summaryrefslogtreecommitdiff
path: root/drivers/block/drbd/drbd_req.c
blob: c00ae8619519eb948523a7f4a51f9c6e72b3b399 (plain)
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
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
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
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
// SPDX-License-Identifier: GPL-2.0-or-later
/*
   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>.


 */

#include <linux/module.h>

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


static bool drbd_may_do_local_read(struct drbd_device *device, sector_t sector, int size);

static struct drbd_request *drbd_req_new(struct drbd_device *device, struct bio *bio_src)
{
	struct drbd_request *req;

	req = mempool_alloc(&drbd_request_mempool, GFP_NOIO);
	if (!req)
		return NULL;
	memset(req, 0, sizeof(*req));

	req->private_bio = bio_alloc_clone(device->ldev->backing_bdev, bio_src,
					   GFP_NOIO, &drbd_io_bio_set);
	req->private_bio->bi_private = req;
	req->private_bio->bi_end_io = drbd_request_endio;

	req->rq_state = (bio_data_dir(bio_src) == WRITE ? RQ_WRITE : 0)
		      | (bio_op(bio_src) == REQ_OP_WRITE_SAME ? RQ_WSAME : 0)
		      | (bio_op(bio_src) == REQ_OP_WRITE_ZEROES ? RQ_ZEROES : 0)
		      | (bio_op(bio_src) == REQ_OP_DISCARD ? RQ_UNMAP : 0);
	req->device = device;
	req->master_bio = bio_src;
	req->epoch = 0;

	drbd_clear_interval(&req->i);
	req->i.sector     = bio_src->bi_iter.bi_sector;
	req->i.size      = bio_src->bi_iter.bi_size;
	req->i.local = true;
	req->i.waiting = false;

	INIT_LIST_HEAD(&req->tl_requests);
	INIT_LIST_HEAD(&req->w.list);
	INIT_LIST_HEAD(&req->req_pending_master_completion);
	INIT_LIST_HEAD(&req->req_pending_local);

	/* one reference to be put by __drbd_make_request */
	atomic_set(&req->completion_ref, 1);
	/* one kref as long as completion_ref > 0 */
	kref_init(&req->kref);
	return req;
}

static void drbd_remove_request_interval(struct rb_root *root,
					 struct drbd_request *req)
{
	struct drbd_device *device = req->device;
	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(&device->misc_wait);
}

void drbd_req_destroy(struct kref *kref)
{
	struct drbd_request *req = container_of(kref, struct drbd_request, kref);
	struct drbd_device *device = req->device;
	const unsigned s = req->rq_state;

	if ((req->master_bio && !(s & RQ_POSTPONED)) ||
		atomic_read(&req->completion_ref) ||
		(s & RQ_LOCAL_PENDING) ||
		((s & RQ_NET_MASK) && !(s & RQ_NET_DONE))) {
		drbd_err(device, "drbd_req_destroy: Logic BUG rq_state = 0x%x, completion_ref = %d\n",
				s, atomic_read(&req->completion_ref));
		return;
	}

	/* If called from mod_rq_state (expected normal case) or
	 * drbd_send_and_submit (the less likely normal path), this holds the
	 * req_lock, and req->tl_requests will typicaly be on ->transfer_log,
	 * though it may be still empty (never added to the transfer log).
	 *
	 * If called from do_retry(), we do NOT hold the req_lock, but we are
	 * still allowed to unconditionally list_del(&req->tl_requests),
	 * because it will be on a local on-stack list only. */
	list_del_init(&req->tl_requests);

	/* finally remove the request from the conflict detection
	 * respective block_id verification interval tree. */
	if (!drbd_interval_empty(&req->i)) {
		struct rb_root *root;

		if (s & RQ_WRITE)
			root = &device->write_requests;
		else
			root = &device->read_requests;
		drbd_remove_request_interval(root, req);
	} else if (s & (RQ_NET_MASK & ~RQ_NET_DONE) && req->i.size != 0)
		drbd_err(device, "drbd_req_destroy: Logic BUG: interval empty, but: rq_state=0x%x, sect=%llu, size=%u\n",
			s, (unsigned long long)req->i.sector, req->i.size);

	/* 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 (s & RQ_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 */

		/* There is a special case:
		 * we may notice late that IO was suspended,
		 * and postpone, or schedule for retry, a write,
		 * before it even was submitted or sent.
		 * In that case we do not want to touch the bitmap at all.
		 */
		if ((s & (RQ_POSTPONED|RQ_LOCAL_MASK|RQ_NET_MASK)) != RQ_POSTPONED) {
			if (!(s & RQ_NET_OK) || !(s & RQ_LOCAL_OK))
				drbd_set_out_of_sync(device, req->i.sector, req->i.size);

			if ((s & RQ_NET_OK) && (s & RQ_LOCAL_OK) && (s & RQ_NET_SIS))
				drbd_set_in_sync(device, req->i.sector, req->i.size);
		}

		/* one might be tempted to move the drbd_al_complete_io
		 * to the local io completion callback drbd_request_endio.
		 * 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_IN_ACT_LOG) {
			if (get_ldev_if_state(device, D_FAILED)) {
				drbd_al_complete_io(device, &req->i);
				put_ldev(device);
			} else if (__ratelimit(&drbd_ratelimit_state)) {
				drbd_warn(device, "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);
			}
		}
	}

	mempool_free(req, &drbd_request_mempool);
}

static void wake_all_senders(struct drbd_connection *connection)
{
	wake_up(&connection->sender_work.q_wait);
}

/* must hold resource->req_lock */
void start_new_tl_epoch(struct drbd_connection *connection)
{
	/* no point closing an epoch, if it is empty, anyways. */
	if (connection->current_tle_writes == 0)
		return;

	connection->current_tle_writes = 0;
	atomic_inc(&connection->current_tle_nr);
	wake_all_senders(connection);
}

void complete_master_bio(struct drbd_device *device,
		struct bio_and_error *m)
{
	m->bio->bi_status = errno_to_blk_status(m->error);
	bio_endio(m->bio);
	dec_ap_bio(device);
}


/* 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.
 */
static
void drbd_req_complete(struct drbd_request *req, struct bio_and_error *m)
{
	const unsigned s = req->rq_state;
	struct drbd_device *device = req->device;
	int error, ok;

	/* 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.
	 */
	if ((s & RQ_LOCAL_PENDING && !(s & RQ_LOCAL_ABORTED)) ||
	    (s & RQ_NET_QUEUED) || (s & RQ_NET_PENDING) ||
	    (s & RQ_COMPLETION_SUSP)) {
		drbd_err(device, "drbd_req_complete: Logic BUG rq_state = 0x%x\n", s);
		return;
	}

	if (!req->master_bio) {
		drbd_err(device, "drbd_req_complete: Logic BUG, master_bio == NULL!\n");
		return;
	}

	/*
	 * 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
	 * in private_bio within drbd_request_endio.
	 */
	ok = (s & RQ_LOCAL_OK) || (s & RQ_NET_OK);
	error = PTR_ERR(req->private_bio);

	/* Before we can signal completion to the upper layers,
	 * we may need to close the current transfer log epoch.
	 * We are within the request lock, so we can simply compare
	 * the request epoch number with the current transfer log
	 * epoch number.  If they match, increase the current_tle_nr,
	 * and reset the transfer log epoch write_cnt.
	 */
	if (op_is_write(bio_op(req->master_bio)) &&
	    req->epoch == atomic_read(&first_peer_device(device)->connection->current_tle_nr))
		start_new_tl_epoch(first_peer_device(device)->connection);

