public inbox for passt-dev@passt.top
 help / color / mirror / code / Atom feed
blob 1f26ff4c6f54c78e7e09fcdf5205506716395315 23538 bytes (raw)

  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
 
// SPDX-License-Identifier: AGPL-3.0-or-later

/* PASTA - Pack A Subtle Tap Abstraction
 *  for network namespace/tap device mode
 *
 * tcp_splice.c - direct namespace forwarding for local connections
 *
 * Copyright (c) 2020-2022 Red Hat GmbH
 * Author: Stefano Brivio <sbrivio@redhat.com>
 */

/**
 * DOC: Theory of Operation
 *
 *
 * For local traffic directed to TCP ports configured for direct mapping between
 * namespaces, packets are directly translated between L4 sockets using a pair
 * of splice() syscalls. These connections are tracked in the @tc array of
 * struct tcp_splice_conn, using these events:
 *
 * - SPLICE_CONNECT:		connection accepted, connecting to target
 * - SPLICE_ESTABLISHED:	connection to target established
 * - SPLICE_A_OUT_WAIT:		pipe to accepted socket full, wait for EPOLLOUT
 * - SPLICE_B_OUT_WAIT:		pipe to target socket full, wait for EPOLLOUT
 * - SPLICE_A_FIN_RCVD:		FIN (EPOLLRDHUP) seen from accepted socket
 * - SPLICE_B_FIN_RCVD:		FIN (EPOLLRDHUP) seen from target socket
 * - SPLICE_A_FIN_RCVD:		FIN (write shutdown) sent to accepted socket
 * - SPLICE_B_FIN_RCVD:		FIN (write shutdown) sent to target socket
 *
 * #syscalls:pasta pipe2|pipe fcntl armv6l:fcntl64 armv7l:fcntl64 ppc64:fcntl64
 */

#include <sched.h>
#include <errno.h>
#include <fcntl.h>
#include <limits.h>
#include <stdint.h>
#include <stdbool.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
#include <net/ethernet.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <sys/epoll.h>
#include <sys/types.h>
#include <sys/socket.h>

#include "util.h"
#include "passt.h"
#include "log.h"

#define MAX_PIPE_SIZE			(8UL * 1024 * 1024)
#define TCP_SPLICE_MAX_CONNS		(128 * 1024)
#define TCP_SPLICE_PIPE_POOL_SIZE	16
#define TCP_SPLICE_CONN_PRESSURE	30	/* % of splice_conn_count */
#define TCP_SPLICE_FILE_PRESSURE	30	/* % of c->nofile */

/* From tcp.c */
extern int init_sock_pool4		[TCP_SOCK_POOL_SIZE];
extern int init_sock_pool6		[TCP_SOCK_POOL_SIZE];
extern int ns_sock_pool4		[TCP_SOCK_POOL_SIZE];
extern int ns_sock_pool6		[TCP_SOCK_POOL_SIZE];

/* Pool of pre-opened pipes */
static int splice_pipe_pool		[TCP_SPLICE_PIPE_POOL_SIZE][2][2];

/**
 * struct tcp_splice_conn - Descriptor for a spliced TCP connection
 * @a:			File descriptor number of socket for accepted connection
 * @pipe_a_b:		Pipe ends for splice() from @a to @b
 * @b:			File descriptor number of peer connected socket
 * @pipe_b_a:		Pipe ends for splice() from @b to @a
 * @events:		Events observed/actions performed on connection
 * @flags:		Connection flags (attributes, not events)
 * @a_read:		Bytes read from @a (not fully written to @b in one shot)
 * @a_written:		Bytes written to @a (not fully written from one @b read)
 * @b_read:		Bytes read from @b (not fully written to @a in one shot)
 * @b_written:		Bytes written to @b (not fully written from one @a read)
*/
struct tcp_splice_conn {
	int a;
	int pipe_a_b[2];
	int b;
	int pipe_b_a[2];

