_io(mdev, &req->i);
P
Philipp Reisner 已提交
132 133
				put_ldev(mdev);
			} else if (__ratelimit(&drbd_ratelimit_state)) {
134 135 136
				dev_warn(DEV, "Should have called drbd_al_complete_io(, %llu, %u), "
					 "but my Disk seems to have failed :(\n",
					 (unsigned long long) req->i.sector, req->i.size);
P
Philipp Reisner 已提交
137 138 139 140
			}
		}
	}

141
	drbd_req_free(req);
P
Philipp Reisner 已提交
142 143 144 145 146 147 148 149 150 151 152 153 154 155
}

static void queue_barrier(struct drbd_conf *mdev)
{
	struct drbd_tl_epoch *b;

	/* We are within the req_lock. Once we queued the barrier for sending,
	 * we set the CREATE_BARRIER bit. It is cleared as soon as a new
	 * barrier/epoch object is added. This is the only place this bit is
	 * set. It indicates that the barrier for this epoch is already queued,
	 * and no new epoch has been created yet. */
	if (test_bit(CREATE_BARRIER, &mdev->flags))
		return;

156
	b = mdev->tconn->newest_tle;
P
Philipp Reisner 已提交
157
	b->w.cb = w_send_barrier;
158
	b->w.mdev = mdev;
P
Philipp Reisner 已提交
159 160 161 162 163
	/* inc_ap_pending done here, so we won't
	 * get imbalanced on connection loss.
	 * dec_ap_pending will be done in got_BarrierAck
	 * or (on connection loss) in tl_clear.  */
	inc_ap_pending(mdev);
164
	drbd_queue_work(&mdev->tconn->data.work, &b->w);
P
Philipp Reisner 已提交
165 166 167 168 169 170 171 172
	set_bit(CREATE_BARRIER, &mdev->flags);
}

static void _about_to_complete_local_write(struct drbd_conf *mdev,
	struct drbd_request *req)
{
	const unsigned long s = req->rq_state;

173 174 175 176 177 178 179 180
	/* Before we can signal completion to the upper layers,
	 * we may need to close the current epoch.
	 * We can skip this, if this request has not even been sent, because we
	 * did not have a fully established connection yet/anymore, during
	 * bitmap exchange, or while we are C_AHEAD due to congestion policy.
	 */
	if (mdev->state.conn >= C_CONNECTED &&
	    (s & RQ_NET_SENT) != 0 &&
181
	    req->epoch == mdev->tconn->newest_tle->br_number)
P
Philipp Reisner 已提交
182 183 184 185 186 187 188 189 190 191
		queue_barrier(mdev);
}

void complete_master_bio(struct drbd_conf *mdev,
		struct bio_and_error *m)
{
	bio_endio(m->bio, m->error);
	dec_ap_bio(mdev);
}

192 193 194 195

static void drbd_remove_request_interval(struct rb_root *root,
					 struct drbd_request *req)
{
196
	struct drbd_conf *mdev = req->w.mdev;
197 198 199 200 201 202 203 204 205
	struct drbd_interval *i = &req->i;

	drbd_remove_interval(root, i);

	/* Wake up any processes waiting for this request to complete.  */
	if (i->waiting)
		wake_up(&mdev->misc_wait);
}

P
Philipp Reisner 已提交
206 207 208 209 210 211 212 213 214
/* Helper for __req_mod().
 * Set m->bio to the master bio, if it is fit to be completed,
 * or leave it alone (it is initialized to NULL in __req_mod),
 * if it has already been completed, or cannot be completed yet.
 * If m->bio is set, the error status to be returned is placed in m->error.
 */
void _req_may_be_done(struct drbd_request *req, struct bio_and_error *m)
{
	const unsigned long s = req->rq_state;
215
	struct drbd_conf *mdev = req->w.mdev;
P
Philipp Reisner 已提交
216 217 218 219 220 221 222 223 224 225 226 227
	/* only WRITES may end up here without a master bio (on barrier ack) */
	int rw = req->master_bio ? bio_data_dir(req->master_bio) : WRITE;

	/* we must not complete the master bio, while it is
	 *	still being processed by _drbd_send_zc_bio (drbd_send_dblock)
	 *	not yet acknowledged by the peer
	 *	not yet completed by the local io subsystem
	 * these flags may get cleared in any order by
	 *	the worker,
	 *	the receiver,
	 *	the bio_endio completion callbacks.
	 */
228 229 230 231 232 233
	if (s & RQ_LOCAL_PENDING)
		return;
	if (req->i.waiting) {
		/* Retry all conflicting peer requests.  */
		wake_up(&mdev->misc_wait);
	}
P
Philipp Reisner 已提交
234 235 236 237 238 239
	if (s & RQ_NET_QUEUED)
		return;
	if (s & RQ_NET_PENDING)
		return;

	if (req->master_bio) {
240
		/* this is DATA_RECEIVED (remote read)
P
Philipp Reisner 已提交
241 242
		 * or protocol C P_WRITE_ACK
		 * or protocol B P_RECV_ACK
243
		 * or protocol A "HANDED_OVER_TO_NETWORK" (SendAck)
P
Philipp Reisner 已提交
244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
		 * or canceled or failed,
		 * or killed from the transfer log due to connection loss.
		 */