	/* Update disk stats */
	bio_end_io_acct(req->master_bio, req->start_jif);

	/* If READ failed,
	 * have it be pushed back to the retry work queue,
	 * so it will re-enter __drbd_make_request(),
	 * and be re-assigned to a suitable local or remote path,
	 * or failed if we do not have access to good data anymore.
	 *
	 * Unless it was failed early by __drbd_make_request(),
	 * because no path was available, in which case
	 * it was not even added to the transfer_log.
	 *
	 * read-ahead may fail, and will not be retried.
	 *
	 * WRITE should have used all available paths already.
	 */
	if (!ok &&
	    bio_op(req->master_bio) == REQ_OP_READ &&
	    !(req->master_bio->bi_opf & REQ_RAHEAD) &&
	    !list_empty(&req->tl_requests))
		req->rq_state |= RQ_POSTPONED;

	if (!(req->rq_state & RQ_POSTPONED)) {
		m->error = ok ? 0 : (error ?: -EIO);
		m->bio = req->master_bio;
		req->master_bio = NULL;
		/* We leave it in the tree, to be able to verify later
		 * write-acks in protocol != C during resync.
		 * But we mark it as "complete", so it won't be counted as
		 * conflict in a multi-primary setup. */
		req->i.completed = true;
	}

	if (req->i.waiting)
		wake_up(&device->misc_wait);

	/* Either we are about to complete to upper layers,
	 * or we will restart this request.
	 * In either case, the request object will be destroyed soon,
	 * so better remove it from all lists. */
	list_del_init(&req->req_pending_master_completion);
}

/* still holds resource->req_lock */
static void drbd_req_put_completion_ref(struct drbd_request *req, struct bio_and_error *m, int put)
{
	struct drbd_device *device = req->device;
	D_ASSERT(device, m || (req->rq_state & RQ_POSTPONED));

	if (!put)
		return;

	if (!atomic_sub_and_test(put, &req->completion_ref))
		return;

	drbd_req_complete(req, m);

	/* local completion may still come in later,
	 * we need to keep the req object around. */
	if (req->rq_state & RQ_LOCAL_ABORTED)
		return;

	if (req->rq_state & RQ_POSTPONED) {
		/* don't destroy the req object just yet,
		 * but queue it for retry */
		drbd_restart_request(req);
		return;
	}

	kref_put(&req->kref, drbd_req_destroy);
}

static void set_if_null_req_next(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_next == NULL)
		connection->req_next = req;
}

static void advance_conn_req_next(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_next != req)
		return;
	list_for_each_entry_continue(req, &connection->transfer_log, tl_requests) {
		const unsigned s = req->rq_state;
		if (s & RQ_NET_QUEUED)
			break;
	}
	if (&req->tl_requests == &connection->transfer_log)
		req = NULL;
	connection->req_next = req;
}

static void set_if_null_req_ack_pending(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_ack_pending == NULL)
		connection->req_ack_pending = req;
}

static void advance_conn_req_ack_pending(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_ack_pending != req)
		return;
	list_for_each_entry_continue(req, &connection->transfer_log, tl_requests) {
		const unsigned s = req->rq_state;
		if ((s & RQ_NET_SENT) && (s & RQ_NET_PENDING))
			break;
	}
	if (&req->tl_requests == &connection->transfer_log)
		req = NULL;
	connection->req_ack_pending = req;
}

static void set_if_null_req_not_net_done(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_not_net_done == NULL)
		connection->req_not_net_done = req;
}

static void advance_conn_req_not_net_done(struct drbd_peer_device *peer_device, struct drbd_request *req)
{
	struct drbd_connection *connection = peer_device ? peer_device->connection : NULL;
	if (!connection)
		return;
	if (connection->req_not_net_done != req)
		return;
	list_for_each_entry_continue(req, &connection->transfer_log, tl_requests) {
		const unsigned s = req->rq_state;
		if ((s & RQ_NET_SENT) && !(s & RQ_NET_DONE))
			break;
	}
	if (&req->tl_requests == &connection->transfer_log)
		req = NULL;
	connection->req_not_net_done = req;
}

/* I'd like this to be the only place that manipulates
 * req->completion_ref and req->kref. */
static void mod_rq_state(struct drbd_request *req, struct bio_and_error *m,
		int clear, int set)
{
	struct drbd_device *device = req->device;
	struct drbd_peer_device *peer_device = first_peer_device(device);
	unsigned s = req->rq_state;
	int c_put = 0;

	if (drbd_suspended(device) && !((s | clear) & RQ_COMPLETION_SUSP))
		set |= RQ_COMPLETION_SUSP;

	/* apply */

	req->rq_state &= ~clear;
	req->rq_state |= set;

	/* no change? */
	if (req->rq_state == s)
		return;

	/* intent: get references */

	kref_get(&req->kref);

	if (!(s & RQ_LOCAL_PENDING) && (set & RQ_LOCAL_PENDING))
		atomic_inc(&req->completion_ref);

	if (!(s & RQ_NET_PENDING) && (set & RQ_NET_PENDING)) {
		inc_ap_pending(device);
		atomic_inc(&req->completion_ref);
	}

	if (!(s & RQ_NET_QUEUED) && (set & RQ_NET_QUEUED)) {
		atomic_inc(&req->completion_ref);
		set_if_null_req_next(peer_device, req);
	}

	if (!(s & RQ_EXP_BARR_ACK) && (set & RQ_EXP_BARR_ACK))
		kref_get(&req->kref); /* wait for the DONE */

	if (!(s & RQ_NET_SENT) && (set & RQ_NET_SENT)) {
		/* potentially already completed in the ack_receiver thread */
		if (!(s & RQ_NET_DONE)) {
			atomic_add(req->i.size >> 9, &device->ap_in_flight);
			set_if_null_req_not_net_done(peer_device, req);
		}
		if (req->rq_state & RQ_NET_PENDING)
			set_if_null_req_ack_pending(peer_device, req);
	}

	if (!(s & RQ_COMPLETION_SUSP) && (set & RQ_COMPLETION_SUSP))
		atomic_inc(&req->completion_ref);

	/* progress: put references */

	if ((s & RQ_COMPLETION_SUSP) && (clear & RQ_COMPLETION_SUSP))
		++c_put;

	if (!(s & RQ_LOCAL_ABORTED) && (set & RQ_LOCAL_ABORTED)) {
		D_ASSERT(device, req->rq_state & RQ_LOCAL_PENDING);
		++c_put;
	}

	if ((s & RQ_LOCAL_PENDING) && (clear & RQ_LOCAL_PENDING)) {
		if (req->rq_state & RQ_LOCAL_ABORTED)
			kref_put(&req->kref, drbd_req_destroy);
		else
			++c_put;
		list_del_init(&req->req_pending_local);
	}

	if ((s & RQ_NET_PENDING) && (clear & RQ_NET_PENDING)) {
		dec_ap_pending(device);
		++c_put;
		req->acked_jif = jiffies;
		advance_conn_req_ack_pending(peer_device, req);
	}

	if ((s & RQ_NET_QUEUED) && (clear & RQ_NET_QUEUED)) {
		++c_put;
		advance_conn_req_next(peer_device, req);
	}

	if (!(s & RQ_NET_DONE) && (set & RQ_NET_DONE)) {
		if (s & RQ_NET_SENT)
			atomic_sub(req->i.size >> 9, &device->ap_in_flight);
		if (s & RQ_EXP_BARR_ACK)
			kref_put(&req->kref, drbd_req_destroy);
		req->net_done_jif = jiffies;