	uint8_t events;
#define CLOSED				0
#define CONNECT				BIT(0)
#define ESTABLISHED			BIT(1)
#define A_OUT_WAIT			BIT(2)
#define B_OUT_WAIT			BIT(3)
#define A_FIN_RCVD			BIT(4)
#define B_FIN_RCVD			BIT(5)
#define A_FIN_SENT			BIT(6)
#define B_FIN_SENT			BIT(7)

	uint8_t flags;
#define SOCK_V6				BIT(0)
#define IN_EPOLL			BIT(1)
#define RCVLOWAT_SET_A			BIT(2)
#define RCVLOWAT_SET_B			BIT(3)
#define RCVLOWAT_ACT_A			BIT(4)
#define RCVLOWAT_ACT_B			BIT(5)
#define CLOSING				BIT(6)

	uint32_t a_read;
	uint32_t a_written;
	uint32_t b_read;
	uint32_t b_written;
};

#define CONN_V6(x)			(x->flags & SOCK_V6)
#define CONN_V4(x)			(!CONN_V6(x))
#define CONN_HAS(conn, set)		((conn->events & (set)) == (set))
#define CONN(index)			(tc + (index))

/* Spliced connections */
static struct tcp_splice_conn tc[TCP_SPLICE_MAX_CONNS];

/* Display strings for connection events */
static const char *tcp_splice_event_str[] __attribute((__unused__)) = {
	"CONNECT", "ESTABLISHED", "A_OUT_WAIT", "B_OUT_WAIT",
	"A_FIN_RCVD", "B_FIN_RCVD", "A_FIN_SENT", "B_FIN_SENT",
};

/* Display strings for connection flags */
static const char *tcp_splice_flag_str[] __attribute((__unused__)) = {
	"SOCK_V6", "IN_EPOLL", "RCVLOWAT_SET_A", "RCVLOWAT_SET_B",
	"RCVLOWAT_ACT_A", "RCVLOWAT_ACT_B", "CLOSING",
};

/**
 * tcp_splice_conn_epoll_events() - epoll events masks for given state
 * @events:	Connection event flags
 * @a:		Event mask for socket with accepted connection, set on return
 * @b:		Event mask for connection target socket, set on return
 */
static void tcp_splice_conn_epoll_events(uint16_t events,
					 uint32_t *a, uint32_t *b)
{
	*a = *b = 0;

	if (events & ESTABLISHED) {
		if (!(events & B_FIN_SENT))
			*a = EPOLLIN | EPOLLRDHUP;
		if (!(events & A_FIN_SENT))
			*b = EPOLLIN | EPOLLRDHUP;
	} else if (events & CONNECT) {
		*b = EPOLLOUT;
	}

	*a |= (events & A_OUT_WAIT) ? EPOLLOUT : 0;
	*b |= (events & B_OUT_WAIT) ? EPOLLOUT : 0;
}

static void tcp_splice_destroy(struct ctx *c, struct tcp_splice_conn *conn);
static int tcp_splice_epoll_ctl(const struct ctx *c,
				struct tcp_splice_conn *conn);

/**
 * conn_flag_do() - Set/unset given flag, log, update epoll on CLOSING flag
 * @c:		Execution context
 * @conn:	Connection pointer
 * @flag:	Flag to set, or ~flag to unset
 */
static void conn_flag_do(const struct ctx *c, struct tcp_splice_conn *conn,
			 unsigned long flag)
{
	if (flag & (flag - 1)) {
		if (!(conn->flags & ~flag))
			return;

		conn->flags &= flag;
		if (fls(~flag) >= 0) {
			debug("TCP (spliced): index %li: %s dropped", conn - tc,
			      tcp_splice_flag_str[fls(~flag)]);
		}
	} else {
		if (conn->flags & flag)
			return;

		conn->flags |= flag;
		if (fls(flag) >= 0) {
			debug("TCP (spliced): index %li: %s", conn - tc,
			      tcp_splice_flag_str[fls(flag)]);
		}
	}

	if (flag == CLOSING)
		tcp_splice_epoll_ctl(c, conn);
}

#define conn_flag(c, conn, flag)					\
	do {								\
		trace("TCP (spliced): flag at %s:%i",			\
		      __func__, __LINE__);				\
		conn_flag_do(c, conn, flag);				\
	} while (0)

