summaryrefslogtreecommitdiffstats
path: root/src/quic_tx.c
blob: 39b917652f09d2321945a8a823f9b4c7dae10ac2 (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
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
/*
 * QUIC protocol implementation. Lower layer with internal features implemented
 * here such as QUIC encryption, idle timeout, acknowledgement and
 * retransmission.
 *
 * Copyright 2020 HAProxy Technologies, Frederic Lecaille <flecaille@haproxy.com>
 *
 * This program 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 of the License, or (at your option) any later version.
 *
 */

#include <haproxy/quic_tx.h>

#include <haproxy/pool.h>
#include <haproxy/trace.h>
#include <haproxy/quic_cid.h>
#include <haproxy/quic_conn.h>
#include <haproxy/quic_retransmit.h>
#include <haproxy/quic_retry.h>
#include <haproxy/quic_sock.h>
#include <haproxy/quic_tls.h>
#include <haproxy/quic_trace.h>
#include <haproxy/ssl_sock-t.h>

DECLARE_POOL(pool_head_quic_tx_packet, "quic_tx_packet", sizeof(struct quic_tx_packet));
DECLARE_POOL(pool_head_quic_cc_buf, "quic_cc_buf", QUIC_MAX_CC_BUFSIZE);

static struct quic_tx_packet *qc_build_pkt(unsigned char **pos, const unsigned char *buf_end,
                                           struct quic_enc_level *qel, struct quic_tls_ctx *ctx,
                                           struct list *frms, struct quic_conn *qc,
                                           const struct quic_version *ver, size_t dglen, int pkt_type,
                                           int must_ack, int padding, int probe, int cc, int *err);

static void quic_packet_encrypt(unsigned char *payload, size_t payload_len,
                                unsigned char *aad, size_t aad_len, uint64_t pn,
                                struct quic_tls_ctx *tls_ctx, struct quic_conn *qc,
                                int *fail)
{
	unsigned char iv[QUIC_TLS_IV_LEN];
	unsigned char *tx_iv = tls_ctx->tx.iv;
	size_t tx_iv_sz = tls_ctx->tx.ivlen;
	struct enc_debug_info edi;

	TRACE_ENTER(QUIC_EV_CONN_ENCPKT, qc);
	*fail = 0;

	quic_aead_iv_build(iv, sizeof iv, tx_iv, tx_iv_sz, pn);

	if (!quic_tls_encrypt(payload, payload_len, aad, aad_len,
	                      tls_ctx->tx.ctx, tls_ctx->tx.aead, iv)) {
		TRACE_ERROR("QUIC packet encryption failed", QUIC_EV_CONN_ENCPKT, qc);
		*fail = 1;
		enc_debug_info_init(&edi, payload, payload_len, aad, aad_len, pn);
	}

	TRACE_LEAVE(QUIC_EV_CONN_ENCPKT, qc);
}

/* Free <pkt> TX packet and its attached frames.
 * This is the responsibility of the caller to remove this packet of
 * any data structure it was possibly attached to.
 */
static inline void free_quic_tx_packet(struct quic_conn *qc,
                                       struct quic_tx_packet *pkt)
{
	struct quic_frame *frm, *frmbak;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	if (!pkt)
		goto leave;

	list_for_each_entry_safe(frm, frmbak, &pkt->frms, list)
		qc_frm_free(qc, &frm);
	pool_free(pool_head_quic_tx_packet, pkt);

 leave:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
}

/* Allocate Tx buffer from <qc> quic-conn if needed.
 *
 * Returns allocated buffer or NULL on error.
 */
struct buffer *qc_txb_alloc(struct quic_conn *qc)
{
	struct buffer *buf = &qc->tx.buf;
	if (!b_alloc(buf, DB_MUX_TX))
		return NULL;

	return buf;
}

/* Free Tx buffer from <qc> if it is empty. */
void qc_txb_release(struct quic_conn *qc)
{
	struct buffer *buf = &qc->tx.buf;

	/* For the moment sending function is responsible to purge the buffer
	 * entirely. It may change in the future but this requires to be able
	 * to reuse old data.
	 * For the moment we do not care to leave data in the buffer for
	 * a connection which is supposed to be killed asap.
	 */
	BUG_ON_HOT(buf && b_data(buf));

	if (!b_data(buf)) {
		b_free(buf);
		offer_buffers(NULL, 1);
	}
}

/* Return the TX buffer dedicated to the "connection close" datagram to be built
 * if an immediate close is required after having allocated it or directly
 * allocate a TX buffer if an immediate close is not required.
 */
struct buffer *qc_get_txb(struct quic_conn *qc)
{
	struct buffer *buf;

	if (qc->flags & QUIC_FL_CONN_IMMEDIATE_CLOSE) {
		TRACE_PROTO("Immediate close required", QUIC_EV_CONN_PHPKTS, qc);
		buf = &qc->tx.cc_buf;
		if (b_is_null(buf)) {
			qc->tx.cc_buf_area = pool_alloc(pool_head_quic_cc_buf);
			if (!qc->tx.cc_buf_area)
				goto err;
		}

		/* In every case, initialize ->tx.cc_buf */
		qc->tx.cc_buf = b_make(qc->tx.cc_buf_area, QUIC_MAX_CC_BUFSIZE, 0, 0);
	}
	else {
		buf = qc_txb_alloc(qc);
		if (!buf)
			goto err;
	}

	return buf;
 err:
	return NULL;
}

/* Commit a datagram payload written into <buf> of length <length>. <first_pkt>
 * must contains the address of the first packet stored in the payload.
 *
 * Caller is responsible that there is enough space in the buffer.
 */
static void qc_txb_store(struct buffer *buf, uint16_t length,
                         struct quic_tx_packet *first_pkt)
{
	const size_t hdlen = sizeof(uint16_t) + sizeof(void *);
	BUG_ON_HOT(b_contig_space(buf) < hdlen); /* this must not happen */

	write_u16(b_tail(buf), length);
	write_ptr(b_tail(buf) + sizeof(length), first_pkt);
	b_add(buf, hdlen + length);
}

/* Returns 1 if a packet may be built for <qc> from <qel> encryption level
 * with <frms> as ack-eliciting frame list to send, 0 if not.
 * <cc> must equal to 1 if an immediate close was asked, 0 if not.
 * <probe> must equalt to 1 if a probing packet is required, 0 if not.
 * Also set <*must_ack> to inform the caller if an acknowledgement should be sent.
 */
static int qc_may_build_pkt(struct quic_conn *qc, struct list *frms,
                            struct quic_enc_level *qel, int cc, int probe,
                            int *must_ack)
{
	int force_ack = qel == qc->iel || qel == qc->hel;
	int nb_aepkts_since_last_ack = qel->pktns->rx.nb_aepkts_since_last_ack;

	/* An acknowledgement must be sent if this has been forced by the caller,
	 * typically during the handshake when the packets must be acknowledged as
	 * soon as possible. This is also the case when the ack delay timer has been
	 * triggered, or at least every QUIC_MAX_RX_AEPKTS_SINCE_LAST_ACK packets.
	 */
	*must_ack = (qc->flags & QUIC_FL_CONN_ACK_TIMER_FIRED) ||
		((qel->pktns->flags & QUIC_FL_PKTNS_ACK_REQUIRED) &&
		 (force_ack || nb_aepkts_since_last_ack >= QUIC_MAX_RX_AEPKTS_SINCE_LAST_ACK));

	TRACE_PRINTF(TRACE_LEVEL_DEVELOPER, QUIC_EV_CONN_PHPKTS, qc, 0, 0, 0,
	             "has_sec=%d cc=%d probe=%d must_ack=%d frms=%d prep_in_fligh=%llu cwnd=%llu",
	             quic_tls_has_tx_sec(qel), cc, probe, *must_ack, LIST_ISEMPTY(frms),
	             (ullong)qc->path->prep_in_flight, (ullong)qc->path->cwnd);

	/* Do not build any more packet if the TX secrets are not available or
	 * if there is nothing to send, i.e. if no CONNECTION_CLOSE or ACK are required
	 * and if there is no more packets to send upon PTO expiration
	 * and if there is no more ack-eliciting frames to send or in flight
	 * congestion control limit is reached for prepared data
	 */
	if (!quic_tls_has_tx_sec(qel) ||
	    (!cc && !probe && !*must_ack &&
	     (LIST_ISEMPTY(frms) || qc->path->prep_in_flight >= qc->path->cwnd))) {
		return 0;
	}

	return 1;
}

/* Free all frames in <l> list. In addition also remove all these frames
 * from the original ones if they are the results of duplications.
 */
static inline void qc_free_frm_list(struct quic_conn *qc, struct list *l)
{
	struct quic_frame *frm, *frmbak;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	list_for_each_entry_safe(frm, frmbak, l, list) {
		LIST_DEL_INIT(&frm->ref);
		qc_frm_free(qc, &frm);
	}

	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
}

/* Free <pkt> TX packet and all the packets coalesced to it. */
static inline void qc_free_tx_coalesced_pkts(struct quic_conn *qc,
                                             struct quic_tx_packet *p)
{
	struct quic_tx_packet *pkt, *nxt_pkt;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	for (pkt = p; pkt; pkt = nxt_pkt) {
		qc_free_frm_list(qc, &pkt->frms);
		nxt_pkt = pkt->next;
		pool_free(pool_head_quic_tx_packet, pkt);
	}

	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
}

/* Purge <buf> TX buffer from its prepare packets. */
static void qc_purge_tx_buf(struct quic_conn *qc, struct buffer *buf)
{
	while (b_contig_data(buf, 0)) {
		uint16_t dglen;
		struct quic_tx_packet *pkt;
		size_t headlen = sizeof dglen + sizeof pkt;

		dglen = read_u16(b_head(buf));
		pkt = read_ptr(b_head(buf) + sizeof dglen);
		qc_free_tx_coalesced_pkts(qc, pkt);
		b_del(buf, dglen + headlen);
	}