		/*
		 * figure out whether to report success or failure.
		 *
		 * report success when at least one of the operations succeeded.
		 * or, to put the other way,
		 * only report failure, when both operations failed.
		 *
		 * what to do about the failures is handled elsewhere.
		 * what we need to do here is just: complete the master_bio.
		 *
		 * local completion error, if any, has been stored as ERR_PTR
259
		 * in private_bio within drbd_request_endio.
P
Philipp Reisner 已提交
260 261 262 263 264 265
		 */
		int ok = (s & RQ_LOCAL_OK) || (s & RQ_NET_OK);
		int error = PTR_ERR(req->private_bio);

		/* remove the request from the conflict detection
		 * respective block_id verification hash */
266 267 268 269 270 271 272
		if (!drbd_interval_empty(&req->i)) {
			struct rb_root *root;

			if (rw == WRITE)
				root = &mdev->write_requests;
			else
				root = &mdev->read_requests;
273
			drbd_remove_request_interval(root, req);
274
		} else if (!(s & RQ_POSTPONED))
P
Philipp Reisner 已提交
275
			D_ASSERT((s & (RQ_NET_MASK & ~RQ_NET_DONE)) == 0);
P
Philipp Reisner 已提交
276 277 278 279 280 281 282 283

		/* for writes we need to do some extra housekeeping */
		if (rw == WRITE)
			_about_to_complete_local_write(mdev, req);

		/* Update disk stats */
		_drbd_end_io_acct(mdev, req);

284 285 286 287
		if (!(s & RQ_POSTPONED)) {
			m->error = ok ? 0 : (error ?: -EIO);
			m->bio = req->master_bio;
		}
P
Philipp Reisner 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301
		req->master_bio = NULL;
	}

	if ((s & RQ_NET_MASK) == 0 || (s & RQ_NET_DONE)) {
		/* this is disconnected (local only) operation,
		 * or protocol C P_WRITE_ACK,
		 * or protocol A or B P_BARRIER_ACK,
		 * or killed from the transfer log due to connection loss. */
		_req_is_done(mdev, req, rw);
	}
	/* else: network part and not DONE yet. that is
	 * protocol A or B, barrier ack still pending... */
}

302 303
static void _req_may_be_done_not_susp(struct drbd_request *req, struct bio_and_error *m)
{
304
	struct drbd_conf *mdev = req->w.mdev;
305

306
	if (!is_susp(mdev->state))
307 308 309
		_req_may_be_done(req, m);
}

P
Philipp Reisner 已提交
310 311 312 313 314 315 316 317 318 319 320 321
/* obviously this could be coded as many single functions
 * instead of one huge switch,
 * or by putting the code directly in the respective locations
 * (as it has been before).
 *
 * but having it this way
 *  enforces that it is all in this one place, where it is easier to audit,
 *  it makes it obvious that whatever "event" "happens" to a request should
 *  happen "atomically" within the req_lock,
 *  and it enforces that we have to think in a very structured manner
 *  about the "events" that may happen to a request during its life time ...
 */
322
int __req_mod(struct drbd_request *req, enum drbd_req_event what,
P
Philipp Reisner 已提交
323 324
		struct bio_and_error *m)
{
325
	struct drbd_conf *mdev = req->w.mdev;
326
	int rv = 0;
327 328 329

	if (m)
		m->bio = NULL;
P
Philipp Reisner 已提交
330 331 332 333 334 335 336 337

	switch (what) {
	default:
		dev_err(DEV, "LOGIC BUG in %s:%u\n", __FILE__ , __LINE__);
		break;

	/* does not happen...
	 * initialization done in drbd_req_new
338
	case CREATED:
P
Philipp Reisner 已提交
339 340 341
		break;
		*/

342
	case TO_BE_SENT: /* via network */
343
		/* reached via __drbd_make_request
P
Philipp Reisner 已提交
344 345 346 347 348 349
		 * and from w_read_retry_remote */
		D_ASSERT(!(req->rq_state & RQ_NET_MASK));
		req->rq_state |= RQ_NET_PENDING;
		inc_ap_pending(mdev);
		break;

350
	case TO_BE_SUBMITTED: /* locally */
351
		/* reached via __drbd_make_request */
P
Philipp Reisner 已提交
352 353 354 355
		D_ASSERT(!(req->rq_state & RQ_LOCAL_MASK));
		req->rq_state |= RQ_LOCAL_PENDING;
		break;

356
	case COMPLETED_OK:
P
Philipp Reisner 已提交
357
		if (bio_data_dir(req->master_bio) == WRITE)
358
			mdev->writ_cnt += req->i.size >> 9;
P
Philipp Reisner 已提交
359
		else
360
			mdev->read_cnt += req->i.size >> 9;
P
Philipp Reisner 已提交
361 362 363 364

		req->rq_state |= (RQ_LOCAL_COMPLETED|RQ_LOCAL_OK);
		req->rq_state &= ~RQ_LOCAL_PENDING;

365
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
366 367 368
		put_ldev(mdev);
		break;

369
	case WRITE_COMPLETED_WITH_ERROR:
P
Philipp Reisner 已提交
370 371 372
		req->rq_state |= RQ_LOCAL_COMPLETED;
		req->rq_state &= ~RQ_LOCAL_PENDING;

373
		__drbd_chk_io_error(mdev, false);
374
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
375 376 377
		put_ldev(mdev);
		break;