/**
 * tcp_splice_epoll_ctl() - Add/modify/delete epoll state from connection events
 * @c:		Execution context
 * @conn:	Connection pointer
 *
 * Return: 0 on success, negative error code on failure (not on deletion)
 */
static int tcp_splice_epoll_ctl(const struct ctx *c,
				struct tcp_splice_conn *conn)
{
	int m = (conn->flags & IN_EPOLL) ? EPOLL_CTL_MOD : EPOLL_CTL_ADD;
	union epoll_ref ref_a = { .r.proto = IPPROTO_TCP, .r.s = conn->a,
				  .r.p.tcp.tcp.splice = 1,
				  .r.p.tcp.tcp.index = conn - tc, };
	union epoll_ref ref_b = { .r.proto = IPPROTO_TCP, .r.s = conn->b,
				  .r.p.tcp.tcp.splice = 1,
				  .r.p.tcp.tcp.index = conn - tc, };
	struct epoll_event ev_a = { .data.u64 = ref_a.u64 };
	struct epoll_event ev_b = { .data.u64 = ref_b.u64 };
	uint32_t events_a, events_b;

	if (conn->flags & CLOSING)
		goto delete;

	tcp_splice_conn_epoll_events(conn->events, &events_a, &events_b);
	ev_a.events = events_a;
	ev_b.events = events_b;

	if (epoll_ctl(c->epollfd, m, conn->a, &ev_a) ||
	    epoll_ctl(c->epollfd, m, conn->b, &ev_b))
		goto delete;

	conn->flags |= IN_EPOLL;		/* No need to log this */

	return 0;

delete:
	epoll_ctl(c->epollfd, EPOLL_CTL_DEL, conn->a, &ev_a);
	epoll_ctl(c->epollfd, EPOLL_CTL_DEL, conn->b, &ev_b);
	return -errno;
}

/**
 * conn_event_do() - Set and log connection events, update epoll state
 * @c:		Execution context
 * @conn:	Connection pointer
 * @event:	Connection event
 */
static void conn_event_do(const struct ctx *c, struct tcp_splice_conn *conn,
			  unsigned long event)
{
	if (event & (event - 1)) {
		if (!(conn->events & ~event))
			return;

		conn->events &= event;
		if (fls(~event) >= 0) {
			debug("TCP (spliced): index %li, ~%s", conn - tc,
			      tcp_splice_event_str[fls(~event)]);
		}
	} else {
		if (conn->events & event)
			return;

		conn->events |= event;
		if (fls(event) >= 0) {
			debug("TCP (spliced): index %li, %s", conn - tc,
			      tcp_splice_event_str[fls(event)]);
		}
	}

	if (tcp_splice_epoll_ctl(c, conn))
		conn_flag(c, conn, CLOSING);
}

#define conn_event(c, conn, event)					\
	do {								\
		trace("TCP (spliced): event at %s:%i",			\
		      __func__, __LINE__);				\
		conn_event_do(c, conn, event);				\
	} while (0)

/**
 * tcp_table_splice_compact - Compact spliced connection table
 * @c:		Execution context
 * @hole:	Pointer to recently closed connection
 */
static void tcp_table_splice_compact(struct ctx *c,
				     struct tcp_splice_conn *hole)
{
	struct tcp_splice_conn *move;

	if ((hole - tc) == --c->tcp.splice_conn_count) {
		debug("TCP (spliced): index %li (max) removed", hole - tc);
		return;
	}

	move = CONN(c->tcp.splice_conn_count);

	memcpy(hole, move, sizeof(*hole));

	move->a = move->b = -1;
	move->a_read = move->a_written = move->b_read = move->b_written = 0;
	move->pipe_a_b[0] = move->pipe_a_b[1] = -1;
	move->pipe_b_a[0] = move->pipe_b_a[1] = -1;
	move->flags = move->events = 0;

	debug("TCP (spliced): index %li moved to %li", move - tc, hole - tc);
	tcp_splice_epoll_ctl(c, hole);
	if (tcp_splice_epoll_ctl(c, hole))
		conn_flag(c, hole, CLOSING);
}