	BUG_ON(b_data(buf));
}

/* Send datagrams stored in <buf>.
 *
 * This function returns 1 for success. On error, there is several behavior
 * depending on underlying sendto() error :
 * - for an unrecoverable error, 0 is returned and connection is killed.
 * - a transient error is handled differently if connection has its owned
 *   socket. If this is the case, 0 is returned and socket is subscribed on the
 *   poller. The other case is assimilated to a success case with 1 returned.
 *   Remaining data are purged from the buffer and will eventually be detected
 *   as lost which gives the opportunity to retry sending.
 */
static int qc_send_ppkts(struct buffer *buf, struct ssl_sock_ctx *ctx)
{
	int ret = 0;
	struct quic_conn *qc;
	char skip_sendto = 0;

	qc = ctx->qc;
	TRACE_ENTER(QUIC_EV_CONN_SPPKTS, qc);
	while (b_contig_data(buf, 0)) {
		unsigned char *pos;
		struct buffer tmpbuf = { };
		struct quic_tx_packet *first_pkt, *pkt, *next_pkt;
		uint16_t dglen;
		size_t headlen = sizeof dglen + sizeof first_pkt;
		unsigned int time_sent;

		pos = (unsigned char *)b_head(buf);
		dglen = read_u16(pos);
		BUG_ON_HOT(!dglen); /* this should not happen */

		pos += sizeof dglen;
		first_pkt = read_ptr(pos);
		pos += sizeof first_pkt;
		tmpbuf.area = (char *)pos;
		tmpbuf.size = tmpbuf.data = dglen;

		TRACE_PROTO("TX dgram", QUIC_EV_CONN_SPPKTS, qc);
		/* If sendto is on error just skip the call to it for the rest
		 * of the loop but continue to purge the buffer. Data will be
		 * transmitted when QUIC packets are detected as lost on our
		 * side.
		 *
		 * TODO use fd-monitoring to detect when send operation can be
		 * retry. This should improve the bandwidth without relying on
		 * retransmission timer. However, it requires a major rework on
		 * quic-conn fd management.
		 */
		if (!skip_sendto) {
			int ret = qc_snd_buf(qc, &tmpbuf, tmpbuf.data, 0);
			if (ret < 0) {
				TRACE_ERROR("sendto fatal error", QUIC_EV_CONN_SPPKTS, qc, first_pkt);
				qc_kill_conn(qc);
				qc_free_tx_coalesced_pkts(qc, first_pkt);
				b_del(buf, dglen + headlen);
				qc_purge_tx_buf(qc, buf);
				goto leave;
			}
			else if (!ret) {
				/* Connection owned socket : poller will wake us up when transient error is cleared. */
				if (qc_test_fd(qc)) {
					TRACE_ERROR("sendto error, subscribe to poller", QUIC_EV_CONN_SPPKTS, qc);
					goto leave;
				}

				/* No connection owned-socket : rely on retransmission to retry sending. */
				skip_sendto = 1;
				TRACE_ERROR("sendto error, simulate sending for the rest of data", QUIC_EV_CONN_SPPKTS, qc);
			}
		}

		b_del(buf, dglen + headlen);
		qc->bytes.tx += tmpbuf.data;
		time_sent = now_ms;

		for (pkt = first_pkt; pkt; pkt = next_pkt) {
			struct quic_cc *cc = &qc->path->cc;
			/* RFC 9000 14.1 Initial datagram size
			 * a server MUST expand the payload of all UDP datagrams carrying ack-eliciting
			 * Initial packets to at least the smallest allowed maximum datagram size of
			 * 1200 bytes.
			 */
			qc->cntrs.sent_pkt++;
			BUG_ON_HOT(pkt->type == QUIC_PACKET_TYPE_INITIAL &&
			           (pkt->flags & QUIC_FL_TX_PACKET_ACK_ELICITING) &&
			           dglen < QUIC_INITIAL_PACKET_MINLEN);

			pkt->time_sent = time_sent;
			if (pkt->flags & QUIC_FL_TX_PACKET_ACK_ELICITING) {
				pkt->pktns->tx.time_of_last_eliciting = time_sent;
				qc->path->ifae_pkts++;
				if (qc->flags & QUIC_FL_CONN_IDLE_TIMER_RESTARTED_AFTER_READ)
					qc_idle_timer_rearm(qc, 0, 0);
			}
			if (!(qc->flags & QUIC_FL_CONN_CLOSING) &&
			    (pkt->flags & QUIC_FL_TX_PACKET_CC)) {
				qc->flags |= QUIC_FL_CONN_CLOSING;
				qc_detach_th_ctx_list(qc, 1);

				/* RFC 9000 10.2. Immediate Close:
				 * The closing and draining connection states exist to ensure
				 * that connections close cleanly and that delayed or reordered
				 * packets are properly discarded. These states SHOULD persist
				 * for at least three times the current PTO interval...
				 *
				 * Rearm the idle timeout only one time when entering closing
				 * state.
				 */
				qc_idle_timer_do_rearm(qc, 0);
				if (qc->timer_task) {
					task_destroy(qc->timer_task);
					qc->timer_task = NULL;
				}
			}
			qc->path->in_flight += pkt->in_flight_len;
			pkt->pktns->tx.in_flight += pkt->in_flight_len;
			if ((global.tune.options & GTUNE_QUIC_CC_HYSTART) && pkt->pktns == qc->apktns)
				cc->algo->hystart_start_round(cc, pkt->pn_node.key);
			if (pkt->in_flight_len)
				qc_set_timer(qc);
			TRACE_PROTO("TX pkt", QUIC_EV_CONN_SPPKTS, qc, pkt);
			next_pkt = pkt->next;
			quic_tx_packet_refinc(pkt);
			eb64_insert(&pkt->pktns->tx.pkts, &pkt->pn_node);
		}
	}

	ret = 1;
leave:
	TRACE_LEAVE(QUIC_EV_CONN_SPPKTS, qc);

	return ret;
}

/* Flush txbuf for <qc> connection. This must be called prior to a packet
 * preparation when txbuf contains older data. A send will be conducted for
 * these data.
 *
 * Returns 1 on success : buffer is empty and can be use for packet
 * preparation. On error 0 is returned.
 */
int qc_purge_txbuf(struct quic_conn *qc, struct buffer *buf)
{
	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	/* This operation can only be conducted if txbuf is not empty. This
	 * case only happens for connection with their owned socket due to an
	 * older transient sendto() error.
	 */
	BUG_ON(!qc_test_fd(qc));

	if (b_data(buf) && !qc_send_ppkts(buf, qc->xprt_ctx)) {
		if (qc->flags & QUIC_FL_CONN_TO_KILL)
			qc_txb_release(qc);
		TRACE_DEVEL("leaving in error", QUIC_EV_CONN_TXPKT, qc);
		return 0;
	}

	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return 1;
}

/* Try to send application frames from list <frms> on connection <qc>. This
 * function is provided for MUX upper layer usage only.
 *
 * Returns the result from qc_send() function.
 */
int qc_send_mux(struct quic_conn *qc, struct list *frms)
{
	struct list send_list = LIST_HEAD_INIT(send_list);
	int ret;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);
	BUG_ON(qc->mux_state != QC_MUX_READY); /* Only MUX can uses this function so it must be ready. */

	if (qc->conn->flags & CO_FL_SOCK_WR_SH) {
		qc->conn->flags |= CO_FL_ERROR | CO_FL_SOCK_RD_SH;
		TRACE_DEVEL("connection on error", QUIC_EV_CONN_TXPKT, qc);
		return 0;
	}

	/* Try to send post handshake frames first unless on 0-RTT. */
	if ((qc->flags & QUIC_FL_CONN_NEED_POST_HANDSHAKE_FRMS) &&
	    qc->state >= QUIC_HS_ST_COMPLETE) {
		quic_build_post_handshake_frames(qc);
		qel_register_send(&send_list, qc->ael, &qc->ael->pktns->tx.frms);
		qc_send(qc, 0, &send_list);
	}

	TRACE_STATE("preparing data (from MUX)", QUIC_EV_CONN_TXPKT, qc);
	qc->flags |= QUIC_FL_CONN_TX_MUX_CONTEXT;
	qel_register_send(&send_list, qc->ael, frms);
	ret = qc_send(qc, 0, &send_list);
	qc->flags &= ~QUIC_FL_CONN_TX_MUX_CONTEXT;

	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return ret;
}

/* Select <*tls_ctx> and <*ver> for the encryption level <qel> of <qc> QUIC
 * connection, depending on its state, especially the negotiated version.
 */
static inline void qc_select_tls_ver(struct quic_conn *qc,
                                     struct quic_enc_level *qel,
                                     struct quic_tls_ctx **tls_ctx,
                                     const struct quic_version **ver)
{
	if (qc->negotiated_version) {
		*ver = qc->negotiated_version;
		if (qel == qc->iel)
			*tls_ctx = qc->nictx;
		else
			*tls_ctx = &qel->tls_ctx;
	}
	else {
		*ver = qc->original_version;
		*tls_ctx = &qel->tls_ctx;
	}
}

/* Prepare as much as possible QUIC datagrams/packets for sending from <qels>
 * list of encryption levels. Several packets can be coalesced into a single
 * datagram. The result is written into <buf>.
 *
 * Each datagram is prepended by a two fields header : the datagram length and
 * the address of first packet in the datagram.
 *
 * Returns the number of bytes prepared in datragrams/packets if succeeded
 * (may be 0), or -1 if something wrong happened.
 */
static int qc_prep_pkts(struct quic_conn *qc, struct buffer *buf,
                         struct list *qels)
{
	int ret, cc, padding;
	struct quic_tx_packet *first_pkt, *prv_pkt;
	unsigned char *end, *pos;
	uint16_t dglen;
	size_t total;
	struct quic_enc_level *qel, *tmp_qel;

	TRACE_ENTER(QUIC_EV_CONN_IO_CB, qc);
	/* Currently qc_prep_pkts() does not handle buffer wrapping so the
	 * caller must ensure that buf is reset.
	 */
	BUG_ON_HOT(buf->head || buf->data);

	ret = -1;
	cc =  qc->flags & QUIC_FL_CONN_IMMEDIATE_CLOSE;
	padding = 0;
	first_pkt = prv_pkt = NULL;
	end = pos = (unsigned char *)b_head(buf);
	dglen = 0;
	total = 0;

	list_for_each_entry_safe(qel, tmp_qel, qels, el_send) {
		struct quic_tls_ctx *tls_ctx;
		const struct quic_version *ver;
		struct list *frms = qel->send_frms, *next_frms;
		struct quic_enc_level *next_qel;

		if (qel == qc->eel) {
			/* Next encryption level */
			continue;
		}

		qc_select_tls_ver(qc, qel, &tls_ctx, &ver);

		/* Retrieve next QEL. Set it to NULL if on qels last element. */
		if (qel->el_send.n != qels) {
			next_qel = LIST_ELEM(qel->el_send.n, struct quic_enc_level *, el_send);
			next_frms = next_qel->send_frms;
		}
		else {
			next_qel  = NULL;
			next_frms = NULL;
		}