378
	case READ_AHEAD_COMPLETED_WITH_ERROR:
P
Philipp Reisner 已提交
379 380 381
		/* it is legal to fail READA */
		req->rq_state |= RQ_LOCAL_COMPLETED;
		req->rq_state &= ~RQ_LOCAL_PENDING;
382
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
383 384 385
		put_ldev(mdev);
		break;

386
	case READ_COMPLETED_WITH_ERROR:
387
		drbd_set_out_of_sync(mdev, req->i.sector, req->i.size);
P
Philipp Reisner 已提交
388 389 390 391 392 393

		req->rq_state |= RQ_LOCAL_COMPLETED;
		req->rq_state &= ~RQ_LOCAL_PENDING;

		D_ASSERT(!(req->rq_state & RQ_NET_MASK));

394
		__drbd_chk_io_error(mdev, false);
P
Philipp Reisner 已提交
395 396
		put_ldev(mdev);

397 398 399
		/* no point in retrying if there is no good remote data,
		 * or we have no connection. */
		if (mdev->state.pdsk != D_UP_TO_DATE) {
400
			_req_may_be_done_not_susp(req, m);
401 402 403
			break;
		}

404
		/* _req_mod(req,TO_BE_SENT); oops, recursion... */
405 406
		req->rq_state |= RQ_NET_PENDING;
		inc_ap_pending(mdev);
407
		/* fall through: _req_mod(req,QUEUE_FOR_NET_READ); */
P
Philipp Reisner 已提交
408

409
	case QUEUE_FOR_NET_READ:
P
Philipp Reisner 已提交
410 411 412 413
		/* READ or READA, and
		 * no local disk,
		 * or target area marked as invalid,
		 * or just got an io-error. */
414
		/* from __drbd_make_request
P
Philipp Reisner 已提交
415 416 417 418
		 * or from bio_endio during read io-error recovery */

		/* so we can verify the handle in the answer packet
		 * corresponding hlist_del is in _req_may_be_done() */
419
		drbd_insert_interval(&mdev->read_requests, &req->i);
P
Philipp Reisner 已提交
420

421
		set_bit(UNPLUG_REMOTE, &mdev->flags);
P
Philipp Reisner 已提交
422 423 424 425 426 427

		D_ASSERT(req->rq_state & RQ_NET_PENDING);
		req->rq_state |= RQ_NET_QUEUED;
		req->w.cb = (req->rq_state & RQ_LOCAL_MASK)
			? w_read_retry_remote
			: w_send_read_req;
428
		drbd_queue_work(&mdev->tconn->data.work, &req->w);
P
Philipp Reisner 已提交
429 430
		break;

431
	case QUEUE_FOR_NET_WRITE:
P
Philipp Reisner 已提交
432
		/* assert something? */
433
		/* from __drbd_make_request only */
P
Philipp Reisner 已提交
434 435

		/* corresponding hlist_del is in _req_may_be_done() */
436
		drbd_insert_interval(&mdev->write_requests, &req->i);
P
Philipp Reisner 已提交
437 438 439 440 441 442 443 444 445 446

		/* NOTE
		 * In case the req ended up on the transfer log before being
		 * queued on the worker, it could lead to this request being
		 * missed during cleanup after connection loss.
		 * So we have to do both operations here,
		 * within the same lock that protects the transfer log.
		 *
		 * _req_add_to_epoch(req); this has to be after the
		 * _maybe_start_new_epoch(req); which happened in
447
		 * __drbd_make_request, because we now may set the bit
P
Philipp Reisner 已提交
448 449 450 451
		 * again ourselves to close the current epoch.
		 *
		 * Add req to the (now) current epoch (barrier). */

452 453 454 455 456
		/* otherwise we may lose an unplug, which may cause some remote
		 * io-scheduler timeout to expire, increasing maximum latency,
		 * hurting performance. */
		set_bit(UNPLUG_REMOTE, &mdev->flags);

457
		/* see __drbd_make_request,
P
Philipp Reisner 已提交
458 459 460
		 * just after it grabs the req_lock */
		D_ASSERT(test_bit(CREATE_BARRIER, &mdev->flags) == 0);

461
		req->epoch = mdev->tconn->newest_tle->br_number;
P
Philipp Reisner 已提交
462 463

		/* increment size of current epoch */
464
		mdev->tconn->newest_tle->n_writes++;
P
Philipp Reisner 已提交
465 466 467 468 469

		/* queue work item to send data */
		D_ASSERT(req->rq_state & RQ_NET_PENDING);
		req->rq_state |= RQ_NET_QUEUED;
		req->w.cb =  w_send_dblock;
470
		drbd_queue_work(&mdev->tconn->data.work, &req->w);
P
Philipp Reisner 已提交
471 472

		/* close the epoch, in case it outgrew the limit */
473
		if (mdev->tconn->newest_tle->n_writes >= mdev->tconn->net_conf->max_epoch_size)
P
Philipp Reisner 已提交
474 475 476 477
			queue_barrier(mdev);

		break;

478
	case QUEUE_FOR_SEND_OOS:
479
		req->rq_state |= RQ_NET_QUEUED;
480
		req->w.cb =  w_send_out_of_sync;
481
		drbd_queue_work(&mdev->tconn->data.work, &req->w);
482 483
		break;