/**
 * tcp_splice_destroy() - Close spliced connection and pipes, clear
 * @c:		Execution context
 * @conn:	Connection pointer
 */
static void tcp_splice_destroy(struct ctx *c, struct tcp_splice_conn *conn)
{
	if (conn->events & ESTABLISHED) {
		/* Flushing might need to block: don't recycle them. */
		if (conn->pipe_a_b[0] != -1) {
			close(conn->pipe_a_b[0]);
			close(conn->pipe_a_b[1]);
			conn->pipe_a_b[0] = conn->pipe_a_b[1] = -1;
		}
		if (conn->pipe_b_a[0] != -1) {
			close(conn->pipe_b_a[0]);
			close(conn->pipe_b_a[1]);
			conn->pipe_b_a[0] = conn->pipe_b_a[1] = -1;
		}
	}

	if (conn->events & CONNECT) {
		close(conn->b);
		conn->b = -1;
	}

	close(conn->a);
	conn->a = -1;
	conn->a_read = conn->a_written = conn->b_read = conn->b_written = 0;

	conn->events = CLOSED;
	conn->flags = 0;
	debug("TCP (spliced): index %li, CLOSED", conn - tc);

	tcp_table_splice_compact(c, conn);
}

/**
 * tcp_splice_connect_finish() - Completion of connect() or call on success
 * @c:		Execution context
 * @conn:	Connection pointer
 *
 * Return: 0 on success, -EIO on failure
 */
static int tcp_splice_connect_finish(const struct ctx *c,
				     struct tcp_splice_conn *conn)
{
	int i;

	conn->pipe_a_b[0] = conn->pipe_b_a[0] = -1;
	conn->pipe_a_b[1] = conn->pipe_b_a[1] = -1;

	for (i = 0; i < TCP_SPLICE_PIPE_POOL_SIZE; i++) {
		if (splice_pipe_pool[i][0][0] >= 0) {
			SWAP(conn->pipe_a_b[0], splice_pipe_pool[i][0][0]);
			SWAP(conn->pipe_a_b[1], splice_pipe_pool[i][0][1]);

			SWAP(conn->pipe_b_a[0], splice_pipe_pool[i][1][0]);
			SWAP(conn->pipe_b_a[1], splice_pipe_pool[i][1][1]);
			break;
		}
	}

	if (conn->pipe_a_b[0] < 0) {
		if (pipe2(conn->pipe_a_b, O_NONBLOCK | O_CLOEXEC) ||
		    pipe2(conn->pipe_b_a, O_NONBLOCK | O_CLOEXEC)) {
			conn_flag(c, conn, CLOSING);
			return -EIO;
		}

		if (fcntl(conn->pipe_a_b[0], F_SETPIPE_SZ, c->tcp.pipe_size)) {
			trace("TCP (spliced): cannot set a->b pipe size to %lu",
			      c->tcp.pipe_size);
		}

		if (fcntl(conn->pipe_b_a[0], F_SETPIPE_SZ, c->tcp.pipe_size)) {
			trace("TCP (spliced): cannot set b->a pipe size to %lu",
			      c->tcp.pipe_size);
		}
	}

	if (!(conn->events & ESTABLISHED))
		conn_event(c, conn, ESTABLISHED);

	return 0;
}

/**
 * tcp_splice_connect() - Create and connect socket for new spliced connection
 * @c:		Execution context
 * @conn:	Connection pointer
 * @s:		Accepted socket
 * @port:	Destination port, host order
 *
 * Return: 0 for connect() succeeded or in progress, negative value on error
 */
static int tcp_splice_connect(const struct ctx *c, struct tcp_splice_conn *conn,
			      int s, in_port_t port)
{
	int sock_conn = (s >= 0) ? s : socket(CONN_V6(conn) ? AF_INET6 :
							      AF_INET,
					      SOCK_STREAM | SOCK_NONBLOCK,
					      IPPROTO_TCP);
	struct sockaddr_in6 addr6 = {
		.sin6_family = AF_INET6,
		.sin6_port = htons(port),
		.sin6_addr = IN6ADDR_LOOPBACK_INIT,
	};
	struct sockaddr_in addr4 = {
		.sin_family = AF_INET,
		.sin_port = htons(port),
		.sin_addr = { .s_addr = htonl(INADDR_LOOPBACK) },
	};
	const struct sockaddr *sa;
	socklen_t sl;