		/* Build as much as datagrams at <qel> encryption level.
		 * Each datagram is prepended with its length followed by the address
		 * of the first packet in the datagram (QUIC_DGRAM_HEADLEN).
		 */
		while ((!cc && b_contig_space(buf) >= (int)qc->path->mtu + QUIC_DGRAM_HEADLEN) ||
		        (cc && b_contig_space(buf) >= QUIC_MIN_CC_PKTSIZE + QUIC_DGRAM_HEADLEN) || prv_pkt) {
			int err, probe, must_ack;
			enum quic_pkt_type pkt_type;
			struct quic_tx_packet *cur_pkt;

			TRACE_PROTO("TX prep pkts", QUIC_EV_CONN_PHPKTS, qc, qel);
			probe = 0;
			/* We do not probe if an immediate close was asked */
			if (!cc)
				probe = qel->pktns->tx.pto_probe;

			if (!qc_may_build_pkt(qc, frms, qel, cc, probe, &must_ack)) {
				/* Remove qel from send_list if nothing to send. */
				LIST_DEL_INIT(&qel->el_send);
				qel->send_frms = NULL;

				if (prv_pkt && !next_qel) {
					qc_txb_store(buf, dglen, first_pkt);
					/* Build only one datagram when an immediate close is required. */
					if (cc)
						goto out;
				}

				TRACE_DEVEL("next encryption level", QUIC_EV_CONN_PHPKTS, qc);
				break;
			}

			if (!prv_pkt) {
				/* Leave room for the datagram header */
				pos += QUIC_DGRAM_HEADLEN;
				if (cc) {
					end = pos + QUIC_MIN_CC_PKTSIZE;
				}
				else if (!quic_peer_validated_addr(qc) && qc_is_listener(qc)) {
					end = pos + QUIC_MIN(qc->path->mtu, quic_may_send_bytes(qc));
				}
				else {
					end = pos + qc->path->mtu;
				}
			}

			/* RFC 9000 14.1 Initial datagram size
			 * a server MUST expand the payload of all UDP datagrams carrying ack-eliciting
			 * Initial packets to at least the smallest allowed maximum datagram size of
			 * 1200 bytes.
			 *
			 * Ensure that no ack-eliciting packets are sent into too small datagrams
			 */
			if (qel == qc->iel && !LIST_ISEMPTY(frms)) {
				if (end - pos < QUIC_INITIAL_PACKET_MINLEN) {
					TRACE_PROTO("No more enough room to build an Initial packet",
					            QUIC_EV_CONN_PHPKTS, qc);
					break;
				}

				/* Pad this Initial packet if there is no ack-eliciting frames to send from
				 * the next packet number space.
				 */
				if (!next_frms || LIST_ISEMPTY(next_frms))
					padding = 1;
			}

			pkt_type = quic_enc_level_pkt_type(qc, qel);
			cur_pkt = qc_build_pkt(&pos, end, qel, tls_ctx, frms,
			                       qc, ver, dglen, pkt_type,
			                       must_ack, padding, probe, cc, &err);
			switch (err) {
				case -3:
					if (first_pkt)
						qc_txb_store(buf, dglen, first_pkt);
					qc_purge_tx_buf(qc, buf);
					goto leave;
				case -2:
					// trace already emitted by function above
					goto leave;
				case -1:
					/* If there was already a correct packet present, set the
					 * current datagram as prepared into <cbuf>.
					 */
					if (prv_pkt)
						qc_txb_store(buf, dglen, first_pkt);
					TRACE_PROTO("could not prepare anymore packet", QUIC_EV_CONN_PHPKTS, qc, qel);
					goto out;
				default:
					break;
			}

			/* This is to please to GCC. We cannot have (err >= 0 && !cur_pkt) */
			BUG_ON(!cur_pkt);

			total += cur_pkt->len;
			dglen += cur_pkt->len;

			if (qc->flags & QUIC_FL_CONN_RETRANS_OLD_DATA)
				cur_pkt->flags |= QUIC_FL_TX_PACKET_PROBE_WITH_OLD_DATA;

			/* keep trace of the first packet in the datagram */
			if (!first_pkt)
				first_pkt = cur_pkt;

			/* Attach the current one to the previous one and vice versa */
			if (prv_pkt) {
				prv_pkt->next = cur_pkt;
				cur_pkt->prev = prv_pkt;
				cur_pkt->flags |= QUIC_FL_TX_PACKET_COALESCED;
			}

			/* If there is no more packet to build for this encryption level,
			 * select the next one <next_qel>, if any, to coalesce a packet in
			 * the same datagram, except if <qel> is the Application data
			 * encryption level which cannot be selected to do that.
			 */
			if (LIST_ISEMPTY(frms) && qel != qc->ael && next_qel) {
				if (qel == qc->iel &&
				    (!qc_is_listener(qc) ||
				     cur_pkt->flags & QUIC_FL_TX_PACKET_ACK_ELICITING))
					padding = 1;

				prv_pkt = cur_pkt;
			}
			else {
				qc_txb_store(buf, dglen, first_pkt);
				/* Build only one datagram when an immediate close is required. */
				if (cc)
					goto out;
				first_pkt = NULL;
				dglen = 0;
				padding = 0;
				prv_pkt = NULL;
			}
		}
	}

 out:
	if (cc && total) {
		BUG_ON(buf != &qc->tx.cc_buf);
		BUG_ON(dglen != total);
		qc->tx.cc_dgram_len = dglen;
	}

	ret = total;
 leave:
	TRACE_LEAVE(QUIC_EV_CONN_PHPKTS, qc);
	return ret;
}

/* Encode frames and send them as packets for <qc> connection. Input frames are
 * specified via quic_enc_level <send_list> through their send_frms member. Set
 * <old_data> when reemitted duplicated data.
 *
* Returns 1 on success else 0. Note that <send_list> will always be reset
* after qc_send() exit.
 */
int qc_send(struct quic_conn *qc, int old_data, struct list *send_list)
{
	struct quic_enc_level *qel, *tmp_qel;
	int ret = 0, status = 0;
	struct buffer *buf;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	buf = qc_get_txb(qc);
	if (!buf) {
		TRACE_ERROR("buffer allocation failed", QUIC_EV_CONN_TXPKT, qc);
		goto out;
	}

	if (b_data(buf) && !qc_purge_txbuf(qc, buf)) {
		TRACE_ERROR("Could not purge TX buffer", QUIC_EV_CONN_TXPKT, qc);
		goto out;
	}

	if (old_data) {
		TRACE_STATE("old data for probing asked", QUIC_EV_CONN_TXPKT, qc);
		qc->flags |= QUIC_FL_CONN_RETRANS_OLD_DATA;
	}

	/* Prepare and send packets until we could not further prepare packets. */
	while (!LIST_ISEMPTY(send_list)) {
		/* Buffer must always be empty before qc_prep_pkts() usage.
		 * qc_send_ppkts() ensures it is cleared on success.
		 */
		BUG_ON_HOT(b_data(buf));
		b_reset(buf);

		ret = qc_prep_pkts(qc, buf, send_list);

		if (b_data(buf) && !qc_send_ppkts(buf, qc->xprt_ctx)) {
			if (qc->flags & QUIC_FL_CONN_TO_KILL)
				qc_txb_release(qc);
			goto out;
		}

		if (ret <= 0) {
			TRACE_DEVEL("stopping on qc_prep_pkts() return", QUIC_EV_CONN_TXPKT, qc);
			break;
		}
	}

	qc_txb_release(qc);
	if (ret < 0)
		goto out;

	status = 1;

 out:
	if (old_data) {
		TRACE_STATE("no more need old data for probing", QUIC_EV_CONN_TXPKT, qc);
		qc->flags &= ~QUIC_FL_CONN_RETRANS_OLD_DATA;
	}

	/* Always reset QEL sending list. */
	list_for_each_entry_safe(qel, tmp_qel, send_list, el_send) {
		LIST_DEL_INIT(&qel->el_send);
		qel->send_frms = NULL;
	}

	TRACE_DEVEL((status ? "leaving" : "leaving in error"), QUIC_EV_CONN_TXPKT, qc);
	return status;
}

/* Insert <qel> into <send_list> in preparation for sending. Set its send
 * frames list pointer to <frms>.
 */
void qel_register_send(struct list *send_list, struct quic_enc_level *qel,
                       struct list *frms)
{
	/* Ensure QEL is not already registered for sending. */
	BUG_ON(LIST_INLIST(&qel->el_send));

	LIST_APPEND(send_list, &qel->el_send);
	qel->send_frms = frms;
}

/* Returns true if <qel> should be registered for sending. This is the case if
 * frames are prepared, probing is set, <qc> ACK timer has fired or a
 * CONNECTION_CLOSE is required.
 */
int qel_need_sending(struct quic_enc_level *qel, struct quic_conn *qc)
{
	return !LIST_ISEMPTY(&qel->pktns->tx.frms) ||
	       qel->pktns->tx.pto_probe ||
	       (qel->pktns->flags & QUIC_FL_PKTNS_ACK_REQUIRED) ||
	       (qc->flags & (QUIC_FL_CONN_ACK_TIMER_FIRED|QUIC_FL_CONN_IMMEDIATE_CLOSE));
}