484
	case OOS_HANDED_TO_NETWORK:
485
		/* actually the same */
486
	case SEND_CANCELED:
P
Philipp Reisner 已提交
487
		/* treat it the same */
488
	case SEND_FAILED:
P
Philipp Reisner 已提交
489 490 491 492 493
		/* real cleanup will be done from tl_clear.  just update flags
		 * so it is no longer marked as on the worker queue */
		req->rq_state &= ~RQ_NET_QUEUED;
		/* if we did it right, tl_clear should be scheduled only after
		 * this, so this should not be necessary! */
494
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
495 496
		break;

497
	case HANDED_OVER_TO_NETWORK:
P
Philipp Reisner 已提交
498
		/* assert something? */
499
		if (bio_data_dir(req->master_bio) == WRITE)
500
			atomic_add(req->i.size >> 9, &mdev->ap_in_flight);
501

P
Philipp Reisner 已提交
502
		if (bio_data_dir(req->master_bio) == WRITE &&
503
		    mdev->tconn->net_conf->wire_protocol == DRBD_PROT_A) {
P
Philipp Reisner 已提交
504 505 506 507 508 509 510 511 512 513 514 515 516 517
			/* this is what is dangerous about protocol A:
			 * pretend it was successfully written on the peer. */
			if (req->rq_state & RQ_NET_PENDING) {
				dec_ap_pending(mdev);
				req->rq_state &= ~RQ_NET_PENDING;
				req->rq_state |= RQ_NET_OK;
			} /* else: neg-ack was faster... */
			/* it is still not yet RQ_NET_DONE until the
			 * corresponding epoch barrier got acked as well,
			 * so we know what to dirty on connection loss */
		}
		req->rq_state &= ~RQ_NET_QUEUED;
		req->rq_state |= RQ_NET_SENT;
		/* because _drbd_send_zc_bio could sleep, and may want to
518 519
		 * dereference the bio even after the "WRITE_ACKED_BY_PEER" and
		 * "COMPLETED_OK" events came in, once we return from
P
Philipp Reisner 已提交
520 521
		 * _drbd_send_zc_bio (drbd_send_dblock), we have to check
		 * whether it is done already, and end it.  */
522
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
523 524
		break;

525
	case READ_RETRY_REMOTE_CANCELED:
526 527
		req->rq_state &= ~RQ_NET_QUEUED;
		/* fall through, in case we raced with drbd_disconnect */
528
	case CONNECTION_LOST_WHILE_PENDING:
P
Philipp Reisner 已提交
529 530 531 532 533 534
		/* transfer log cleanup after connection loss */
		/* assert something? */
		if (req->rq_state & RQ_NET_PENDING)
			dec_ap_pending(mdev);
		req->rq_state &= ~(RQ_NET_OK|RQ_NET_PENDING);
		req->rq_state |= RQ_NET_DONE;
535
		if (req->rq_state & RQ_NET_SENT && req->rq_state & RQ_WRITE)
536
			atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
537

P
Philipp Reisner 已提交
538 539 540
		/* if it is still queued, we may not complete it here.
		 * it will be canceled soon. */
		if (!(req->rq_state & RQ_NET_QUEUED))
541
			_req_may_be_done(req, m); /* Allowed while state.susp */
P
Philipp Reisner 已提交
542 543
		break;

544
	case WRITE_ACKED_BY_PEER_AND_SIS:
P
Philipp Reisner 已提交
545
		req->rq_state |= RQ_NET_SIS;
546
	case DISCARD_WRITE:
P
Philipp Reisner 已提交
547 548 549 550 551
		/* for discarded conflicting writes of multiple primaries,
		 * there is no need to keep anything in the tl, potential
		 * node crashes are covered by the activity log. */
		req->rq_state |= RQ_NET_DONE;
		/* fall through */
552
	case WRITE_ACKED_BY_PEER:
P
Philipp Reisner 已提交
553 554 555 556 557 558 559 560 561 562 563
		/* protocol C; successfully written on peer.
		 * Nothing to do here.
		 * We want to keep the tl in place for all protocols, to cater
		 * for volatile write-back caches on lower level devices.
		 *
		 * A barrier request is expected to have forced all prior
		 * requests onto stable storage, so completion of a barrier
		 * request could set NET_DONE right here, and not wait for the
		 * P_BARRIER_ACK, but that is an unnecessary optimization. */

		/* this makes it effectively the same as for: */
564
	case RECV_ACKED_BY_PEER:
P
Philipp Reisner 已提交
565
		/* protocol B; pretends to be successfully written on peer.
566
		 * see also notes above in HANDED_OVER_TO_NETWORK about
P
Philipp Reisner 已提交
567 568 569 570
		 * protocol != C */
		req->rq_state |= RQ_NET_OK;
		D_ASSERT(req->rq_state & RQ_NET_PENDING);
		dec_ap_pending(mdev);
571
		atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
P
Philipp Reisner 已提交
572
		req->rq_state &= ~RQ_NET_PENDING;
573
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
574 575
		break;

576 577 578 579 580 581 582 583 584 585 586
	case POSTPONE_WRITE:
		/*
		 * If this node has already detected the write conflict, the
		 * worker will be waiting on misc_wait.  Wake it up once this
		 * request has completed locally.
		 */
		D_ASSERT(req->rq_state & RQ_NET_PENDING);
		req->rq_state |= RQ_POSTPONED;
		_req_may_be_done_not_susp(req, m);
		break;