		/* in ahead/behind mode, or just in case,
		 * before we finally destroy this request,
		 * the caching pointers must not reference it anymore */
		advance_conn_req_next(peer_device, req);
		advance_conn_req_ack_pending(peer_device, req);
		advance_conn_req_not_net_done(peer_device, req);
	}

	/* potentially complete and destroy */

	/* If we made progress, retry conflicting peer requests, if any. */
	if (req->i.waiting)
		wake_up(&device->misc_wait);

	drbd_req_put_completion_ref(req, m, c_put);
	kref_put(&req->kref, drbd_req_destroy);
}

static void drbd_report_io_error(struct drbd_device *device, struct drbd_request *req)
{
        char b[BDEVNAME_SIZE];

	if (!__ratelimit(&drbd_ratelimit_state))
		return;

	drbd_warn(device, "local %s IO error sector %llu+%u on %s\n",
			(req->rq_state & RQ_WRITE) ? "WRITE" : "READ",
			(unsigned long long)req->i.sector,
			req->i.size >> 9,
			bdevname(device->ldev->backing_bdev, b));
}

/* Helper for HANDED_OVER_TO_NETWORK.
 * Is this a protocol A write (neither WRITE_ACK nor RECEIVE_ACK expected)?
 * Is it also still "PENDING"?
 * --> If so, clear PENDING and set NET_OK below.
 * If it is a protocol A write, but not RQ_PENDING anymore, neg-ack was faster
 * (and we must not set RQ_NET_OK) */
static inline bool is_pending_write_protocol_A(struct drbd_request *req)
{
	return (req->rq_state &
		   (RQ_WRITE|RQ_NET_PENDING|RQ_EXP_WRITE_ACK|RQ_EXP_RECEIVE_ACK))
		== (RQ_WRITE|RQ_NET_PENDING);
}

/* 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 ...
 */
int __req_mod(struct drbd_request *req, enum drbd_req_event what,
		struct bio_and_error *m)
{
	struct drbd_device *const device = req->device;
	struct drbd_peer_device *const peer_device = first_peer_device(device);
	struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL;
	struct net_conf *nc;
	int p, rv = 0;

	if (m)
		m->bio = NULL;

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

	/* does not happen...
	 * initialization done in drbd_req_new
	case CREATED:
		break;
		*/

	case TO_BE_SENT: /* via network */
		/* reached via __drbd_make_request
		 * and from w_read_retry_remote */
		D_ASSERT(device, !(req->rq_state & RQ_NET_MASK));
		rcu_read_lock();
		nc = rcu_dereference(connection->net_conf);
		p = nc->wire_protocol;
		rcu_read_unlock();
		req->rq_state |=
			p == DRBD_PROT_C ? RQ_EXP_WRITE_ACK :
			p == DRBD_PROT_B ? RQ_EXP_RECEIVE_ACK : 0;
		mod_rq_state(req, m, 0, RQ_NET_PENDING);
		break;

	case TO_BE_SUBMITTED: /* locally */
		/* reached via __drbd_make_request */
		D_ASSERT(device, !(req->rq_state & RQ_LOCAL_MASK));
		mod_rq_state(req, m, 0, RQ_LOCAL_PENDING);
		break;

	case COMPLETED_OK:
		if (req->rq_state & RQ_WRITE)
			device->writ_cnt += req->i.size >> 9;
		else
			device->read_cnt += req->i.size >> 9;

		mod_rq_state(req, m, RQ_LOCAL_PENDING,
				RQ_LOCAL_COMPLETED|RQ_LOCAL_OK);
		break;

	case ABORT_DISK_IO:
		mod_rq_state(req, m, 0, RQ_LOCAL_ABORTED);
		break;

	case WRITE_COMPLETED_WITH_ERROR:
		drbd_report_io_error(device, req);
		__drbd_chk_io_error(device, DRBD_WRITE_ERROR);
		mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED);
		break;

	case READ_COMPLETED_WITH_ERROR:
		drbd_set_out_of_sync(device, req->i.sector, req->i.size);
		drbd_report_io_error(device, req);
		__drbd_chk_io_error(device, DRBD_READ_ERROR);
		fallthrough;
	case READ_AHEAD_COMPLETED_WITH_ERROR:
		/* it is legal to fail read-ahead, no __drbd_chk_io_error in that case. */
		mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED);
		break;

	case DISCARD_COMPLETED_NOTSUPP:
	case DISCARD_COMPLETED_WITH_ERROR:
		/* I'd rather not detach from local disk just because it
		 * failed a REQ_OP_DISCARD. */
		mod_rq_state(req, m, RQ_LOCAL_PENDING, RQ_LOCAL_COMPLETED);
		break;

	case QUEUE_FOR_NET_READ:
		/* READ, and
		 * no local disk,
		 * or target area marked as invalid,
		 * or just got an io-error. */
		/* from __drbd_make_request
		 * or from bio_endio during read io-error recovery */

		/* So we can verify the handle in the answer packet.
		 * Corresponding drbd_remove_request_interval is in
		 * drbd_req_complete() */
		D_ASSERT(device, drbd_interval_empty(&req->i));
		drbd_insert_interval(&device->read_requests, &req->i);

		set_bit(UNPLUG_REMOTE, &device->flags);

		D_ASSERT(device, req->rq_state & RQ_NET_PENDING);
		D_ASSERT(device, (req->rq_state & RQ_LOCAL_MASK) == 0);
		mod_rq_state(req, m, 0, RQ_NET_QUEUED);
		req->w.cb = w_send_read_req;
		drbd_queue_work(&connection->sender_work,
				&req->w);
		break;

	case QUEUE_FOR_NET_WRITE:
		/* assert something? */
		/* from __drbd_make_request only */

		/* Corresponding drbd_remove_request_interval is in
		 * drbd_req_complete() */
		D_ASSERT(device, drbd_interval_empty(&req->i));
		drbd_insert_interval(&device->write_requests, &req->i);

		/* 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
		 * __drbd_make_request, because we now may set the bit
		 * again ourselves to close the current epoch.
		 *
		 * Add req to the (now) current epoch (barrier). */

		/* 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, &device->flags);

		/* queue work item to send data */
		D_ASSERT(device, req->rq_state & RQ_NET_PENDING);
		mod_rq_state(req, m, 0, RQ_NET_QUEUED|RQ_EXP_BARR_ACK);
		req->w.cb =  w_send_dblock;
		drbd_queue_work(&connection->sender_work,
				&req->w);