/* Retransmit up to two datagrams depending on packet number space.
 * Return 0 when failed, 0 if not.
 */
int qc_dgrams_retransmit(struct quic_conn *qc)
{
	int ret = 0;
	int sret;
	struct quic_pktns *ipktns = qc->ipktns;
	struct quic_pktns *hpktns = qc->hpktns;
	struct quic_pktns *apktns = qc->apktns;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	/* Note that if the Initial packet number space is not discarded,
	 * this is also the case for the Handshake packet number space.
	 */
	if (ipktns && (ipktns->flags & QUIC_FL_PKTNS_PROBE_NEEDED)) {
		int i;

		for (i = 0; i < QUIC_MAX_NB_PTO_DGRAMS; i++) {
			struct list send_list = LIST_HEAD_INIT(send_list);
			struct list ifrms = LIST_HEAD_INIT(ifrms);
			struct list hfrms = LIST_HEAD_INIT(hfrms);

			qc_prep_hdshk_fast_retrans(qc, &ifrms, &hfrms);
			TRACE_DEVEL("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, &ifrms);
			TRACE_DEVEL("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, &hfrms);
			if (!LIST_ISEMPTY(&ifrms)) {
				ipktns->tx.pto_probe = 1;
				if (!LIST_ISEMPTY(&hfrms))
					hpktns->tx.pto_probe = 1;

				qel_register_send(&send_list, qc->iel, &ifrms);
				if (qc->hel)
					qel_register_send(&send_list, qc->hel, &hfrms);

				sret = qc_send(qc, 1, &send_list);
				qc_free_frm_list(qc, &ifrms);
				qc_free_frm_list(qc, &hfrms);
				if (!sret)
					goto leave;
			}
			else {
				/* No frame to send due to amplification limit
				 * or allocation failure. A PING frame will be
				 * emitted for probing.
				 */
				ipktns->tx.pto_probe = 1;
				qel_register_send(&send_list, qc->iel, &ifrms);
				sret = qc_send(qc, 0, &send_list);
				qc_free_frm_list(qc, &ifrms);
				qc_free_frm_list(qc, &hfrms);
				if (!sret)
					goto leave;

				break;
			}
		}
		TRACE_STATE("no more need to probe Initial packet number space",
					QUIC_EV_CONN_TXPKT, qc);
		ipktns->flags &= ~QUIC_FL_PKTNS_PROBE_NEEDED;
		if (hpktns)
			hpktns->flags &= ~QUIC_FL_PKTNS_PROBE_NEEDED;
	}
	else {
		int i;

		if (hpktns && (hpktns->flags & QUIC_FL_PKTNS_PROBE_NEEDED)) {
			hpktns->tx.pto_probe = 0;
			for (i = 0; i < QUIC_MAX_NB_PTO_DGRAMS; i++) {
				struct list send_list = LIST_HEAD_INIT(send_list);
				struct list frms1 = LIST_HEAD_INIT(frms1);

				qc_prep_fast_retrans(qc, hpktns, &frms1, NULL);
				TRACE_DEVEL("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, &frms1);
				if (!LIST_ISEMPTY(&frms1)) {
					hpktns->tx.pto_probe = 1;
					qel_register_send(&send_list, qc->hel, &frms1);
					sret = qc_send(qc, 1, &send_list);
					qc_free_frm_list(qc, &frms1);
					if (!sret)
						goto leave;
				}
			}
			TRACE_STATE("no more need to probe Handshake packet number space",
			            QUIC_EV_CONN_TXPKT, qc);
			hpktns->flags &= ~QUIC_FL_PKTNS_PROBE_NEEDED;
		}
		else if (apktns && (apktns->flags & QUIC_FL_PKTNS_PROBE_NEEDED)) {
			struct list send_list = LIST_HEAD_INIT(send_list);
			struct list frms2 = LIST_HEAD_INIT(frms2);
			struct list frms1 = LIST_HEAD_INIT(frms1);

			apktns->tx.pto_probe = 0;
			qc_prep_fast_retrans(qc, apktns, &frms1, &frms2);
			TRACE_PROTO("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, &frms1);
			TRACE_PROTO("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, &frms2);

			if (!LIST_ISEMPTY(&frms1)) {
				apktns->tx.pto_probe = 1;
				qel_register_send(&send_list, qc->ael, &frms1);
				sret = qc_send(qc, 1, &send_list);
				qc_free_frm_list(qc, &frms1);
				if (!sret) {
					qc_free_frm_list(qc, &frms2);
					goto leave;
				}
			}

			if (!LIST_ISEMPTY(&frms2)) {
				apktns->tx.pto_probe = 1;
				qel_register_send(&send_list, qc->ael, &frms2);
				sret = qc_send(qc, 1, &send_list);
				qc_free_frm_list(qc, &frms2);
				if (!sret)
					goto leave;
			}
			TRACE_STATE("no more need to probe 01RTT packet number space",
			            QUIC_EV_CONN_TXPKT, qc);
			apktns->flags &= ~QUIC_FL_PKTNS_PROBE_NEEDED;
		}
	}

	ret = 1;
 leave:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return ret;
}

/*
 * Send a Version Negotiation packet on response to <pkt> on socket <fd> to
 * address <addr>.
 * Implementation of RFC9000 6. Version Negotiation
 *
 * TODO implement a rate-limiting sending of Version Negotiation packets
 *
 * Returns 0 on success else non-zero
 */
int send_version_negotiation(int fd, struct sockaddr_storage *addr,
                             struct quic_rx_packet *pkt)
{
	char buf[256];
	int ret = 0, i = 0, j;
	uint32_t version;
	const socklen_t addrlen = get_addr_len(addr);

	TRACE_ENTER(QUIC_EV_CONN_TXPKT);
	/*
	 * header form
	 * long header, fixed bit to 0 for Version Negotiation
	 */
	/* TODO: RAND_bytes() should be replaced? */
	if (RAND_bytes((unsigned char *)buf, 1) != 1) {
		TRACE_ERROR("RAND_bytes() error", QUIC_EV_CONN_TXPKT);
		goto out;
	}

	buf[i++] |= '\x80';
	/* null version for Version Negotiation */
	buf[i++] = '\x00';
	buf[i++] = '\x00';
	buf[i++] = '\x00';
	buf[i++] = '\x00';

	/* source connection id */
	buf[i++] = pkt->scid.len;
	memcpy(&buf[i], pkt->scid.data, pkt->scid.len);
	i += pkt->scid.len;

	/* destination connection id */
	buf[i++] = pkt->dcid.len;
	memcpy(&buf[i], pkt->dcid.data, pkt->dcid.len);
	i += pkt->dcid.len;

	/* supported version */
	for (j = 0; j < quic_versions_nb; j++) {
		version = htonl(quic_versions[j].num);
		memcpy(&buf[i], &version, sizeof(version));
		i += sizeof(version);
	}

	if (sendto(fd, buf, i, 0, (struct sockaddr *)addr, addrlen) < 0)
		goto out;

	ret = 1;
 out:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT);
	return !ret;
}

/* Send a stateless reset packet depending on <pkt> RX packet information
 * from <fd> UDP socket to <dst>
 * Return 1 if succeeded, 0 if not.
 */
int send_stateless_reset(struct listener *l, struct sockaddr_storage *dstaddr,
                         struct quic_rx_packet *rxpkt)
{
	int ret = 0, pktlen, rndlen;
	unsigned char pkt[64];
	const socklen_t addrlen = get_addr_len(dstaddr);
	struct proxy *prx;
	struct quic_counters *prx_counters;

	TRACE_ENTER(QUIC_EV_STATELESS_RST);

	/* RFC 9000 10.3. Stateless Reset
	 *
	 * Endpoints MUST discard packets that are too small to be valid QUIC
	 * packets. To give an example, with the set of AEAD functions defined
	 * in [QUIC-TLS], short header packets that are smaller than 21 bytes
	 * are never valid.
	 *
	 * [...]
	 *
	 * RFC 9000 10.3.3. Looping
	 *
	 * An endpoint MUST ensure that every Stateless Reset that it sends is
	 * smaller than the packet that triggered it, unless it maintains state
	 * sufficient to prevent looping. In the event of a loop, this results
	 * in packets eventually being too small to trigger a response.
	 */
	if (rxpkt->len <= QUIC_STATELESS_RESET_PACKET_MINLEN) {
		TRACE_DEVEL("rxpkt too short", QUIC_EV_STATELESS_RST);
		goto leave;
	}

	prx = l->bind_conf->frontend;
	prx_counters = EXTRA_COUNTERS_GET(prx->extra_counters_fe, &quic_stats_module);

	/* RFC 9000 10.3. Stateless Reset
	 *
	 * An endpoint that sends a Stateless Reset in response to a packet that is
	 * 43 bytes or shorter SHOULD send a Stateless Reset that is one byte shorter
	 * than the packet it responds to.
	 */
	pktlen = rxpkt->len <= 43 ? rxpkt->len - 1 :
	                            QUIC_STATELESS_RESET_PACKET_MINLEN;
	rndlen = pktlen - QUIC_STATELESS_RESET_TOKEN_LEN;

	/* Put a header of random bytes */
	/* TODO: RAND_bytes() should be replaced */
	if (RAND_bytes(pkt, rndlen) != 1) {
		TRACE_ERROR("RAND_bytes() failed", QUIC_EV_STATELESS_RST);
		goto leave;
	}

	/* Clear the most significant bit, and set the second one */
	*pkt = (*pkt & ~0x80) | 0x40;
	if (!quic_stateless_reset_token_cpy(pkt + rndlen, QUIC_STATELESS_RESET_TOKEN_LEN,
	                                    rxpkt->dcid.data, rxpkt->dcid.len))
		goto leave;

	if (sendto(l->rx.fd, pkt, pktlen, 0, (struct sockaddr *)dstaddr, addrlen) < 0)
		goto leave;

    ret = 1;
	HA_ATOMIC_INC(&prx_counters->stateless_reset_sent);
	TRACE_PROTO("stateless reset sent", QUIC_EV_STATELESS_RST, NULL, &rxpkt->dcid);
 leave:
	TRACE_LEAVE(QUIC_EV_STATELESS_RST);
	return ret;
}

/* Return the long packet type matching with <qv> version and <type> */
static inline int quic_pkt_type(int type, uint32_t version)
{
	if (version != QUIC_PROTOCOL_VERSION_2)
		return type;

	switch (type) {
	case QUIC_PACKET_TYPE_INITIAL:
		return 1;
	case QUIC_PACKET_TYPE_0RTT:
		return 2;
	case QUIC_PACKET_TYPE_HANDSHAKE:
		return 3;
	case QUIC_PACKET_TYPE_RETRY:
		return 0;
	}

	return -1;
}


/* Generate a Retry packet and send it on <fd> socket to <addr> in response to
 * the Initial <pkt> packet.
 *
 * Returns 0 on success else non-zero.
 */
int send_retry(int fd, struct sockaddr_storage *addr,
               struct quic_rx_packet *pkt, const struct quic_version *qv)
{
	int ret = 0;
	unsigned char buf[128];
	int i = 0, token_len;
	const socklen_t addrlen = get_addr_len(addr);
	struct quic_cid scid;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT);

	/* long header(1) | fixed bit(1) | packet type QUIC_PACKET_TYPE_RETRY(2) | unused random bits(4)*/
	buf[i++] = (QUIC_PACKET_LONG_HEADER_BIT | QUIC_PACKET_FIXED_BIT) |
		(quic_pkt_type(QUIC_PACKET_TYPE_RETRY, qv->num) << QUIC_PACKET_TYPE_SHIFT) |
		statistical_prng_range(16);
	/* version */
	write_n32(&buf[i], qv->num);
	i += sizeof(uint32_t);

	/* Use the SCID from <pkt> for Retry DCID. */
	buf[i++] = pkt->scid.len;
	memcpy(&buf[i], pkt->scid.data, pkt->scid.len);
	i += pkt->scid.len;