	if (sock_conn < 0)
		return -errno;

	if (sock_conn > SOCKET_MAX) {
		close(sock_conn);
		return -EIO;
	}

	conn->b = sock_conn;

	if (s < 0)
		tcp_sock_set_bufsize(c, conn->b);

	if (setsockopt(conn->b, SOL_TCP, TCP_QUICKACK,
		       &((int){ 1 }), sizeof(int))) {
		trace("TCP (spliced): failed to set TCP_QUICKACK on socket %i",
		      conn->b);
	}

	if (CONN_V6(conn)) {
		sa = (struct sockaddr *)&addr6;
		sl = sizeof(addr6);
	} else {
		sa = (struct sockaddr *)&addr4;
		sl = sizeof(addr4);
	}

	if (connect(conn->b, sa, sl)) {
		if (errno != EINPROGRESS) {
			int ret = -errno;

			close(sock_conn);
			return ret;
		}
		conn_event(c, conn, CONNECT);
	} else {
		conn_event(c, conn, ESTABLISHED);
		return tcp_splice_connect_finish(c, conn);
	}

	return 0;
}

/**
 * struct tcp_splice_connect_ns_arg - Arguments for tcp_splice_connect_ns()
 * @c:		Execution context
 * @conn:	Accepted inbound connection
 * @port:	Destination port, host order
 * @ret:	Return value of tcp_splice_connect_ns()
 */
struct tcp_splice_connect_ns_arg {
	const struct ctx *c;
	struct tcp_splice_conn *conn;
	in_port_t port;
	int ret;
};

/**
 * tcp_splice_connect_ns() - Enter namespace and call tcp_splice_connect()
 * @arg:	See struct tcp_splice_connect_ns_arg
 *
 * Return: 0
 */
static int tcp_splice_connect_ns(void *arg)
{
	struct tcp_splice_connect_ns_arg *a;

	a = (struct tcp_splice_connect_ns_arg *)arg;
	ns_enter(a->c);
	a->ret = tcp_splice_connect(a->c, a->conn, -1, a->port);
	return 0;
}

/**
 * tcp_splice_new() - Handle new spliced connection
 * @c:		Execution context
 * @conn:	Connection pointer
 * @port:	Destination port, host order
 * @outbound:	Connection request coming from namespace
 *
 * Return: return code from connect()
 */
static int tcp_splice_new(const struct ctx *c, struct tcp_splice_conn *conn,
			  in_port_t port, int outbound)
{
	int *p, i, s = -1;

	if (outbound)
		p = CONN_V6(conn) ? init_sock_pool6 : init_sock_pool4;
	else
		p = CONN_V6(conn) ? ns_sock_pool6 : ns_sock_pool4;

	for (i = 0; i < TCP_SOCK_POOL_SIZE; i++, p++) {
		SWAP(s, *p);
		if (s >= 0)
			break;
	}

	/* No socket available in namespace: create a new one for connect() */
	if (s < 0 && !outbound) {
		struct tcp_splice_connect_ns_arg ns_arg = { c, conn, port, 0 };

		NS_CALL(tcp_splice_connect_ns, &ns_arg);
		return ns_arg.ret;
	}

	/* Otherwise, the socket will connect on the side it was created on */
	return tcp_splice_connect(c, conn, s, port);
}