		/* close the epoch, in case it outgrew the limit */
		rcu_read_lock();
		nc = rcu_dereference(connection->net_conf);
		p = nc->max_epoch_size;
		rcu_read_unlock();
		if (connection->current_tle_writes >= p)
			start_new_tl_epoch(connection);

		break;

	case QUEUE_FOR_SEND_OOS:
		mod_rq_state(req, m, 0, RQ_NET_QUEUED);
		req->w.cb =  w_send_out_of_sync;
		drbd_queue_work(&connection->sender_work,
				&req->w);
		break;

	case READ_RETRY_REMOTE_CANCELED:
	case SEND_CANCELED:
	case SEND_FAILED:
		/* real cleanup will be done from tl_clear.  just update flags
		 * so it is no longer marked as on the worker queue */
		mod_rq_state(req, m, RQ_NET_QUEUED, 0);
		break;

	case HANDED_OVER_TO_NETWORK:
		/* assert something? */
		if (is_pending_write_protocol_A(req))
			/* this is what is dangerous about protocol A:
			 * pretend it was successfully written on the peer. */
			mod_rq_state(req, m, RQ_NET_QUEUED|RQ_NET_PENDING,
						RQ_NET_SENT|RQ_NET_OK);
		else
			mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_SENT);
		/* 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. */
		break;

	case OOS_HANDED_TO_NETWORK:
		/* Was not set PENDING, no longer QUEUED, so is now DONE
		 * as far as this connection is concerned. */
		mod_rq_state(req, m, RQ_NET_QUEUED, RQ_NET_DONE);
		break;

	case CONNECTION_LOST_WHILE_PENDING:
		/* transfer log cleanup after connection loss */
		mod_rq_state(req, m,
				RQ_NET_OK|RQ_NET_PENDING|RQ_COMPLETION_SUSP,
				RQ_NET_DONE);
		break;

	case CONFLICT_RESOLVED:
		/* for superseded conflicting writes of multiple primaries,
		 * there is no need to keep anything in the tl, potential
		 * node crashes are covered by the activity log.
		 *
		 * If this request had been marked as RQ_POSTPONED before,
		 * it will actually not be completed, but "restarted",
		 * resubmitted from the retry worker context. */
		D_ASSERT(device, req->rq_state & RQ_NET_PENDING);
		D_ASSERT(device, req->rq_state & RQ_EXP_WRITE_ACK);
		mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_DONE|RQ_NET_OK);
		break;

	case WRITE_ACKED_BY_PEER_AND_SIS:
		req->rq_state |= RQ_NET_SIS;
		fallthrough;
	case WRITE_ACKED_BY_PEER:
		/* Normal operation protocol C: successfully written on peer.
		 * During resync, even in protocol != C,
		 * we requested an explicit write ack anyways.
		 * Which means we cannot even assert anything here.
		 * Nothing more 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. */
		goto ack_common;
	case RECV_ACKED_BY_PEER:
		D_ASSERT(device, req->rq_state & RQ_EXP_RECEIVE_ACK);
		/* protocol B; pretends to be successfully written on peer.
		 * see also notes above in HANDED_OVER_TO_NETWORK about
		 * protocol != C */
	ack_common:
		mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK);
		break;

	case POSTPONE_WRITE:
		D_ASSERT(device, req->rq_state & RQ_EXP_WRITE_ACK);
		/* 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(device, req->rq_state & RQ_NET_PENDING);
		req->rq_state |= RQ_POSTPONED;
		if (req->i.waiting)
			wake_up(&device->misc_wait);
		/* Do not clear RQ_NET_PENDING. This request will make further
		 * progress via restart_conflicting_writes() or
		 * fail_postponed_requests(). Hopefully. */
		break;

	case NEG_ACKED:
		mod_rq_state(req, m, RQ_NET_OK|RQ_NET_PENDING, 0);
		break;

	case FAIL_FROZEN_DISK_IO:
		if (!(req->rq_state & RQ_LOCAL_COMPLETED))
			break;
		mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0);
		break;

	case RESTART_FROZEN_DISK_IO:
		if (!(req->rq_state & RQ_LOCAL_COMPLETED))
			break;

		mod_rq_state(req, m,
				RQ_COMPLETION_SUSP|RQ_LOCAL_COMPLETED,
				RQ_LOCAL_PENDING);

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

		get_ldev(device); /* always succeeds in this call path */
		req->w.cb = w_restart_disk_io;
		drbd_queue_work(&connection->sender_work,
				&req->w);
		break;

	case RESEND:
		/* Simply complete (local only) READs. */
		if (!(req->rq_state & RQ_WRITE) && !req->w.cb) {
			mod_rq_state(req, m, RQ_COMPLETION_SUSP, 0);
			break;
		}

		/* If RQ_NET_OK is already set, we got a P_WRITE_ACK or P_RECV_ACK
		   before the connection loss (B&C only); only P_BARRIER_ACK
		   (or the local completion?) was missing when we suspended.
		   Throwing them out of the TL here by pretending we got a BARRIER_ACK.
		   During connection handshake, we ensure that the peer was not rebooted. */
		if (!(req->rq_state & RQ_NET_OK)) {
			/* FIXME could this possibly be a req->dw.cb == w_send_out_of_sync?
			 * in that case we must not set RQ_NET_PENDING. */

			mod_rq_state(req, m, RQ_COMPLETION_SUSP, RQ_NET_QUEUED|RQ_NET_PENDING);
			if (req->w.cb) {
				/* w.cb expected to be w_send_dblock, or w_send_read_req */
				drbd_queue_work(&connection->sender_work,
						&req->w);
				rv = req->rq_state & RQ_WRITE ? MR_WRITE : MR_READ;
			} /* else: FIXME can this happen? */
			break;
		}
		fallthrough;	/* to BARRIER_ACKED */

	case BARRIER_ACKED:
		/* barrier ack for READ requests does not make sense */
		if (!(req->rq_state & RQ_WRITE))
			break;

		if (req->rq_state & RQ_NET_PENDING) {
			/* barrier came in before all requests were acked.
			 * this is bad, because if the connection is lost now,
			 * we won't be able to clean them up... */
			drbd_err(device, "FIXME (BARRIER_ACKED but pending)\n");
		}
		/* Allowed to complete requests, even while suspended.
		 * As this is called for all requests within a matching epoch,
		 * we need to filter, and only set RQ_NET_DONE for those that
		 * have actually been on the wire. */
		mod_rq_state(req, m, RQ_COMPLETION_SUSP,
				(req->rq_state & RQ_NET_MASK) ? RQ_NET_DONE : 0);
		break;

	case DATA_RECEIVED:
		D_ASSERT(device, req->rq_state & RQ_NET_PENDING);
		mod_rq_state(req, m, RQ_NET_PENDING, RQ_NET_OK|RQ_NET_DONE);
		break;

	case QUEUE_AS_DRBD_BARRIER:
		start_new_tl_epoch(connection);
		mod_rq_state(req, m, 0, RQ_NET_OK|RQ_NET_DONE);
		break;
	}

	return rv;
}

/* 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.
 */
static bool drbd_may_do_local_read(struct drbd_device *device, sector_t sector, int size)
{
	unsigned long sbnr, ebnr;
	sector_t esector, nr_sectors;

	if (device->state.disk == D_UP_TO_DATE)
		return true;
	if (device->state.disk != D_INCONSISTENT)
		return false;
	esector = sector + (size >> 9) - 1;
	nr_sectors = get_capacity(device->vdisk);
	D_ASSERT(device, sector  < nr_sectors);
	D_ASSERT(device, esector < nr_sectors);

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

	return drbd_bm_count_bits(device, sbnr, ebnr) == 0;
}

static bool remote_due_to_read_balancing(struct drbd_device *device, sector_t sector,
		enum drbd_read_balancing rbm)
{
	int stripe_shift;

	switch (rbm) {
	case RB_CONGESTED_REMOTE:
		return bdi_read_congested(
			device->ldev->backing_bdev->bd_disk->bdi);
	case RB_LEAST_PENDING:
		return atomic_read(&device->local_cnt) >
			atomic_read(&device->ap_pending_cnt) + atomic_read(&device->rs_pending_cnt);
	case RB_32K_STRIPING:  /* stripe_shift = 15 */
	case RB_64K_STRIPING:
	case RB_128K_STRIPING:
	case RB_256K_STRIPING:
	case RB_512K_STRIPING:
	case RB_1M_STRIPING:   /* stripe_shift = 20 */
		stripe_shift = (rbm - RB_32K_STRIPING + 15);
		return (sector >> (stripe_shift - 9)) & 1;
	case RB_ROUND_ROBIN:
		return test_and_change_bit(READ_BALANCE_RR, &device->flags);
	case RB_PREFER_REMOTE:
		return true;
	case RB_PREFER_LOCAL:
	default:
		return false;
	}
}