	/* Generate a new CID to be used as SCID for the Retry packet. */
	scid.len = QUIC_HAP_CID_LEN;
	/* TODO: RAND_bytes() should be replaced */
	if (RAND_bytes(scid.data, scid.len) != 1) {
		TRACE_ERROR("RAND_bytes() failed", QUIC_EV_CONN_TXPKT);
		goto out;
	}

	buf[i++] = scid.len;
	memcpy(&buf[i], scid.data, scid.len);
	i += scid.len;

	/* token */
	if (!(token_len = quic_generate_retry_token(&buf[i], sizeof(buf) - i, qv->num,
	                                            &pkt->dcid, &pkt->scid, addr))) {
		TRACE_ERROR("quic_generate_retry_token() failed", QUIC_EV_CONN_TXPKT);
		goto out;
	}

	i += token_len;

	/* token integrity tag */
	if ((sizeof(buf) - i < QUIC_TLS_TAG_LEN) ||
	    !quic_tls_generate_retry_integrity_tag(pkt->dcid.data,
	                                           pkt->dcid.len, buf, i, qv)) {
		TRACE_ERROR("quic_tls_generate_retry_integrity_tag() failed", QUIC_EV_CONN_TXPKT);
		goto out;
	}

	i += QUIC_TLS_TAG_LEN;

	if (sendto(fd, buf, i, 0, (struct sockaddr *)addr, addrlen) < 0) {
		TRACE_ERROR("quic_tls_generate_retry_integrity_tag() failed", QUIC_EV_CONN_TXPKT);
		goto out;
	}

	ret = 1;
 out:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT);
	return !ret;
}

/* Write a 32-bits integer to a buffer with <buf> as address.
 * Make <buf> point to the data after this 32-buts value if succeeded.
 * Note that these 32-bits integers are networkg bytes ordered.
 * Returns 0 if failed (not enough room in the buffer), 1 if succeeded.
 */
static inline int quic_write_uint32(unsigned char **buf,
                                    const unsigned char *end, uint32_t val)
{
	if (end - *buf < sizeof val)
		return 0;

	write_u32(*buf, htonl(val));
	*buf += sizeof val;

	return 1;
}

/* Return the maximum number of bytes we must use to completely fill a
 * buffer with <sz> as size for a data field of bytes prefixed by its QUIC
 * variable-length (may be 0).
 * Also put in <*len_sz> the size of this QUIC variable-length.
 * So after returning from this function we have : <*len_sz> + <ret> <= <sz>
 * (<*len_sz> = { max(i), i + ret <= <sz> }) .
 */
static inline size_t max_available_room(size_t sz, size_t *len_sz)
{
	size_t sz_sz, ret;
	size_t diff;

	sz_sz = quic_int_getsize(sz);
	if (sz <= sz_sz)
		return 0;

	ret = sz - sz_sz;
	*len_sz = quic_int_getsize(ret);
	/* Difference between the two sizes. Note that <sz_sz> >= <*len_sz>. */
	diff = sz_sz - *len_sz;
	if (unlikely(diff > 0)) {
		/* Let's try to take into an account remaining bytes.
		 *
		 *                  <----------------> <sz_sz>
		 *  <--------------><-------->  +----> <max_int>
		 *       <ret>       <len_sz>   |
		 *  +---------------------------+-----------....
		 *  <--------------------------------> <sz>
		 */
		size_t max_int = quic_max_int(*len_sz);

		if (max_int + *len_sz <= sz)
			ret = max_int;
		else
			ret = sz - diff;
	}

	return ret;
}

/* This function computes the maximum data we can put into a buffer with <sz> as
 * size prefixed with a variable-length field "Length" whose value is the
 * remaining data length, already filled of <ilen> bytes which must be taken
 * into an account by "Length" field, and finally followed by the data we want
 * to put in this buffer prefixed again by a variable-length field.
 * <sz> is the size of the buffer to fill.
 * <ilen> the number of bytes already put after the "Length" field.
 * <dlen> the number of bytes we want to at most put in the buffer.
 * Also set <*dlen_sz> to the size of the data variable-length we want to put in
 * the buffer. This is typically this function which must be used to fill as
 * much as possible a QUIC packet made of only one CRYPTO or STREAM frames.
 * Returns this computed size if there is enough room in the buffer, 0 if not.
 */
static inline size_t max_stream_data_size(size_t sz, size_t ilen, size_t dlen)
{
	size_t ret, len_sz, dlen_sz;

	/*
	 * The length of variable-length QUIC integers are powers of two.
	 * Look for the first 3length" field value <len_sz> which match our need.
	 * As we must put <ilen> bytes in our buffer, the minimum value for
	 * <len_sz> is the number of bytes required to encode <ilen>.
	 */
	for (len_sz = quic_int_getsize(ilen);
	     len_sz <= QUIC_VARINT_MAX_SIZE;
	     len_sz <<= 1) {
		if (sz < len_sz + ilen)
			return 0;

		ret = max_available_room(sz - len_sz - ilen, &dlen_sz);
		if (!ret)
			return 0;

		/* Check that <*len_sz> matches <ret> value */
		if (len_sz + ilen + dlen_sz + ret <= quic_max_int(len_sz))
			return ret < dlen ? ret : dlen;
	}

	return 0;
}

/* Return the length in bytes of <pn> packet number depending on
 * <largest_acked_pn> the largest ackownledged packet number.
 */
static inline size_t quic_packet_number_length(int64_t pn,
                                               int64_t largest_acked_pn)
{
	int64_t max_nack_pkts;

	/* About packet number encoding, the RFC says:
	 * The sender MUST use a packet number size able to represent more than
	 * twice as large a range than the difference between the largest
	 * acknowledged packet and packet number being sent.
	 */
	max_nack_pkts = 2 * (pn - largest_acked_pn) + 1;
	if (max_nack_pkts > 0xffffff)
		return 4;
	if (max_nack_pkts > 0xffff)
		return 3;
	if (max_nack_pkts > 0xff)
		return 2;

	return 1;
}

/* Encode <pn> packet number with <pn_len> as length in byte into a buffer with
 * <buf> as current copy address and <end> as pointer to one past the end of
 * this buffer. This is the responsibility of the caller to check there is
 * enough room in the buffer to copy <pn_len> bytes.
 * Never fails.
 */
static inline int quic_packet_number_encode(unsigned char **buf,
                                            const unsigned char *end,
                                            uint64_t pn, size_t pn_len)
{
	if (end - *buf < pn_len)
		return 0;

	/* Encode the packet number. */
	switch (pn_len) {
	case 1:
		**buf = pn;
		break;
	case 2:
		write_n16(*buf, pn);
		break;
	case 3:
		(*buf)[0] = pn >> 16;
		(*buf)[1] = pn >> 8;
		(*buf)[2] = pn;
		break;
	case 4:
		write_n32(*buf, pn);
		break;
	}
	*buf += pn_len;

	return 1;
}

/* This function builds into a buffer at <pos> position a QUIC long packet header,
 * <end> being one byte past the end of this buffer.
 * Return 1 if enough room to build this header, 0 if not.
 */
static int quic_build_packet_long_header(unsigned char **pos, const unsigned char *end,
                                         int type, size_t pn_len,
                                         struct quic_conn *qc, const struct quic_version *ver)
{
	int ret = 0;

	TRACE_ENTER(QUIC_EV_CONN_LPKT, qc);

	if (end - *pos < sizeof ver->num + qc->dcid.len + qc->scid.len + 3) {
		TRACE_DEVEL("not enough room", QUIC_EV_CONN_LPKT, qc);
		goto leave;
	}

	type = quic_pkt_type(type, ver->num);
	/* #0 byte flags */
	*(*pos)++ = QUIC_PACKET_FIXED_BIT | QUIC_PACKET_LONG_HEADER_BIT |
		(type << QUIC_PACKET_TYPE_SHIFT) | (pn_len - 1);
	/* Version */
	quic_write_uint32(pos, end, ver->num);
	*(*pos)++ = qc->dcid.len;
	/* Destination connection ID */
	if (qc->dcid.len) {
		memcpy(*pos, qc->dcid.data, qc->dcid.len);
		*pos += qc->dcid.len;
	}
	/* Source connection ID */
	*(*pos)++ = qc->scid.len;
	if (qc->scid.len) {
		memcpy(*pos, qc->scid.data, qc->scid.len);
		*pos += qc->scid.len;
	}

	ret = 1;
 leave:
	TRACE_LEAVE(QUIC_EV_CONN_LPKT, qc);
	return ret;
}

/* This function builds into a buffer at <pos> position a QUIC short packet header,
 * <end> being one byte past the end of this buffer.
 * Return 1 if enough room to build this header, 0 if not.
 */
static int quic_build_packet_short_header(unsigned char **pos, const unsigned char *end,
                                          size_t pn_len, struct quic_conn *qc,
                                          unsigned char tls_flags)
{
	int ret = 0;
	unsigned char spin_bit =
		(qc->flags & QUIC_FL_CONN_SPIN_BIT) ? QUIC_PACKET_SPIN_BIT : 0;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	if (end - *pos < 1 + qc->dcid.len) {
		TRACE_DEVEL("not enough room", QUIC_EV_CONN_LPKT, qc);
		goto leave;
	}

	/* #0 byte flags */
	*(*pos)++ = QUIC_PACKET_FIXED_BIT | spin_bit |
		((tls_flags & QUIC_FL_TLS_KP_BIT_SET) ? QUIC_PACKET_KEY_PHASE_BIT : 0) | (pn_len - 1);
	/* Destination connection ID */
	if (qc->dcid.len) {
		memcpy(*pos, qc->dcid.data, qc->dcid.len);
		*pos += qc->dcid.len;
	}

	ret = 1;
 leave:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return ret;
}

/* Apply QUIC header protection to the packet with <pos> as first byte address,
 * <pn> as address of the Packet number field, <pnlen> being this field length
 * with <aead> as AEAD cipher and <key> as secret key.
 *
 * TODO no error is expected as encryption is done in place but encryption
 * manual is unclear. <fail> will be set to true if an error is detected.
 */
void quic_apply_header_protection(struct quic_conn *qc, unsigned char *pos,
                                        unsigned char *pn, size_t pnlen,
                                        struct quic_tls_ctx *tls_ctx, int *fail)

{
	int i;
	/* We need an IV of at least 5 bytes: one byte for bytes #0
	 * and at most 4 bytes for the packet number
	 */
	unsigned char mask[5] = {0};
	EVP_CIPHER_CTX *aes_ctx = tls_ctx->tx.hp_ctx;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	*fail = 0;

	if (!quic_tls_aes_encrypt(mask, pn + QUIC_PACKET_PN_MAXLEN, sizeof mask, aes_ctx)) {
		TRACE_ERROR("could not apply header protection", QUIC_EV_CONN_TXPKT, qc);
		*fail = 1;
		goto out;
	}