/**
 * tcp_splice_dir() - Set sockets/pipe pointers reflecting flow direction
 * @conn:	Connection pointers
 * @ref_sock:	Socket returned as reference from epoll
 * @reverse:	Reverse direction: @ref_sock is used as destination
 * @from:	Destination socket pointer to set
 * @to:		Source socket pointer to set
 * @pipes:	Pipe set, assigned on return
 */
static void tcp_splice_dir(struct tcp_splice_conn *conn, int ref_sock,
			   int reverse, int *from, int *to, int **pipes)
{
	if (!reverse) {
		*from = ref_sock;
		*to   = (*from == conn->a) ? conn->b : conn->a;
	} else {
		*to   = ref_sock;
		*from = (*to   == conn->a) ? conn->b : conn->a;
	}

	*pipes = *from == conn->a ? conn->pipe_a_b : conn->pipe_b_a;
}

/**
 * tcp_sock_handler_splice() - Handler for socket mapped to spliced connection
 * @c:		Execution context
 * @ref:	epoll reference
 * @events:	epoll events bitmap
 *
 * #syscalls:pasta splice
 */
void tcp_sock_handler_splice(struct ctx *c, union epoll_ref ref,
			     uint32_t events)
{
	uint8_t lowat_set_flag, lowat_act_flag;
	int from, to, *pipes, eof, never_read;
	uint32_t *seq_read, *seq_write;
	struct tcp_splice_conn *conn;

	if (ref.r.p.tcp.tcp.listen) {
		struct sockaddr sa;
		socklen_t sl = sizeof(sa);
		int s;

		if (c->tcp.splice_conn_count >= TCP_SPLICE_MAX_CONNS)
			return;

		sa.sa_family = AF_UNSPEC; /* FIXME: stop clang-tidy complaining */
		if ((s = accept4(ref.r.s, &sa, &sl, SOCK_NONBLOCK)) < 0)
			return;

		if (setsockopt(s, SOL_TCP, TCP_QUICKACK, &((int){ 1 }),
			       sizeof(int))) {
			trace("TCP (spliced): failed to set TCP_QUICKACK on %i",
			      s);
		}

		conn = CONN(c->tcp.splice_conn_count++);
		conn->a = s;
		conn->flags = (sa.sa_family == AF_INET6) ? SOCK_V6 : 0;

		if (tcp_splice_new(c, conn, ref.r.p.tcp.tcp.index,
				   ref.r.p.tcp.tcp.outbound))
			conn_flag(c, conn, CLOSING);

		return;
	}

	conn = CONN(ref.r.p.tcp.tcp.index);

	if (conn->events == CLOSED)
		return;

	if (events & EPOLLERR)
		goto close;

	if (conn->events == CONNECT) {
		if (!(events & EPOLLOUT))
			goto close;
		if (tcp_splice_connect_finish(c, conn))
			goto close;
	}

	if (events & EPOLLOUT) {
		if (ref.r.s == conn->a)
			conn_event(c, conn, ~A_OUT_WAIT);
		else
			conn_event(c, conn, ~B_OUT_WAIT);

		tcp_splice_dir(conn, ref.r.s, 1, &from, &to, &pipes);
	} else {
		tcp_splice_dir(conn, ref.r.s, 0, &from, &to, &pipes);
	}

	if (events & EPOLLRDHUP) {
		if (ref.r.s == conn->a)
			conn_event(c, conn, A_FIN_RCVD);
		else
			conn_event(c, conn, B_FIN_RCVD);
	}

	if (events & EPOLLHUP) {
		if (ref.r.s == conn->a)
			conn_event(c, conn, A_FIN_SENT); /* Fake, but implied */
		else
			conn_event(c, conn, B_FIN_SENT);
	}

swap:
	eof = 0;
	never_read = 1;

	if (from == conn->a) {
		seq_read = &conn->a_read;
		seq_write = &conn->a_written;
		lowat_set_flag = RCVLOWAT_SET_A;
		lowat_act_flag = RCVLOWAT_ACT_A;
	} else {
		seq_read = &conn->b_read;
		seq_write = &conn->b_written;
		lowat_set_flag = RCVLOWAT_SET_B;
		lowat_act_flag = RCVLOWAT_ACT_B;
	}