587
	case NEG_ACKED:
P
Philipp Reisner 已提交
588
		/* assert something? */
589
		if (req->rq_state & RQ_NET_PENDING) {
P
Philipp Reisner 已提交
590
			dec_ap_pending(mdev);
591
			atomic_sub(req->i.size >> 9, &mdev->ap_in_flight);
592
		}
P
Philipp Reisner 已提交
593 594 595
		req->rq_state &= ~(RQ_NET_OK|RQ_NET_PENDING);

		req->rq_state |= RQ_NET_DONE;
596
		_req_may_be_done_not_susp(req, m);
597
		/* else: done by HANDED_OVER_TO_NETWORK */
P
Philipp Reisner 已提交
598 599
		break;

600
	case FAIL_FROZEN_DISK_IO:
601 602 603
		if (!(req->rq_state & RQ_LOCAL_COMPLETED))
			break;

604
		_req_may_be_done(req, m); /* Allowed while state.susp */
605 606
		break;

607
	case RESTART_FROZEN_DISK_IO:
608 609 610 611 612 613 614 615 616 617 618
		if (!(req->rq_state & RQ_LOCAL_COMPLETED))
			break;

		req->rq_state &= ~RQ_LOCAL_COMPLETED;

		rv = MR_READ;
		if (bio_data_dir(req->master_bio) == WRITE)
			rv = MR_WRITE;

		get_ldev(mdev);
		req->w.cb = w_restart_disk_io;
619
		drbd_queue_work(&mdev->tconn->data.work, &req->w);
620 621
		break;

622
	case RESEND:
623
		/* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK
624
		   before the connection loss (B&C only); only P_BARRIER_ACK was missing.
625
		   Trowing them out of the TL here by pretending we got a BARRIER_ACK
626
		   We ensure that the peer was not rebooted */
627 628
		if (!(req->rq_state & RQ_NET_OK)) {
			if (req->w.cb) {
629
				drbd_queue_work(&mdev->tconn->data.work, &req->w);
630 631 632 633
				rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ;
			}
			break;
		}
634
		/* else, fall through to BARRIER_ACKED */
635

636
	case BARRIER_ACKED:
637 638 639
		if (!(req->rq_state & RQ_WRITE))
			break;

P
Philipp Reisner 已提交
640 641 642 643
		if (req->rq_state & RQ_NET_PENDING) {
			/* barrier came in before all requests have been acked.
			 * this is bad, because if the connection is lost now,
			 * we won't be able to clean them up... */
644
			dev_err(DEV, "FIXME (BARRIER_ACKED but pending)\n");
645
			list_move(&req->tl_requests, &mdev->tconn->out_of_sequence_requests);
P
Philipp Reisner 已提交
646
		}
647 648
		if ((req->rq_state & RQ_NET_MASK) != 0) {
			req->rq_state |= RQ_NET_DONE;
649 650
			if (mdev->tconn->net_conf->wire_protocol == DRBD_PROT_A)
				atomic_sub(req->i.size>>9, &mdev->ap_in_flight);
651
		}
652
		_req_may_be_done(req, m); /* Allowed while state.susp */
P
Philipp Reisner 已提交
653 654
		break;

655
	case DATA_RECEIVED:
P
Philipp Reisner 已提交
656 657 658 659
		D_ASSERT(req->rq_state & RQ_NET_PENDING);
		dec_ap_pending(mdev);
		req->rq_state &= ~RQ_NET_PENDING;
		req->rq_state |= (RQ_NET_OK|RQ_NET_DONE);
660
		_req_may_be_done_not_susp(req, m);
P
Philipp Reisner 已提交
661 662
		break;
	};
663 664

	return rv;
P
Philipp Reisner 已提交
665 666 667 668 669 670 671 672 673
}

/* we may do a local read if:
 * - we are consistent (of course),
 * - or we are generally inconsistent,
 *   BUT we are still/already IN SYNC for this area.
 *   since size may be bigger than BM_BLOCK_SIZE,
 *   we may need to check several bits.
 */
674
static bool drbd_may_do_local_read(struct drbd_conf *mdev, sector_t sector, int size)
P
Philipp Reisner 已提交
675 676 677 678 679
{
	unsigned long sbnr, ebnr;
	sector_t esector, nr_sectors;

	if (mdev->state.disk == D_UP_TO_DATE)
680
		return true;
681
	if (mdev->state.disk != D_INCONSISTENT)
682
		return false;
P
Philipp Reisner 已提交
683
	esector = sector + (size >> 9) - 1;
684
	nr_sectors = drbd_get_capacity(mdev->this_bdev);
P
Philipp Reisner 已提交
685 686 687 688 689 690
	D_ASSERT(sector  < nr_sectors);
	D_ASSERT(esector < nr_sectors);

	sbnr = BM_SECT_TO_BIT(sector);
	ebnr = BM_SECT_TO_BIT(esector);

691
	return drbd_bm_count_bits(mdev, sbnr, ebnr) == 0;
P
Philipp Reisner 已提交
692 693
}

694 695 696 697 698 699 700 701 702 703 704 705
/*
 * complete_conflicting_writes  -  wait for any conflicting write requests
 *
 * The write_requests tree contains all active write requests which we
 * currently know about.  Wait for any requests to complete which conflict with
 * the new one.
 */
static int complete_conflicting_writes(struct drbd_conf *mdev,
				       sector_t sector, int size)
{
	for(;;) {
		struct drbd_interval *i;
706
		int err;
707 708 709 710

		i = drbd_find_overlap(&mdev->write_requests, sector, size);
		if (!i)
			return 0;
711 712 713
		err = drbd_wait_misc(mdev, i);
		if (err)
			return err;
714 715 716
	}
}