/*
 * 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.
 *
 * Only way out: remove the conflicting intervals from the tree.
 */
static void complete_conflicting_writes(struct drbd_request *req)
{
	DEFINE_WAIT(wait);
	struct drbd_device *device = req->device;
	struct drbd_interval *i;
	sector_t sector = req->i.sector;
	int size = req->i.size;

	for (;;) {
		drbd_for_each_overlap(i, &device->write_requests, sector, size) {
			/* Ignore, if already completed to upper layers. */
			if (i->completed)
				continue;
			/* Handle the first found overlap.  After the schedule
			 * we have to restart the tree walk. */
			break;
		}
		if (!i)	/* if any */
			break;

		/* Indicate to wake up device->misc_wait on progress.  */
		prepare_to_wait(&device->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
		i->waiting = true;
		spin_unlock_irq(&device->resource->req_lock);
		schedule();
		spin_lock_irq(&device->resource->req_lock);
	}
	finish_wait(&device->misc_wait, &wait);
}

/* called within req_lock */
static void maybe_pull_ahead(struct drbd_device *device)
{
	struct drbd_connection *connection = first_peer_device(device)->connection;
	struct net_conf *nc;
	bool congested = false;
	enum drbd_on_congestion on_congestion;

	rcu_read_lock();
	nc = rcu_dereference(connection->net_conf);
	on_congestion = nc ? nc->on_congestion : OC_BLOCK;
	rcu_read_unlock();
	if (on_congestion == OC_BLOCK ||
	    connection->agreed_pro_version < 96)
		return;

	if (on_congestion == OC_PULL_AHEAD && device->state.conn == C_AHEAD)
		return; /* nothing to do ... */

	/* If I don't even have good local storage, we can not reasonably try
	 * to pull ahead of the peer. We also need the local reference to make
	 * sure device->act_log is there.
	 */
	if (!get_ldev_if_state(device, D_UP_TO_DATE))
		return;

	if (nc->cong_fill &&
	    atomic_read(&device->ap_in_flight) >= nc->cong_fill) {
		drbd_info(device, "Congestion-fill threshold reached\n");
		congested = true;
	}

	if (device->act_log->used >= nc->cong_extents) {
		drbd_info(device, "Congestion-extents threshold reached\n");
		congested = true;
	}

	if (congested) {
		/* start a new epoch for non-mirrored writes */
		start_new_tl_epoch(first_peer_device(device)->connection);

		if (on_congestion == OC_PULL_AHEAD)
			_drbd_set_state(_NS(device, conn, C_AHEAD), 0, NULL);
		else  /*nc->on_congestion == OC_DISCONNECT */
			_drbd_set_state(_NS(device, conn, C_DISCONNECTING), 0, NULL);
	}
	put_ldev(device);
}

/* If this returns false, and req->private_bio is still set,
 * this should be submitted locally.
 *
 * If it returns false, but req->private_bio is not set,
 * we do not have access to good data :(
 *
 * Otherwise, this destroys req->private_bio, if any,
 * and returns true.
 */
static bool do_remote_read(struct drbd_request *req)
{
	struct drbd_device *device = req->device;
	enum drbd_read_balancing rbm;

	if (req->private_bio) {
		if (!drbd_may_do_local_read(device,
					req->i.sector, req->i.size)) {
			bio_put(req->private_bio);
			req->private_bio = NULL;
			put_ldev(device);
		}
	}

	if (device->state.pdsk != D_UP_TO_DATE)
		return false;

	if (req->private_bio == NULL)
		return true;

	/* TODO: improve read balancing decisions, take into account drbd
	 * protocol, pending requests etc. */

	rcu_read_lock();
	rbm = rcu_dereference(device->ldev->disk_conf)->read_balancing;
	rcu_read_unlock();

	if (rbm == RB_PREFER_LOCAL && req->private_bio)
		return false; /* submit locally */

	if (remote_due_to_read_balancing(device, req->i.sector, rbm)) {
		if (req->private_bio) {
			bio_put(req->private_bio);
			req->private_bio = NULL;
			put_ldev(device);
		}
		return true;
	}

	return false;
}

bool drbd_should_do_remote(union drbd_dev_state s)
{
	return s.pdsk == D_UP_TO_DATE ||
		(s.pdsk >= D_INCONSISTENT &&
		 s.conn >= C_WF_BITMAP_T &&
		 s.conn < C_AHEAD);
	/* Before proto 96 that was >= CONNECTED instead of >= C_WF_BITMAP_T.
	   That is equivalent since before 96 IO was frozen in the C_WF_BITMAP*
	   states. */
}

static bool drbd_should_send_out_of_sync(union drbd_dev_state s)
{
	return s.conn == C_AHEAD || s.conn == C_WF_BITMAP_S;
	/* pdsk = D_INCONSISTENT as a consequence. Protocol 96 check not necessary
	   since we enter state C_AHEAD only if proto >= 96 */
}

/* returns number of connections (== 1, for drbd 8.4)
 * expected to actually write this data,
 * which does NOT include those that we are L_AHEAD for. */
static int drbd_process_write_request(struct drbd_request *req)
{
	struct drbd_device *device = req->device;
	int remote, send_oos;

	remote = drbd_should_do_remote(device->state);
	send_oos = drbd_should_send_out_of_sync(device->state);

	/* Need to replicate writes.  Unless it is an empty flush,
	 * which is better mapped to a DRBD P_BARRIER packet,
	 * also for drbd wire protocol compatibility reasons.
	 * If this was a flush, just start a new epoch.
	 * Unless the current epoch was empty anyways, or we are not currently
	 * replicating, in which case there is no point. */
	if (unlikely(req->i.size == 0)) {
		/* The only size==0 bios we expect are empty flushes. */
		D_ASSERT(device, req->master_bio->bi_opf & REQ_PREFLUSH);
		if (remote)
			_req_mod(req, QUEUE_AS_DRBD_BARRIER);
		return remote;
	}

	if (!remote && !send_oos)
		return 0;

	D_ASSERT(device, !(remote && send_oos));

	if (remote) {
		_req_mod(req, TO_BE_SENT);
		_req_mod(req, QUEUE_FOR_NET_WRITE);
	} else if (drbd_set_out_of_sync(device, req->i.sector, req->i.size))
		_req_mod(req, QUEUE_FOR_SEND_OOS);

	return remote;
}

static void drbd_process_discard_or_zeroes_req(struct drbd_request *req, int flags)
{
	int err = drbd_issue_discard_or_zero_out(req->device,
				req->i.sector, req->i.size >> 9, flags);
	if (err)
		req->private_bio->bi_status = BLK_STS_IOERR;
	bio_endio(req->private_bio);
}

static void
drbd_submit_req_private_bio(struct drbd_request *req)
{
	struct drbd_device *device = req->device;
	struct bio *bio = req->private_bio;
	unsigned int type;

	if (bio_op(bio) != REQ_OP_READ)
		type = DRBD_FAULT_DT_WR;
	else if (bio->bi_opf & REQ_RAHEAD)
		type = DRBD_FAULT_DT_RA;
	else
		type = DRBD_FAULT_DT_RD;