	*pos ^= mask[0] & (*pos & QUIC_PACKET_LONG_HEADER_BIT ? 0xf : 0x1f);
	for (i = 0; i < pnlen; i++)
		pn[i] ^= mask[i + 1];

 out:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
}

/* Prepare into <outlist> as most as possible ack-eliciting frame from their
 * <inlist> prebuilt frames for <qel> encryption level to be encoded in a buffer
 * with <room> as available room, and <*len> the packet Length field initialized
 * with the number of bytes already present in this buffer which must be taken
 * into an account for the Length packet field value. <headlen> is the number of
 * bytes already present in this packet before building frames.
 *
 * Update consequently <*len> to reflect the size of these frames built
 * by this function. Also attach these frames to <l> frame list.
 * Return 1 if at least one ack-eleciting frame could be built, 0 if not.
 */
static int qc_build_frms(struct list *outlist, struct list *inlist,
                         size_t room, size_t *len, size_t headlen,
                         struct quic_enc_level *qel,
                         struct quic_conn *qc)
{
	int ret;
	struct quic_frame *cf, *cfbak;

	TRACE_ENTER(QUIC_EV_CONN_BCFRMS, qc);

	ret = 0;
	if (*len > room)
		goto leave;
	room -= *len;

	/* If we are not probing we must take into an account the congestion
	 * control window.
	 */
	if (!qel->pktns->tx.pto_probe) {
		size_t remain = quic_cc_path_prep_data(qc->path);

		if (headlen > remain)
			goto leave;

		room = QUIC_MIN(room, remain - headlen);
	}

	TRACE_PROTO("TX frms build (headlen)",
	            QUIC_EV_CONN_BCFRMS, qc, &headlen);

	/* NOTE: switch/case block inside a loop, a successful status must be
	 * returned by this function only if at least one frame could be built
	 * in the switch/case block.
	 */
	list_for_each_entry_safe(cf, cfbak, inlist, list) {
		/* header length, data length, frame length. */
		size_t hlen, dlen, dlen_sz, avail_room, flen;

		if (!room)
			break;

		switch (cf->type) {
		case QUIC_FT_CRYPTO:
			TRACE_DEVEL("          New CRYPTO frame build (room, len)",
			            QUIC_EV_CONN_BCFRMS, qc, &room, len);
			/* Compute the length of this CRYPTO frame header */
			hlen = 1 + quic_int_getsize(cf->crypto.offset);
			/* Compute the data length of this CRYPTO frame. */
			dlen = max_stream_data_size(room, hlen, cf->crypto.len);
			TRACE_DEVEL(" CRYPTO data length (hlen, crypto.len, dlen)",
			            QUIC_EV_CONN_BCFRMS, qc, &hlen, &cf->crypto.len, &dlen);
			if (!dlen)
				continue;

			/* CRYPTO frame length. */
			flen = hlen + quic_int_getsize(dlen) + dlen;
			TRACE_DEVEL("                 CRYPTO frame length (flen)",
			            QUIC_EV_CONN_BCFRMS, qc, &flen);
			/* Add the CRYPTO data length and its encoded length to the packet
			 * length and the length of this length.
			 */
			*len += flen;
			room -= flen;
			if (dlen == cf->crypto.len) {
				/* <cf> CRYPTO data have been consumed. */
				LIST_DEL_INIT(&cf->list);
				LIST_APPEND(outlist, &cf->list);
			}
			else {
				struct quic_frame *new_cf;

				new_cf = qc_frm_alloc(QUIC_FT_CRYPTO);
				if (!new_cf) {
					TRACE_ERROR("No memory for new crypto frame", QUIC_EV_CONN_BCFRMS, qc);
					continue;
				}

				new_cf->crypto.len = dlen;
				new_cf->crypto.offset = cf->crypto.offset;
				new_cf->crypto.qel = qel;
				TRACE_DEVEL("split frame", QUIC_EV_CONN_PRSAFRM, qc, new_cf);
				if (cf->origin) {
					TRACE_DEVEL("duplicated frame", QUIC_EV_CONN_PRSAFRM, qc);
					/* This <cf> frame was duplicated */
					LIST_APPEND(&cf->origin->reflist, &new_cf->ref);
					new_cf->origin = cf->origin;
					/* Detach the remaining CRYPTO frame from its original frame */
					LIST_DEL_INIT(&cf->ref);
					cf->origin = NULL;
				}
				LIST_APPEND(outlist, &new_cf->list);
				/* Consume <dlen> bytes of the current frame. */
				cf->crypto.len -= dlen;
				cf->crypto.offset += dlen;
			}
			break;

		case QUIC_FT_STREAM_8 ... QUIC_FT_STREAM_F:
			if (cf->stream.dup) {
				struct eb64_node *node = NULL;
				struct qc_stream_desc *stream_desc = NULL;
				struct qf_stream *strm_frm = &cf->stream;

				/* As this frame has been already lost, ensure the stream is always
				 * available or the range of this frame is not consumed before
				 * resending it.
				 */
				node = eb64_lookup(&qc->streams_by_id, strm_frm->id);
				if (!node) {
					TRACE_DEVEL("released stream", QUIC_EV_CONN_PRSAFRM, qc, cf);
					qc_frm_free(qc, &cf);
					continue;
				}

				stream_desc = eb64_entry(node, struct qc_stream_desc, by_id);
				if (strm_frm->offset.key + strm_frm->len <= stream_desc->ack_offset) {
					TRACE_DEVEL("ignored frame frame in already acked range",
					            QUIC_EV_CONN_PRSAFRM, qc, cf);
					qc_frm_free(qc, &cf);
					continue;
				}
				else if (strm_frm->offset.key < stream_desc->ack_offset) {
					uint64_t diff = stream_desc->ack_offset - strm_frm->offset.key;

					qc_stream_frm_mv_fwd(cf, diff);
					TRACE_DEVEL("updated partially acked frame",
					            QUIC_EV_CONN_PRSAFRM, qc, cf);
				}
			}
			/* Note that these frames are accepted in short packets only without
			 * "Length" packet field. Here, <*len> is used only to compute the
			 * sum of the lengths of the already built frames for this packet.
			 *
			 * Compute the length of this STREAM frame "header" made a all the field
			 * excepting the variable ones. Note that +1 is for the type of this frame.
			 */
			hlen = 1 + quic_int_getsize(cf->stream.id) +
				((cf->type & QUIC_STREAM_FRAME_TYPE_OFF_BIT) ? quic_int_getsize(cf->stream.offset.key) : 0);
			/* Compute the data length of this STREAM frame. */
			avail_room = room - hlen;
			if ((ssize_t)avail_room <= 0)
				continue;

			TRACE_DEVEL("          New STREAM frame build (room, len)",
			            QUIC_EV_CONN_BCFRMS, qc, &room, len);

			/* hlen contains STREAM id and offset. Ensure there is
			 * enough room for length field.
			 */
			if (cf->type & QUIC_STREAM_FRAME_TYPE_LEN_BIT) {
				dlen = QUIC_MIN((uint64_t)max_available_room(avail_room, &dlen_sz),
				                cf->stream.len);
				dlen_sz = quic_int_getsize(dlen);
				flen = hlen + dlen_sz + dlen;
			}
			else {
				dlen = QUIC_MIN((uint64_t)avail_room, cf->stream.len);
				flen = hlen + dlen;
			}

			if (cf->stream.len && !dlen) {
				/* Only a small gap is left on buffer, not
				 * enough to encode the STREAM data length.
				 */
				continue;
			}

			TRACE_DEVEL(" STREAM data length (hlen, stream.len, dlen)",
			            QUIC_EV_CONN_BCFRMS, qc, &hlen, &cf->stream.len, &dlen);
			TRACE_DEVEL("                 STREAM frame length (flen)",
			            QUIC_EV_CONN_BCFRMS, qc, &flen);
			/* Add the STREAM data length and its encoded length to the packet
			 * length and the length of this length.
			 */
			*len += flen;
			room -= flen;
			if (dlen == cf->stream.len) {
				/* <cf> STREAM data have been consumed. */
				LIST_DEL_INIT(&cf->list);
				LIST_APPEND(outlist, &cf->list);

				/* Do not notify MUX on retransmission. */
				if (qc->flags & QUIC_FL_CONN_TX_MUX_CONTEXT) {
					qcc_streams_sent_done(cf->stream.stream->ctx,
					                      cf->stream.len,
					                      cf->stream.offset.key);
				}
			}
			else {
				struct quic_frame *new_cf;
				struct buffer cf_buf;

				new_cf = qc_frm_alloc(cf->type);
				if (!new_cf) {
					TRACE_ERROR("No memory for new STREAM frame", QUIC_EV_CONN_BCFRMS, qc);
					continue;
				}

				new_cf->stream.stream = cf->stream.stream;
				new_cf->stream.buf = cf->stream.buf;
				new_cf->stream.id = cf->stream.id;
				new_cf->stream.offset = cf->stream.offset;
				new_cf->stream.len = dlen;
				new_cf->type |= QUIC_STREAM_FRAME_TYPE_LEN_BIT;
				/* FIN bit reset */
				new_cf->type &= ~QUIC_STREAM_FRAME_TYPE_FIN_BIT;
				new_cf->stream.data = cf->stream.data;
				new_cf->stream.dup = cf->stream.dup;
				TRACE_DEVEL("split frame", QUIC_EV_CONN_PRSAFRM, qc, new_cf);
				if (cf->origin) {
					TRACE_DEVEL("duplicated frame", QUIC_EV_CONN_PRSAFRM, qc);
					/* This <cf> frame was duplicated */
					LIST_APPEND(&cf->origin->reflist, &new_cf->ref);
					new_cf->origin = cf->origin;
					/* Detach this STREAM frame from its origin */
					LIST_DEL_INIT(&cf->ref);
					cf->origin = NULL;
				}
				LIST_APPEND(outlist, &new_cf->list);
				cf->type |= QUIC_STREAM_FRAME_TYPE_OFF_BIT;
				/* Consume <dlen> bytes of the current frame. */
				cf_buf = b_make(b_orig(cf->stream.buf),
				                b_size(cf->stream.buf),
				                (char *)cf->stream.data - b_orig(cf->stream.buf), 0);
				cf->stream.len -= dlen;
				cf->stream.offset.key += dlen;
				cf->stream.data = (unsigned char *)b_peek(&cf_buf, dlen);

				/* Do not notify MUX on retransmission. */
				if (qc->flags & QUIC_FL_CONN_TX_MUX_CONTEXT) {
					qcc_streams_sent_done(new_cf->stream.stream->ctx,
					                      new_cf->stream.len,
					                      new_cf->stream.offset.key);
				}
			}