717
int __drbd_make_request(struct drbd_conf *mdev, struct bio *bio, unsigned long start_time)
P
Philipp Reisner 已提交
718 719 720 721 722 723
{
	const int rw = bio_rw(bio);
	const int size = bio->bi_size;
	const sector_t sector = bio->bi_sector;
	struct drbd_tl_epoch *b = NULL;
	struct drbd_request *req;
724
	int local, remote, send_oos = 0;
725
	int err;
726
	int ret = 0;
P
Philipp Reisner 已提交
727 728 729 730 731 732 733 734 735 736 737

	/* allocate outside of all locks; */
	req = drbd_req_new(mdev, bio);
	if (!req) {
		dec_ap_bio(mdev);
		/* only pass the error to the upper layers.
		 * if user cannot handle io errors, that's not our business. */
		dev_err(DEV, "could not kmalloc() req\n");
		bio_endio(bio, -ENOMEM);
		return 0;
	}
738
	req->start_time = start_time;
P
Philipp Reisner 已提交
739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782

	local = get_ldev(mdev);
	if (!local) {
		bio_put(req->private_bio); /* or we get a bio leak */
		req->private_bio = NULL;
	}
	if (rw == WRITE) {
		remote = 1;
	} else {
		/* READ || READA */
		if (local) {
			if (!drbd_may_do_local_read(mdev, sector, size)) {
				/* we could kick the syncer to
				 * sync this extent asap, wait for
				 * it, then continue locally.
				 * Or just issue the request remotely.
				 */
				local = 0;
				bio_put(req->private_bio);
				req->private_bio = NULL;
				put_ldev(mdev);
			}
		}
		remote = !local && mdev->state.pdsk >= D_UP_TO_DATE;
	}

	/* If we have a disk, but a READA request is mapped to remote,
	 * we are R_PRIMARY, D_INCONSISTENT, SyncTarget.
	 * Just fail that READA request right here.
	 *
	 * THINK: maybe fail all READA when not local?
	 *        or make this configurable...
	 *        if network is slow, READA won't do any good.
	 */
	if (rw == READA && mdev->state.disk >= D_INCONSISTENT && !local) {
		err = -EWOULDBLOCK;
		goto fail_and_free_req;
	}

	/* For WRITES going to the local disk, grab a reference on the target
	 * extent.  This waits for any resync activity in the corresponding
	 * resync extent to finish, and, if necessary, pulls in the target
	 * extent into the activity log, which involves further disk io because
	 * of transactional on-disk meta data updates. */
783 784
	if (rw == WRITE && local && !test_bit(AL_SUSPENDED, &mdev->flags)) {
		req->rq_state |= RQ_IN_ACT_LOG;
785
		drbd_al_begin_io(mdev, &req->i);
786
	}
P
Philipp Reisner 已提交
787

788
	remote = remote && drbd_should_do_remote(mdev->state);
789
	send_oos = rw == WRITE && drbd_should_send_out_of_sync(mdev->state);
790
	D_ASSERT(!(remote && send_oos));
P
Philipp Reisner 已提交
791

792
	if (!(local || remote) && !is_susp(mdev->state)) {
L
Lars Ellenberg 已提交
793 794
		if (__ratelimit(&drbd_ratelimit_state))
			dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
795
		err = -EIO;
P
Philipp Reisner 已提交
796 797 798 799 800 801 802 803 804
		goto fail_free_complete;
	}

	/* For WRITE request, we have to make sure that we have an
	 * unused_spare_tle, in case we need to start a new epoch.
	 * I try to be smart and avoid to pre-allocate always "just in case",
	 * but there is a race between testing the bit and pointer outside the
	 * spinlock, and grabbing the spinlock.
	 * if we lost that race, we retry.  */
805
	if (rw == WRITE && (remote || send_oos) &&
806
	    mdev->tconn->unused_spare_tle == NULL &&
P
Philipp Reisner 已提交
807 808 809 810 811 812 813 814 815 816 817
	    test_bit(CREATE_BARRIER, &mdev->flags)) {
allocate_barrier:
		b = kmalloc(sizeof(struct drbd_tl_epoch), GFP_NOIO);
		if (!b) {
			dev_err(DEV, "Failed to alloc barrier.\n");
			err = -ENOMEM;
			goto fail_free_complete;
		}
	}

	/* GOOD, everything prepared, grab the spin_lock */
818
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
819

820 821 822
	if (rw == WRITE) {
		err = complete_conflicting_writes(mdev, sector, size);
		if (err) {
823 824 825 826
			if (err != -ERESTARTSYS)
				_conn_request_state(mdev->tconn,
						    NS(conn, C_TIMEOUT),
						    CS_HARD);
827
			spin_unlock_irq(&mdev->tconn->req_lock);
828
			err = -EIO;
829 830 831 832
			goto fail_free_complete;
		}
	}

833
	if (is_susp(mdev->state)) {
834 835
		/* If we got suspended, use the retry mechanism of
		   generic_make_request() to restart processing of this
836
		   bio. In the next call to drbd_make_request
837 838
		   we sleep in inc_ap_bio() */
		ret = 1;
839
		spin_unlock_irq(&mdev->tconn->req_lock);
840 841 842
		goto fail_free_complete;
	}