	/* State may have changed since we grabbed our reference on the
	 * ->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(device)) {
		if (drbd_insert_fault(device, type))
			bio_io_error(bio);
		else if (bio_op(bio) == REQ_OP_WRITE_ZEROES)
			drbd_process_discard_or_zeroes_req(req, EE_ZEROOUT |
			    ((bio->bi_opf & REQ_NOUNMAP) ? 0 : EE_TRIM));
		else if (bio_op(bio) == REQ_OP_DISCARD)
			drbd_process_discard_or_zeroes_req(req, EE_TRIM);
		else
			submit_bio_noacct(bio);
		put_ldev(device);
	} else
		bio_io_error(bio);
}

static void drbd_queue_write(struct drbd_device *device, struct drbd_request *req)
{
	spin_lock_irq(&device->resource->req_lock);
	list_add_tail(&req->tl_requests, &device->submit.writes);
	list_add_tail(&req->req_pending_master_completion,
			&device->pending_master_completion[1 /* WRITE */]);
	spin_unlock_irq(&device->resource->req_lock);
	queue_work(device->submit.wq, &device->submit.worker);
	/* do_submit() may sleep internally on al_wait, too */
	wake_up(&device->al_wait);
}

/* returns the new drbd_request pointer, if the caller is expected to
 * drbd_send_and_submit() it (to save latency), or NULL if we queued the
 * request on the submitter thread.
 * Returns ERR_PTR(-ENOMEM) if we cannot allocate a drbd_request.
 */
static struct drbd_request *
drbd_request_prepare(struct drbd_device *device, struct bio *bio)
{
	const int rw = bio_data_dir(bio);
	struct drbd_request *req;

	/* allocate outside of all locks; */
	req = drbd_req_new(device, bio);
	if (!req) {
		dec_ap_bio(device);
		/* only pass the error to the upper layers.
		 * if user cannot handle io errors, that's not our business. */
		drbd_err(device, "could not kmalloc() req\n");
		bio->bi_status = BLK_STS_RESOURCE;
		bio_endio(bio);
		return ERR_PTR(-ENOMEM);
	}

	/* Update disk stats */
	req->start_jif = bio_start_io_acct(req->master_bio);

	if (!get_ldev(device)) {
		bio_put(req->private_bio);
		req->private_bio = NULL;
	}

	/* process discards always from our submitter thread */
	if (bio_op(bio) == REQ_OP_WRITE_ZEROES ||
	    bio_op(bio) == REQ_OP_DISCARD)
		goto queue_for_submitter_thread;

	if (rw == WRITE && req->private_bio && req->i.size
	&& !test_bit(AL_SUSPENDED, &device->flags)) {
		if (!drbd_al_begin_io_fastpath(device, &req->i))
			goto queue_for_submitter_thread;
		req->rq_state |= RQ_IN_ACT_LOG;
		req->in_actlog_jif = jiffies;
	}
	return req;

 queue_for_submitter_thread:
	atomic_inc(&device->ap_actlog_cnt);
	drbd_queue_write(device, req);
	return NULL;
}

/* Require at least one path to current data.
 * We don't want to allow writes on C_STANDALONE D_INCONSISTENT:
 * We would not allow to read what was written,
 * we would not have bumped the data generation uuids,
 * we would cause data divergence for all the wrong reasons.
 *
 * If we don't see at least one D_UP_TO_DATE, we will fail this request,
 * which either returns EIO, or, if OND_SUSPEND_IO is set, suspends IO,
 * and queues for retry later.
 */
static bool may_do_writes(struct drbd_device *device)
{
	const union drbd_dev_state s = device->state;
	return s.disk == D_UP_TO_DATE || s.pdsk == D_UP_TO_DATE;
}

struct drbd_plug_cb {
	struct blk_plug_cb cb;
	struct drbd_request *most_recent_req;
	/* do we need more? */
};

static void drbd_unplug(struct blk_plug_cb *cb, bool from_schedule)
{
	struct drbd_plug_cb *plug = container_of(cb, struct drbd_plug_cb, cb);
	struct drbd_resource *resource = plug->cb.data;
	struct drbd_request *req = plug->most_recent_req;

	kfree(cb);
	if (!req)
		return;

	spin_lock_irq(&resource->req_lock);
	/* In case the sender did not process it yet, raise the flag to
	 * have it followed with P_UNPLUG_REMOTE just after. */
	req->rq_state |= RQ_UNPLUG;
	/* but also queue a generic unplug */
	drbd_queue_unplug(req->device);
	kref_put(&req->kref, drbd_req_destroy);
	spin_unlock_irq(&resource->req_lock);
}

static struct drbd_plug_cb* drbd_check_plugged(struct drbd_resource *resource)
{
	/* A lot of text to say
	 * return (struct drbd_plug_cb*)blk_check_plugged(); */
	struct drbd_plug_cb *plug;
	struct blk_plug_cb *cb = blk_check_plugged(drbd_unplug, resource, sizeof(*plug));

	if (cb)
		plug = container_of(cb, struct drbd_plug_cb, cb);
	else
		plug = NULL;
	return plug;
}

static void drbd_update_plug(struct drbd_plug_cb *plug, struct drbd_request *req)
{
	struct drbd_request *tmp = plug->most_recent_req;
	/* Will be sent to some peer.
	 * Remember to tag it with UNPLUG_REMOTE on unplug */
	kref_get(&req->kref);
	plug->most_recent_req = req;
	if (tmp)
		kref_put(&tmp->kref, drbd_req_destroy);
}

static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request *req)
{
	struct drbd_resource *resource = device->resource;
	const int rw = bio_data_dir(req->master_bio);
	struct bio_and_error m = { NULL, };
	bool no_remote = false;
	bool submit_private_bio = false;

	spin_lock_irq(&resource->req_lock);
	if (rw == WRITE) {
		/* This may temporarily give up the req_lock,
		 * but will re-aquire it before it returns here.
		 * Needs to be before the check on drbd_suspended() */
		complete_conflicting_writes(req);
		/* no more giving up req_lock from now on! */

		/* check for congestion, and potentially stop sending
		 * full data updates, but start sending "dirty bits" only. */
		maybe_pull_ahead(device);
	}


	if (drbd_suspended(device)) {
		/* push back and retry: */
		req->rq_state |= RQ_POSTPONED;
		if (req->private_bio) {
			bio_put(req->private_bio);
			req->private_bio = NULL;
			put_ldev(device);
		}
		goto out;
	}

	/* We fail READ early, if we can not serve it.
	 * We must do this before req is registered on any lists.
	 * Otherwise, drbd_req_complete() will queue failed READ for retry. */
	if (rw != WRITE) {
		if (!do_remote_read(req) && !req->private_bio)
			goto nodata;
	}

	/* which transfer log epoch does this belong to? */
	req->epoch = atomic_read(&first_peer_device(device)->connection->current_tle_nr);

	/* no point in adding empty flushes to the transfer log,
	 * they are mapped to drbd barriers already. */
	if (likely(req->i.size!=0)) {
		if (rw == WRITE)
			first_peer_device(device)->connection->current_tle_writes++;

		list_add_tail(&req->tl_requests, &first_peer_device(device)->connection->transfer_log);
	}

	if (rw == WRITE) {
		if (req->private_bio && !may_do_writes(device)) {
			bio_put(req->private_bio);
			req->private_bio = NULL;
			put_ldev(device);
			goto nodata;
		}
		if (!drbd_process_write_request(req))
			no_remote = true;
	} else {
		/* We either have a private_bio, or we can read from remote.
		 * Otherwise we had done the goto nodata above. */
		if (req->private_bio == NULL) {
			_req_mod(req, TO_BE_SENT);
			_req_mod(req, QUEUE_FOR_NET_READ);
		} else
			no_remote = true;
	}

	if (no_remote == false) {
		struct drbd_plug_cb *plug = drbd_check_plugged(resource);
		if (plug)
			drbd_update_plug(plug, req);
	}