			/* TODO the MUX is notified about the frame sending via
			 * previous qcc_streams_sent_done call. However, the
			 * sending can fail later, for example if the sendto
			 * system call returns an error. As the MUX has been
			 * notified, the transport layer is responsible to
			 * bufferize and resent the announced data later.
			 */

			break;

		default:
			flen = qc_frm_len(cf);
			BUG_ON(!flen);
			if (flen > room)
				continue;

			*len += flen;
			room -= flen;
			LIST_DEL_INIT(&cf->list);
			LIST_APPEND(outlist, &cf->list);
			break;
		}

		/* Successful status as soon as a frame could be built */
		ret = 1;
	}

 leave:
	TRACE_LEAVE(QUIC_EV_CONN_BCFRMS, qc);
	return ret;
}

/* Generate a CONNECTION_CLOSE frame for <qc> on <qel> encryption level. <out>
 * is used as return parameter and should be zero'ed by the caller.
 */
static void qc_build_cc_frm(struct quic_conn *qc, struct quic_enc_level *qel,
                            struct quic_frame *out)
{
	/* TODO improve CONNECTION_CLOSE on Initial/Handshake encryption levels
	 *
	 * A CONNECTION_CLOSE frame should be sent in several packets with
	 * different encryption levels depending on the client context. This is
	 * to ensure that the client can decrypt it. See RFC 9000 10.2.3 for
	 * more details on how to implement it.
	 */
	TRACE_ENTER(QUIC_EV_CONN_BFRM, qc);


	if (qc->err.app) {
		if (unlikely(qel == qc->iel || qel == qc->hel)) {
			/* RFC 9000 10.2.3.  Immediate Close during the Handshake
			 *
			 * Sending a CONNECTION_CLOSE of type 0x1d in an Initial or Handshake
			 * packet could expose application state or be used to alter application
			 * state.  A CONNECTION_CLOSE of type 0x1d MUST be replaced by a
			 * CONNECTION_CLOSE of type 0x1c when sending the frame in Initial or
			 * Handshake packets.  Otherwise, information about the application
			 * state might be revealed.  Endpoints MUST clear the value of the
			 * Reason Phrase field and SHOULD use the APPLICATION_ERROR code when
			 * converting to a CONNECTION_CLOSE of type 0x1c.
			 */
			out->type = QUIC_FT_CONNECTION_CLOSE;
			out->connection_close.error_code = QC_ERR_APPLICATION_ERROR;
			out->connection_close.reason_phrase_len = 0;
		}
		else {
			out->type = QUIC_FT_CONNECTION_CLOSE_APP;
			out->connection_close_app.error_code = qc->err.code;
			out->connection_close_app.reason_phrase_len = 0;
		}
	}
	else {
		out->type = QUIC_FT_CONNECTION_CLOSE;
		out->connection_close.error_code = qc->err.code;
		out->connection_close.reason_phrase_len = 0;
	}
	TRACE_LEAVE(QUIC_EV_CONN_BFRM, qc);

}

/* Returns the <ack_delay> field value in microsecond to be set in an ACK frame
 * depending on the time the packet with a new largest packet number was received.
 */
static inline uint64_t quic_compute_ack_delay_us(unsigned int time_received,
                                                 struct quic_conn *conn)
{
	return ((now_ms - time_received) * 1000) >> conn->tx.params.ack_delay_exponent;
}

/* This function builds a clear packet from <pkt> information (its type)
 * into a buffer with <pos> as position pointer and <qel> as QUIC TLS encryption
 * level for <conn> QUIC connection and <qel> as QUIC TLS encryption level,
 * filling the buffer with as much frames as possible from <frms> list of
 * prebuilt frames.
 * The trailing QUIC_TLS_TAG_LEN bytes of this packet are not built. But they are
 * reserved so that to ensure there is enough room to build this AEAD TAG after
 * having returned from this function.
 * This function also updates the value of <buf_pn> pointer to point to the packet
 * number field in this packet. <pn_len> will also have the packet number
 * length as value.
 *
 * Return 1 if succeeded (enough room to buile this packet), O if not.
 */
static int qc_do_build_pkt(unsigned char *pos, const unsigned char *end,
                           size_t dglen, struct quic_tx_packet *pkt,
                           int64_t pn, size_t *pn_len, unsigned char **buf_pn,
                           int must_ack, int padding, int cc, int probe,
                           struct quic_enc_level *qel, struct quic_conn *qc,
                           const struct quic_version *ver, struct list *frms)
{
	unsigned char *beg, *payload;
	size_t len, len_sz, len_frms, padding_len;
	struct quic_frame frm;
	struct quic_frame ack_frm;
	struct quic_frame cc_frm;
	size_t ack_frm_len, head_len;
	int64_t rx_largest_acked_pn;
	int add_ping_frm;
	struct list frm_list = LIST_HEAD_INIT(frm_list);
	struct quic_frame *cf;
	int ret = 0;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);

	/* Length field value with CRYPTO frames if present. */
	len_frms = 0;
	beg = pos;
	/* When not probing, and no immediate close is required, reduce the size of this
	 * buffer to respect the congestion controller window.
	 * This size will be limited if we have ack-eliciting frames to send from <frms>.
	 */
	if (!probe && !LIST_ISEMPTY(frms) && !cc) {
		size_t path_room;

		path_room = quic_cc_path_prep_data(qc->path);
		if (end - beg > path_room)
			end = beg + path_room;
	}

	/* Ensure there is enough room for the TLS encryption tag and a zero token
	 * length field if any.
	 */
	if (end - pos < QUIC_TLS_TAG_LEN +
	    (pkt->type == QUIC_PACKET_TYPE_INITIAL ? 1 : 0))
		goto no_room;

	end -= QUIC_TLS_TAG_LEN;
	rx_largest_acked_pn = qel->pktns->rx.largest_acked_pn;
	/* packet number length */
	*pn_len = quic_packet_number_length(pn, rx_largest_acked_pn);
	/* Build the header */
	if ((pkt->type == QUIC_PACKET_TYPE_SHORT &&
	    !quic_build_packet_short_header(&pos, end, *pn_len, qc, qel->tls_ctx.flags)) ||
	    (pkt->type != QUIC_PACKET_TYPE_SHORT &&
		!quic_build_packet_long_header(&pos, end, pkt->type, *pn_len, qc, ver)))
		goto no_room;

	/* Encode the token length (0) for an Initial packet. */
	if (pkt->type == QUIC_PACKET_TYPE_INITIAL) {
		if (end <= pos)
			goto no_room;

		*pos++ = 0;
	}

	head_len = pos - beg;
	/* Build an ACK frame if required. */
	ack_frm_len = 0;
	/* Do not ack and probe at the same time. */
	if ((must_ack || (qel->pktns->flags & QUIC_FL_PKTNS_ACK_REQUIRED)) && !qel->pktns->tx.pto_probe) {
		struct quic_arngs *arngs = &qel->pktns->rx.arngs;
		BUG_ON(eb_is_empty(&qel->pktns->rx.arngs.root));
		ack_frm.type = QUIC_FT_ACK;
		ack_frm.tx_ack.arngs = arngs;
		if (qel->pktns->flags & QUIC_FL_PKTNS_NEW_LARGEST_PN) {
			qel->pktns->tx.ack_delay =
				quic_compute_ack_delay_us(qel->pktns->rx.largest_time_received, qc);
			qel->pktns->flags &= ~QUIC_FL_PKTNS_NEW_LARGEST_PN;
		}
		ack_frm.tx_ack.ack_delay = qel->pktns->tx.ack_delay;
		/* XXX BE CAREFUL XXX : here we reserved at least one byte for the
		 * smallest frame (PING) and <*pn_len> more for the packet number. Note
		 * that from here, we do not know if we will have to send a PING frame.
		 * This will be decided after having computed the ack-eliciting frames
		 * to be added to this packet.
		 */
		if (end - pos <= 1 + *pn_len)
			goto no_room;

		ack_frm_len = qc_frm_len(&ack_frm);
		if (ack_frm_len > end - 1 - *pn_len - pos)
			goto no_room;
	}

	/* Length field value without the ack-eliciting frames. */
	len = ack_frm_len + *pn_len;
	len_frms = 0;
	if (!cc && !LIST_ISEMPTY(frms)) {
		ssize_t room = end - pos;

		TRACE_PROTO("Avail. ack eliciting frames", QUIC_EV_CONN_FRMLIST, qc, frms);
		/* Initialize the length of the frames built below to <len>.
		 * If any frame could be successfully built by qc_build_frms(),
		 * we will have len_frms > len.
		 */
		len_frms = len;
		if (!qc_build_frms(&frm_list, frms,
		                   end - pos, &len_frms, pos - beg, qel, qc)) {
			TRACE_PROTO("Not enough room", QUIC_EV_CONN_TXPKT,
			            qc, NULL, NULL, &room);
			if (padding) {
				len_frms = 0;
				goto comp_pkt_len;
			}

			if (!ack_frm_len && !qel->pktns->tx.pto_probe)
				goto no_room;
		}
	}

 comp_pkt_len:
	/* Length (of the remaining data). Must not fail because, the buffer size
	 * has been checked above. Note that we have reserved QUIC_TLS_TAG_LEN bytes
	 * for the encryption tag. It must be taken into an account for the length
	 * of this packet.
	 */
	if (len_frms)
		len = len_frms + QUIC_TLS_TAG_LEN;
	else
		len += QUIC_TLS_TAG_LEN;
	/* CONNECTION_CLOSE frame */
	if (cc) {
		qc_build_cc_frm(qc, qel, &cc_frm);
		len += qc_frm_len(&cc_frm);
	}
	add_ping_frm = 0;
	padding_len = 0;
	len_sz = quic_int_getsize(len);
	/* Add this packet size to <dglen> */
	dglen += head_len + len_sz + len;
	/* Note that <padding> is true only when building an Handshake packet
	 * coalesced to an Initial packet.
	 */
	if (padding && dglen < QUIC_INITIAL_PACKET_MINLEN) {
		/* This is a maximum padding size */
		padding_len = QUIC_INITIAL_PACKET_MINLEN - dglen;
		/* The length field value is of this packet is <len> + <padding_len>
		 * the size of which may be greater than the initial computed size
		 * <len_sz>. So, let's deduce the difference between these to packet
		 * sizes from <padding_len>.
		 */
		padding_len -= quic_int_getsize(len + padding_len) - len_sz;
		len += padding_len;
	}
	else if (len_frms && len_frms < QUIC_PACKET_PN_MAXLEN) {
		len += padding_len = QUIC_PACKET_PN_MAXLEN - len_frms;
	}
	else if (LIST_ISEMPTY(&frm_list)) {
		if (qel->pktns->tx.pto_probe) {
			/* If we cannot send a frame, we send a PING frame. */
			add_ping_frm = 1;
			len += 1;
			dglen += 1;
			/* Note that only we are in the case where this Initial packet
			 * is not coalesced to an Handshake packet. We must directly
			 * pad the datragram.
			 */
			if (pkt->type == QUIC_PACKET_TYPE_INITIAL) {
				if (dglen < QUIC_INITIAL_PACKET_MINLEN) {
					padding_len = QUIC_INITIAL_PACKET_MINLEN - dglen;
					padding_len -= quic_int_getsize(len + padding_len) - len_sz;
					len += padding_len;
				}
			}
			else {
				/* Note that +1 is for the PING frame */
				if (*pn_len + 1 < QUIC_PACKET_PN_MAXLEN)
					len += padding_len = QUIC_PACKET_PN_MAXLEN - *pn_len - 1;
			}
		}
		else {
			/* If there is no frame at all to follow, add at least a PADDING frame. */
			if (!ack_frm_len && !cc)
				len += padding_len = QUIC_PACKET_PN_MAXLEN - *pn_len;
		}
	}

	if (pkt->type != QUIC_PACKET_TYPE_SHORT && !quic_enc_int(&pos, end, len))
		goto no_room;