843
	if (remote || send_oos) {
844
		remote = drbd_should_do_remote(mdev->state);
845
		send_oos = rw == WRITE && drbd_should_send_out_of_sync(mdev->state);
846
		D_ASSERT(!(remote && send_oos));
847 848

		if (!(remote || send_oos))
P
Philipp Reisner 已提交
849 850 851
			dev_warn(DEV, "lost connection while grabbing the req_lock!\n");
		if (!(local || remote)) {
			dev_err(DEV, "IO ERROR: neither local nor remote disk\n");
852
			spin_unlock_irq(&mdev->tconn->req_lock);
853
			err = -EIO;
P
Philipp Reisner 已提交
854 855 856 857
			goto fail_free_complete;
		}
	}

858 859
	if (b && mdev->tconn->unused_spare_tle == NULL) {
		mdev->tconn->unused_spare_tle = b;
P
Philipp Reisner 已提交
860 861
		b = NULL;
	}
862
	if (rw == WRITE && (remote || send_oos) &&
863
	    mdev->tconn->unused_spare_tle == NULL &&
P
Philipp Reisner 已提交
864 865 866
	    test_bit(CREATE_BARRIER, &mdev->flags)) {
		/* someone closed the current epoch
		 * while we were grabbing the spinlock */
867
		spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
		goto allocate_barrier;
	}


	/* Update disk stats */
	_drbd_start_io_acct(mdev, req, bio);

	/* _maybe_start_new_epoch(mdev);
	 * If we need to generate a write barrier packet, we have to add the
	 * new epoch (barrier) object, and queue the barrier packet for sending,
	 * and queue the req's data after it _within the same lock_, otherwise
	 * we have race conditions were the reorder domains could be mixed up.
	 *
	 * Even read requests may start a new epoch and queue the corresponding
	 * barrier packet.  To get the write ordering right, we only have to
	 * make sure that, if this is a write request and it triggered a
	 * barrier packet, this request is queued within the same spinlock. */
885
	if ((remote || send_oos) && mdev->tconn->unused_spare_tle &&
P
Philipp Reisner 已提交
886
	    test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) {
887
		_tl_add_barrier(mdev->tconn, mdev->tconn->unused_spare_tle);
888
		mdev->tconn->unused_spare_tle = NULL;
P
Philipp Reisner 已提交
889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909
	} else {
		D_ASSERT(!(remote && rw == WRITE &&
			   test_bit(CREATE_BARRIER, &mdev->flags)));
	}

	/* NOTE
	 * Actually, 'local' may be wrong here already, since we may have failed
	 * to write to the meta data, and may become wrong anytime because of
	 * local io-error for some other request, which would lead to us
	 * "detaching" the local disk.
	 *
	 * 'remote' may become wrong any time because the network could fail.
	 *
	 * This is a harmless race condition, though, since it is handled
	 * correctly at the appropriate places; so it just defers the failure
	 * of the respective operation.
	 */

	/* mark them early for readability.
	 * this just sets some state flags. */
	if (remote)
910
		_req_mod(req, TO_BE_SENT);
P
Philipp Reisner 已提交
911
	if (local)
912
		_req_mod(req, TO_BE_SUBMITTED);
P
Philipp Reisner 已提交
913

914
	list_add_tail(&req->tl_requests, &mdev->tconn->newest_tle->requests);
915

P
Philipp Reisner 已提交
916 917 918 919 920 921 922 923
	/* NOTE remote first: to get the concurrent write detection right,
	 * we must register the request before start of local IO.  */
	if (remote) {
		/* either WRITE and C_CONNECTED,
		 * or READ, and no local disk,
		 * or READ, but not in sync.
		 */
		_req_mod(req, (rw == WRITE)
924 925
				? QUEUE_FOR_NET_WRITE
				: QUEUE_FOR_NET_READ);
P
Philipp Reisner 已提交
926
	}
927
	if (send_oos && drbd_set_out_of_sync(mdev, sector, size))
928
		_req_mod(req, QUEUE_FOR_SEND_OOS);
929

930
	if (remote &&
931
	    mdev->tconn->net_conf->on_congestion != OC_BLOCK && mdev->tconn->agreed_pro_version >= 96) {
932 933
		int congested = 0;

934 935
		if (mdev->tconn->net_conf->cong_fill &&
		    atomic_read(&mdev->ap_in_flight) >= mdev->tconn->net_conf->cong_fill) {
936 937 938 939
			dev_info(DEV, "Congestion-fill threshold reached\n");
			congested = 1;
		}

940
		if (mdev->act_log->used >= mdev->tconn->net_conf->cong_extents) {
941 942 943 944
			dev_info(DEV, "Congestion-extents threshold reached\n");
			congested = 1;
		}

945
		if (congested) {
946
			queue_barrier(mdev); /* last barrier, after mirrored writes */
947

948
			if (mdev->tconn->net_conf->on_congestion == OC_PULL_AHEAD)
949
				_drbd_set_state(_NS(mdev, conn, C_AHEAD), 0, NULL);
950
			else  /*mdev->tconn->net_conf->on_congestion == OC_DISCONNECT */
951 952 953 954
				_drbd_set_state(_NS(mdev, conn, C_DISCONNECTING), 0, NULL);
		}
	}

955
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
956 957 958 959 960
	kfree(b); /* if someone else has beaten us to it... */

	if (local) {
		req->private_bio->bi_bdev = mdev->ldev->backing_bdev;