	/* If it took the fast path in drbd_request_prepare, add it here.
	 * The slow path has added it already. */
	if (list_empty(&req->req_pending_master_completion))
		list_add_tail(&req->req_pending_master_completion,
			&device->pending_master_completion[rw == WRITE]);
	if (req->private_bio) {
		/* needs to be marked within the same spinlock */
		req->pre_submit_jif = jiffies;
		list_add_tail(&req->req_pending_local,
			&device->pending_completion[rw == WRITE]);
		_req_mod(req, TO_BE_SUBMITTED);
		/* but we need to give up the spinlock to submit */
		submit_private_bio = true;
	} else if (no_remote) {
nodata:
		if (__ratelimit(&drbd_ratelimit_state))
			drbd_err(device, "IO ERROR: neither local nor remote data, sector %llu+%u\n",
					(unsigned long long)req->i.sector, req->i.size >> 9);
		/* A write may have been queued for send_oos, however.
		 * So we can not simply free it, we must go through drbd_req_put_completion_ref() */
	}

out:
	drbd_req_put_completion_ref(req, &m, 1);
	spin_unlock_irq(&resource->req_lock);

	/* Even though above is a kref_put(), this is safe.
	 * As long as we still need to submit our private bio,
	 * we hold a completion ref, and the request cannot disappear.
	 * If however this request did not even have a private bio to submit
	 * (e.g. remote read), req may already be invalid now.
	 * That's why we cannot check on req->private_bio. */
	if (submit_private_bio)
		drbd_submit_req_private_bio(req);
	if (m.bio)
		complete_master_bio(device, &m);
}

void __drbd_make_request(struct drbd_device *device, struct bio *bio)
{
	struct drbd_request *req = drbd_request_prepare(device, bio);
	if (IS_ERR_OR_NULL(req))
		return;
	drbd_send_and_submit(device, req);
}

static void submit_fast_path(struct drbd_device *device, struct list_head *incoming)
{
	struct blk_plug plug;
	struct drbd_request *req, *tmp;

	blk_start_plug(&plug);
	list_for_each_entry_safe(req, tmp, incoming, tl_requests) {
		const int rw = bio_data_dir(req->master_bio);

		if (rw == WRITE /* rw != WRITE should not even end up here! */
		&& req->private_bio && req->i.size
		&& !test_bit(AL_SUSPENDED, &device->flags)) {
			if (!drbd_al_begin_io_fastpath(device, &req->i))
				continue;

			req->rq_state |= RQ_IN_ACT_LOG;
			req->in_actlog_jif = jiffies;
			atomic_dec(&device->ap_actlog_cnt);
		}

		list_del_init(&req->tl_requests);
		drbd_send_and_submit(device, req);
	}
	blk_finish_plug(&plug);
}

static bool prepare_al_transaction_nonblock(struct drbd_device *device,
					    struct list_head *incoming,
					    struct list_head *pending,
					    struct list_head *later)
{
	struct drbd_request *req;
	int wake = 0;
	int err;

	spin_lock_irq(&device->al_lock);
	while ((req = list_first_entry_or_null(incoming, struct drbd_request, tl_requests))) {
		err = drbd_al_begin_io_nonblock(device, &req->i);
		if (err == -ENOBUFS)
			break;
		if (err == -EBUSY)
			wake = 1;
		if (err)
			list_move_tail(&req->tl_requests, later);
		else
			list_move_tail(&req->tl_requests, pending);
	}
	spin_unlock_irq(&device->al_lock);
	if (wake)
		wake_up(&device->al_wait);
	return !list_empty(pending);
}

static void send_and_submit_pending(struct drbd_device *device, struct list_head *pending)
{
	struct blk_plug plug;
	struct drbd_request *req;

	blk_start_plug(&plug);
	while ((req = list_first_entry_or_null(pending, struct drbd_request, tl_requests))) {
		req->rq_state |= RQ_IN_ACT_LOG;
		req->in_actlog_jif = jiffies;
		atomic_dec(&device->ap_actlog_cnt);
		list_del_init(&req->tl_requests);
		drbd_send_and_submit(device, req);
	}
	blk_finish_plug(&plug);
}

void do_submit(struct work_struct *ws)
{
	struct drbd_device *device = container_of(ws, struct drbd_device, submit.worker);
	LIST_HEAD(incoming);	/* from drbd_make_request() */
	LIST_HEAD(pending);	/* to be submitted after next AL-transaction commit */
	LIST_HEAD(busy);	/* blocked by resync requests */

	/* grab new incoming requests */
	spin_lock_irq(&device->resource->req_lock);
	list_splice_tail_init(&device->submit.writes, &incoming);
	spin_unlock_irq(&device->resource->req_lock);

	for (;;) {
		DEFINE_WAIT(wait);

		/* move used-to-be-busy back to front of incoming */
		list_splice_init(&busy, &incoming);
		submit_fast_path(device, &incoming);
		if (list_empty(&incoming))
			break;

		for (;;) {
			prepare_to_wait(&device->al_wait, &wait, TASK_UNINTERRUPTIBLE);

			list_splice_init(&busy, &incoming);
			prepare_al_transaction_nonblock(device, &incoming, &pending, &busy);
			if (!list_empty(&pending))
				break;

			schedule();

			/* If all currently "hot" activity log extents are kept busy by
			 * incoming requests, we still must not totally starve new
			 * requests to "cold" extents.
			 * Something left on &incoming means there had not been
			 * enough update slots available, and the activity log
			 * has been marked as "starving".
			 *
			 * Try again now, without looking for new requests,
			 * effectively blocking all new requests until we made
			 * at least _some_ progress with what we currently have.
			 */
			if (!list_empty(&incoming))
				continue;

			/* Nothing moved to pending, but nothing left
			 * on incoming: all moved to busy!
			 * Grab new and iterate. */
			spin_lock_irq(&device->resource->req_lock);
			list_splice_tail_init(&device->submit.writes, &incoming);
			spin_unlock_irq(&device->resource->req_lock);
		}
		finish_wait(&device->al_wait, &wait);

		/* If the transaction was full, before all incoming requests
		 * had been processed, skip ahead to commit, and iterate
		 * without splicing in more incoming requests from upper layers.
		 *
		 * Else, if all incoming have been processed,
		 * they have become either "pending" (to be submitted after
		 * next transaction commit) or "busy" (blocked by resync).
		 *
		 * Maybe more was queued, while we prepared the transaction?
		 * Try to stuff those into this transaction as well.
		 * Be strictly non-blocking here,
		 * we already have something to commit.
		 *
		 * Commit if we don't make any more progres.
		 */

		while (list_empty(&incoming)) {
			LIST_HEAD(more_pending);
			LIST_HEAD(more_incoming);
			bool made_progress;

			/* It is ok to look outside the lock,
			 * it's only an optimization anyways */
			if (list_empty(&device->submit.writes))
				break;

			spin_lock_irq(&device->resource->req_lock);
			list_splice_tail_init(&device->submit.writes, &more_incoming);
			spin_unlock_irq(&device->resource->req_lock);

			if (list_empty(&more_incoming))
				break;

			made_progress = prepare_al_transaction_nonblock(device, &more_incoming, &more_pending, &busy);

			list_splice_tail_init(&more_pending, &pending);
			list_splice_tail_init(&more_incoming, &incoming);
			if (!made_progress)
				break;
		}

		drbd_al_begin_io_commit(device);
		send_and_submit_pending(device, &pending);
	}
}

void drbd_submit_bio(struct bio *bio)
{
	struct drbd_device *device = bio->bi_bdev->bd_disk->private_data;

	blk_queue_split(&bio);