	/* Packet number field address. */
	*buf_pn = pos;

	/* Packet number encoding. */
	if (!quic_packet_number_encode(&pos, end, pn, *pn_len))
		goto no_room;

	/* payload building (ack-eliciting or not frames) */
	payload = pos;
	if (ack_frm_len) {
		if (!qc_build_frm(&pos, end, &ack_frm, pkt, qc))
			goto no_room;

		pkt->largest_acked_pn = quic_pktns_get_largest_acked_pn(qel->pktns);
		pkt->flags |= QUIC_FL_TX_PACKET_ACK;
	}

	/* Ack-eliciting frames */
	if (!LIST_ISEMPTY(&frm_list)) {
		struct quic_frame *tmp_cf;
		list_for_each_entry_safe(cf, tmp_cf, &frm_list, list) {
			if (!qc_build_frm(&pos, end, cf, pkt, qc)) {
				ssize_t room = end - pos;
				TRACE_PROTO("Not enough room", QUIC_EV_CONN_TXPKT,
				            qc, NULL, NULL, &room);
				/* Note that <cf> was added from <frms> to <frm_list> list by
				 * qc_build_frms().
				 */
				LIST_DEL_INIT(&cf->list);
				LIST_INSERT(frms, &cf->list);
				continue;
			}

			quic_tx_packet_refinc(pkt);
			cf->pkt = pkt;
		}
	}

	/* Build a PING frame if needed. */
	if (add_ping_frm) {
		frm.type = QUIC_FT_PING;
		if (!qc_build_frm(&pos, end, &frm, pkt, qc))
			goto no_room;
	}

	/* Build a CONNECTION_CLOSE frame if needed. */
	if (cc) {
		if (!qc_build_frm(&pos, end, &cc_frm, pkt, qc))
			goto no_room;

		pkt->flags |= QUIC_FL_TX_PACKET_CC;
	}

	/* Build a PADDING frame if needed. */
	if (padding_len) {
		frm.type = QUIC_FT_PADDING;
		frm.padding.len = padding_len;
		if (!qc_build_frm(&pos, end, &frm, pkt, qc))
			goto no_room;
	}

	if (pos == payload) {
		/* No payload was built because of congestion control */
		TRACE_PROTO("limited by congestion control", QUIC_EV_CONN_TXPKT, qc);
		goto no_room;
	}

	/* If this packet is ack-eliciting and we are probing let's
	 * decrement the PTO probe counter.
	 */
	if ((pkt->flags & QUIC_FL_TX_PACKET_ACK_ELICITING) &&
	    qel->pktns->tx.pto_probe)
		qel->pktns->tx.pto_probe--;

	pkt->len = pos - beg;
	LIST_SPLICE(&pkt->frms, &frm_list);

	ret = 1;
	TRACE_PROTO("Packet ack-eliciting frames", QUIC_EV_CONN_TXPKT, qc, pkt);
 leave:
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return ret;

 no_room:
	/* Replace the pre-built frames which could not be add to this packet */
	LIST_SPLICE(frms, &frm_list);
	TRACE_PROTO("Remaining ack-eliciting frames", QUIC_EV_CONN_FRMLIST, qc, frms);
	goto leave;
}

static inline void quic_tx_packet_init(struct quic_tx_packet *pkt, int type)
{
	pkt->type = type;
	pkt->len = 0;
	pkt->in_flight_len = 0;
	pkt->pn_node.key = (uint64_t)-1;
	LIST_INIT(&pkt->frms);
	pkt->time_sent = TICK_ETERNITY;
	pkt->next = NULL;
	pkt->prev = NULL;
	pkt->largest_acked_pn = -1;
	pkt->flags = 0;
	pkt->refcnt = 0;
}

/* Build a packet into a buffer at <pos> position, <end> pointing to one byte past
 * the end of this buffer, with <pkt_type> as packet type for <qc> QUIC connection
 * at <qel> encryption level with <frms> list of prebuilt frames.
 *
 * Return -3 if the packet could not be allocated, -2 if could not be encrypted for
 * any reason, -1 if there was not enough room to build a packet.
 * XXX NOTE XXX
 * If you provide provide qc_build_pkt() with a big enough buffer to build a packet as big as
 * possible (to fill an MTU), the unique reason why this function may fail is the congestion
 * control window limitation.
 */
static struct quic_tx_packet *qc_build_pkt(unsigned char **pos,
                                           const unsigned char *end,
                                           struct quic_enc_level *qel,
                                           struct quic_tls_ctx *tls_ctx, struct list *frms,
                                           struct quic_conn *qc, const struct quic_version *ver,
                                           size_t dglen, int pkt_type, int must_ack,
                                           int padding, int probe, int cc, int *err)
{
	struct quic_tx_packet *ret_pkt = NULL;
	/* The pointer to the packet number field. */
	unsigned char *buf_pn;
	unsigned char *first_byte, *last_byte, *payload;
	int64_t pn;
	size_t pn_len, payload_len, aad_len;
	struct quic_tx_packet *pkt;
	int encrypt_failure = 0;

	TRACE_ENTER(QUIC_EV_CONN_TXPKT, qc);
	TRACE_PROTO("TX pkt build", QUIC_EV_CONN_TXPKT, qc, NULL, qel);
	*err = 0;
	pkt = pool_alloc(pool_head_quic_tx_packet);
	if (!pkt) {
		TRACE_DEVEL("Not enough memory for a new packet", QUIC_EV_CONN_TXPKT, qc);
		*err = -3;
		goto err;
	}

	quic_tx_packet_init(pkt, pkt_type);
	first_byte = *pos;
	pn_len = 0;
	buf_pn = NULL;

	pn = qel->pktns->tx.next_pn + 1;
	if (!qc_do_build_pkt(*pos, end, dglen, pkt, pn, &pn_len, &buf_pn,
	                     must_ack, padding, cc, probe, qel, qc, ver, frms)) {
		// trace already emitted by function above
		*err = -1;
		goto err;
	}

	last_byte = first_byte + pkt->len;
	payload = buf_pn + pn_len;
	payload_len = last_byte - payload;
	aad_len = payload - first_byte;

	quic_packet_encrypt(payload, payload_len, first_byte, aad_len, pn, tls_ctx, qc, &encrypt_failure);
	if (encrypt_failure) {
		/* TODO Unrecoverable failure, unencrypted data should be returned to the caller. */
		WARN_ON("quic_packet_encrypt failure");
		*err = -2;
		goto err;
	}

	last_byte += QUIC_TLS_TAG_LEN;
	pkt->len += QUIC_TLS_TAG_LEN;
	quic_apply_header_protection(qc, first_byte, buf_pn, pn_len, tls_ctx, &encrypt_failure);
	if (encrypt_failure) {
		/* TODO Unrecoverable failure, unencrypted data should be returned to the caller. */
		WARN_ON("quic_apply_header_protection failure");
		*err = -2;
		goto err;
	}

	/* Consume a packet number */
	qel->pktns->tx.next_pn++;
	qc->bytes.prep += pkt->len;
	if (qc->bytes.prep >= 3 * qc->bytes.rx && !quic_peer_validated_addr(qc)) {
		qc->flags |= QUIC_FL_CONN_ANTI_AMPLIFICATION_REACHED;
		TRACE_PROTO("anti-amplification limit reached", QUIC_EV_CONN_TXPKT, qc);
	}

	/* Now that a correct packet is built, let us consume <*pos> buffer. */
	*pos = last_byte;
	/* Attach the built packet to its tree. */
	pkt->pn_node.key = pn;
	/* Set the packet in fligth length for in flight packet only. */
	if (pkt->flags & QUIC_FL_TX_PACKET_IN_FLIGHT) {
		pkt->in_flight_len = pkt->len;
		qc->path->prep_in_flight += pkt->len;
	}
	/* Always reset this flag */
	qc->flags &= ~QUIC_FL_CONN_IMMEDIATE_CLOSE;
	if (pkt->flags & QUIC_FL_TX_PACKET_ACK) {
		qel->pktns->flags &= ~QUIC_FL_PKTNS_ACK_REQUIRED;
		qel->pktns->rx.nb_aepkts_since_last_ack = 0;
		qc->flags &= ~QUIC_FL_CONN_ACK_TIMER_FIRED;
		if (tick_isset(qc->ack_expire)) {
		    qc->ack_expire = TICK_ETERNITY;
		    qc->idle_timer_task->expire = qc->idle_expire;
		    task_queue(qc->idle_timer_task);
		    TRACE_PROTO("ack timer cancelled", QUIC_EV_CONN_IDLE_TIMER, qc);
		}
	}

	pkt->pktns = qel->pktns;

	ret_pkt = pkt;
 leave:
	TRACE_PROTO("TX pkt built", QUIC_EV_CONN_TXPKT, qc, ret_pkt);
	TRACE_LEAVE(QUIC_EV_CONN_TXPKT, qc);
	return ret_pkt;

 err:
	/* TODO: what about the frames which have been built
	 * for this packet.
	 */
	free_quic_tx_packet(qc, pkt);
	goto leave;
}
/*
 * Local variables:
 *  c-indent-level: 8
 *  c-basic-offset: 8
 * End:
 */