961 962 963 964 965 966
		/* State may have changed since we grabbed our reference on the
		 * mdev->ldev member. Double check, and short-circuit to endio.
		 * In case the last activity log transaction failed to get on
		 * stable storage, and this is a WRITE, we may not even submit
		 * this bio. */
		if (get_ldev(mdev)) {
967 968 969
			if (drbd_insert_fault(mdev,   rw == WRITE ? DRBD_FAULT_DT_WR
						    : rw == READ  ? DRBD_FAULT_DT_RD
						    :               DRBD_FAULT_DT_RA))
970 971 972 973 974
				bio_endio(req->private_bio, -EIO);
			else
				generic_make_request(req->private_bio);
			put_ldev(mdev);
		} else
P
Philipp Reisner 已提交
975 976 977 978 979 980
			bio_endio(req->private_bio, -EIO);
	}

	return 0;

fail_free_complete:
981
	if (req->rq_state & RQ_IN_ACT_LOG)
982
		drbd_al_complete_io(mdev, &req->i);
P
Philipp Reisner 已提交
983 984 985 986 987 988
fail_and_free_req:
	if (local) {
		bio_put(req->private_bio);
		req->private_bio = NULL;
		put_ldev(mdev);
	}
989 990 991
	if (!ret)
		bio_endio(bio, err);

P
Philipp Reisner 已提交
992 993 994 995
	drbd_req_free(req);
	dec_ap_bio(mdev);
	kfree(b);

996
	return ret;
P
Philipp Reisner 已提交
997 998
}

999
int drbd_make_request(struct request_queue *q, struct bio *bio)
P
Philipp Reisner 已提交
1000 1001
{
	struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
1002
	unsigned long start_time;
P
Philipp Reisner 已提交
1003

1004 1005
	start_time = jiffies;

P
Philipp Reisner 已提交
1006 1007 1008 1009
	/*
	 * what we "blindly" assume:
	 */
	D_ASSERT(bio->bi_size > 0);
1010
	D_ASSERT(IS_ALIGNED(bio->bi_size, 512));
P
Philipp Reisner 已提交
1011

1012 1013
	inc_ap_bio(mdev);
	return __drbd_make_request(mdev, bio, start_time);
P
Philipp Reisner 已提交
1014 1015
}

1016 1017 1018
/* This is called by bio_add_page().
 *
 * q->max_hw_sectors and other global limits are already enforced there.
P
Philipp Reisner 已提交
1019
 *
1020 1021 1022 1023
 * We need to call down to our lower level device,
 * in case it has special restrictions.
 *
 * We also may need to enforce configured max-bio-bvecs limits.
P
Philipp Reisner 已提交
1024 1025
 *
 * As long as the BIO is empty we have to allow at least one bvec,
1026
 * regardless of size and offset, so no need to ask lower levels.
P
Philipp Reisner 已提交
1027 1028 1029 1030 1031
 */
int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec)
{
	struct drbd_conf *mdev = (struct drbd_conf *) q->queuedata;
	unsigned int bio_size = bvm->bi_size;
1032 1033 1034 1035
	int limit = DRBD_MAX_BIO_SIZE;
	int backing_limit;

	if (bio_size && get_ldev(mdev)) {
P
Philipp Reisner 已提交
1036 1037
		struct request_queue * const b =
			mdev->ldev->backing_bdev->bd_disk->queue;
1038
		if (b->merge_bvec_fn) {
P
Philipp Reisner 已提交
1039 1040 1041 1042 1043 1044 1045
			backing_limit = b->merge_bvec_fn(b, bvm, bvec);
			limit = min(limit, backing_limit);
		}
		put_ldev(mdev);
	}
	return limit;
}
1046 1047 1048 1049 1050 1051 1052 1053

void request_timer_fn(unsigned long data)
{
	struct drbd_conf *mdev = (struct drbd_conf *) data;
	struct drbd_request *req; /* oldest request */
	struct list_head *le;
	unsigned long et = 0; /* effective timeout = ko_count * timeout */

1054
	if (get_net_conf(mdev->tconn)) {
1055
		et = mdev->tconn->net_conf->timeout*HZ/10 * mdev->tconn->net_conf->ko_count;
1056
		put_net_conf(mdev->tconn);
1057 1058 1059 1060
	}
	if (!et || mdev->state.conn < C_WF_REPORT_PARAMS)
		return; /* Recurring timer stopped */

1061 1062
	spin_lock_irq(&mdev->tconn->req_lock);
	le = &mdev->tconn->oldest_tle->requests;
1063
	if (list_empty(le)) {
1064
		spin_unlock_irq(&mdev->tconn->req_lock);
1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
		mod_timer(&mdev->request_timer, jiffies + et);
		return;
	}

	le = le->prev;
	req = list_entry(le, struct drbd_request, tl_requests);
	if (time_is_before_eq_jiffies(req->start_time + et)) {
		if (req->rq_state & RQ_NET_PENDING) {
			dev_warn(DEV, "Remote failed to finish a request within ko-count * timeout\n");
			_drbd_set_state(_NS(mdev, conn, C_TIMEOUT), CS_VERBOSE, NULL);
		} else {
			dev_warn(DEV, "Local backing block device frozen?\n");
			mod_timer(&mdev->request_timer, jiffies + et);
		}
	} else {
		mod_timer(&mdev->request_timer, req->start_time + et);
	}

1083
	spin_unlock_irq(&mdev->tconn->req_lock);
1084
}