	/*
	 * what we "blindly" assume:
	 */
	D_ASSERT(device, IS_ALIGNED(bio->bi_iter.bi_size, 512));

	inc_ap_bio(device);
	__drbd_make_request(device, bio);
}

static bool net_timeout_reached(struct drbd_request *net_req,
		struct drbd_connection *connection,
		unsigned long now, unsigned long ent,
		unsigned int ko_count, unsigned int timeout)
{
	struct drbd_device *device = net_req->device;

	if (!time_after(now, net_req->pre_send_jif + ent))
		return false;

	if (time_in_range(now, connection->last_reconnect_jif, connection->last_reconnect_jif + ent))
		return false;

	if (net_req->rq_state & RQ_NET_PENDING) {
		drbd_warn(device, "Remote failed to finish a request within %ums > ko-count (%u) * timeout (%u * 0.1s)\n",
			jiffies_to_msecs(now - net_req->pre_send_jif), ko_count, timeout);
		return true;
	}

	/* We received an ACK already (or are using protocol A),
	 * but are waiting for the epoch closing barrier ack.
	 * Check if we sent the barrier already.  We should not blame the peer
	 * for being unresponsive, if we did not even ask it yet. */
	if (net_req->epoch == connection->send.current_epoch_nr) {
		drbd_warn(device,
			"We did not send a P_BARRIER for %ums > ko-count (%u) * timeout (%u * 0.1s); drbd kernel thread blocked?\n",
			jiffies_to_msecs(now - net_req->pre_send_jif), ko_count, timeout);
		return false;
	}

	/* Worst case: we may have been blocked for whatever reason, then
	 * suddenly are able to send a lot of requests (and epoch separating
	 * barriers) in quick succession.
	 * The timestamp of the net_req may be much too old and not correspond
	 * to the sending time of the relevant unack'ed barrier packet, so
	 * would trigger a spurious timeout.  The latest barrier packet may
	 * have a too recent timestamp to trigger the timeout, potentially miss
	 * a timeout.  Right now we don't have a place to conveniently store
	 * these timestamps.
	 * But in this particular situation, the application requests are still
	 * completed to upper layers, DRBD should still "feel" responsive.
	 * No need yet to kill this connection, it may still recover.
	 * If not, eventually we will have queued enough into the network for
	 * us to block. From that point of view, the timestamp of the last sent
	 * barrier packet is relevant enough.
	 */
	if (time_after(now, connection->send.last_sent_barrier_jif + ent)) {
		drbd_warn(device, "Remote failed to answer a P_BARRIER (sent at %lu jif; now=%lu jif) within %ums > ko-count (%u) * timeout (%u * 0.1s)\n",
			connection->send.last_sent_barrier_jif, now,
			jiffies_to_msecs(now - connection->send.last_sent_barrier_jif), ko_count, timeout);
		return true;
	}
	return false;
}

/* A request is considered timed out, if
 * - we have some effective timeout from the configuration,
 *   with some state restrictions applied,
 * - the oldest request is waiting for a response from the network
 *   resp. the local disk,
 * - the oldest request is in fact older than the effective timeout,
 * - the connection was established (resp. disk was attached)
 *   for longer than the timeout already.
 * Note that for 32bit jiffies and very stable connections/disks,
 * we may have a wrap around, which is catched by
 *   !time_in_range(now, last_..._jif, last_..._jif + timeout).
 *
 * Side effect: once per 32bit wrap-around interval, which means every
 * ~198 days with 250 HZ, we have a window where the timeout would need
 * to expire twice (worst case) to become effective. Good enough.
 */

void request_timer_fn(struct timer_list *t)
{
	struct drbd_device *device = from_timer(device, t, request_timer);
	struct drbd_connection *connection = first_peer_device(device)->connection;
	struct drbd_request *req_read, *req_write, *req_peer; /* oldest request */
	struct net_conf *nc;
	unsigned long oldest_submit_jif;
	unsigned long ent = 0, dt = 0, et, nt; /* effective timeout = ko_count * timeout */
	unsigned long now;
	unsigned int ko_count = 0, timeout = 0;

	rcu_read_lock();
	nc = rcu_dereference(connection->net_conf);
	if (nc && device->state.conn >= C_WF_REPORT_PARAMS) {
		ko_count = nc->ko_count;
		timeout = nc->timeout;
	}

	if (get_ldev(device)) { /* implicit state.disk >= D_INCONSISTENT */
		dt = rcu_dereference(device->ldev->disk_conf)->disk_timeout * HZ / 10;
		put_ldev(device);
	}
	rcu_read_unlock();


	ent = timeout * HZ/10 * ko_count;
	et = min_not_zero(dt, ent);

	if (!et)
		return; /* Recurring timer stopped */

	now = jiffies;
	nt = now + et;

	spin_lock_irq(&device->resource->req_lock);
	req_read = list_first_entry_or_null(&device->pending_completion[0], struct drbd_request, req_pending_local);
	req_write = list_first_entry_or_null(&device->pending_completion[1], struct drbd_request, req_pending_local);

	/* maybe the oldest request waiting for the peer is in fact still
	 * blocking in tcp sendmsg.  That's ok, though, that's handled via the
	 * socket send timeout, requesting a ping, and bumping ko-count in
	 * we_should_drop_the_connection().
	 */

	/* check the oldest request we did successfully sent,
	 * but which is still waiting for an ACK. */
	req_peer = connection->req_ack_pending;

	/* if we don't have such request (e.g. protocoll A)
	 * check the oldest requests which is still waiting on its epoch
	 * closing barrier ack. */
	if (!req_peer)
		req_peer = connection->req_not_net_done;

	/* evaluate the oldest peer request only in one timer! */
	if (req_peer && req_peer->device != device)
		req_peer = NULL;

	/* do we have something to evaluate? */
	if (req_peer == NULL && req_write == NULL && req_read == NULL)
		goto out;

	oldest_submit_jif =
		(req_write && req_read)
		? ( time_before(req_write->pre_submit_jif, req_read->pre_submit_jif)
		  ? req_write->pre_submit_jif : req_read->pre_submit_jif )
		: req_write ? req_write->pre_submit_jif
		: req_read ? req_read->pre_submit_jif : now;

	if (ent && req_peer && net_timeout_reached(req_peer, connection, now, ent, ko_count, timeout))
		_conn_request_state(connection, NS(conn, C_TIMEOUT), CS_VERBOSE | CS_HARD);

	if (dt && oldest_submit_jif != now &&
		 time_after(now, oldest_submit_jif + dt) &&
		!time_in_range(now, device->last_reattach_jif, device->last_reattach_jif + dt)) {
		drbd_warn(device, "Local backing device failed to meet the disk-timeout\n");
		__drbd_chk_io_error(device, DRBD_FORCE_DETACH);
	}

	/* Reschedule timer for the nearest not already expired timeout.
	 * Fallback to now + min(effective network timeout, disk timeout). */
	ent = (ent && req_peer && time_before(now, req_peer->pre_send_jif + ent))
		? req_peer->pre_send_jif + ent : now + et;
	dt = (dt && oldest_submit_jif != now && time_before(now, oldest_submit_jif + dt))
		? oldest_submit_jif + dt : now + et;
	nt = time_before(ent, dt) ? ent : dt;
out:
	spin_unlock_irq(&device->resource->req_lock);
	mod_timer(&device->request_timer, nt);
}