	while (1) {
		ssize_t readlen, to_write = 0, written;
		int more = 0;

retry:
		readlen = splice(from, NULL, pipes[1], NULL, c->tcp.pipe_size,
				 SPLICE_F_MOVE | SPLICE_F_NONBLOCK);
		trace("TCP (spliced): %li from read-side call", readlen);
		if (readlen < 0) {
			if (errno == EINTR)
				goto retry;

			if (errno != EAGAIN)
				goto close;

			to_write = c->tcp.pipe_size;
		} else if (!readlen) {
			eof = 1;
			to_write = c->tcp.pipe_size;
		} else {
			never_read = 0;
			to_write += readlen;
			if (readlen >= (long)c->tcp.pipe_size * 90 / 100)
				more = SPLICE_F_MORE;

			if (conn->flags & lowat_set_flag)
				conn_flag(c, conn, lowat_act_flag);
		}

eintr:
		written = splice(pipes[0], NULL, to, NULL, to_write,
				 SPLICE_F_MOVE | more | SPLICE_F_NONBLOCK);
		trace("TCP (spliced): %li from write-side call (passed %lu)",
		      written, to_write);

		/* Most common case: skip updating counters. */
		if (readlen > 0 && readlen == written) {
			if (readlen >= (long)c->tcp.pipe_size * 10 / 100)
				continue;

			if (conn->flags & lowat_set_flag &&
			    readlen > (long)c->tcp.pipe_size / 10) {
				int lowat = c->tcp.pipe_size / 4;

				setsockopt(from, SOL_SOCKET, SO_RCVLOWAT,
					   &lowat, sizeof(lowat));

				conn_flag(c, conn, lowat_set_flag);
				conn_flag(c, conn, lowat_act_flag);
			}

			break;
		}

		*seq_read  += readlen > 0 ? readlen : 0;
		*seq_write += written > 0 ? written : 0;

		if (written < 0) {
			if (errno == EINTR)
				goto eintr;

			if (errno != EAGAIN)
				goto close;

			if (never_read)
				break;

			if (to == conn->a)
				conn_event(c, conn, A_OUT_WAIT);
			else
				conn_event(c, conn, B_OUT_WAIT);
			break;
		}

		if (never_read && written == (long)(c->tcp.pipe_size))
			goto retry;

		if (!never_read && written < to_write) {
			to_write -= written;
			goto retry;
		}

		if (eof)
			break;
	}

	if ((conn->events & A_FIN_RCVD) && !(conn->events & B_FIN_SENT)) {
		if (*seq_read == *seq_write && eof) {
			shutdown(conn->b, SHUT_WR);
			conn_event(c, conn, B_FIN_SENT);
		}
	}

	if ((conn->events & B_FIN_RCVD) && !(conn->events & A_FIN_SENT)) {
		if (*seq_read == *seq_write && eof) {
			shutdown(conn->a, SHUT_WR);
			conn_event(c, conn, A_FIN_SENT);
		}
	}

	if (CONN_HAS(conn, A_FIN_SENT | B_FIN_SENT))
		goto close;

	if ((events & (EPOLLIN | EPOLLOUT)) == (EPOLLIN | EPOLLOUT)) {
		events = EPOLLIN;

		SWAP(from, to);
		if (pipes == conn->pipe_a_b)
			pipes = conn->pipe_b_a;
		else
			pipes = conn->pipe_a_b;

		goto swap;
	}

	if (events & EPOLLHUP)
		goto close;

	return;

close:
	conn_flag(c, conn, CLOSING);
}

/**
 * tcp_set_pipe_size() - Set usable pipe size, probe starting from MAX_PIPE_SIZE
 * @c:		Execution context
 */
static void tcp_set_pipe_size(struct ctx *c)
{
	int probe_pipe[TCP_SPLICE_PIPE_POOL_SIZE * 2][2], i, j;

	c->tcp.pipe_size = MAX_PIPE_SIZE;

smaller:
	for (i = 0; i < TCP_SPLICE_PIPE_POOL_SIZE * 2; i++) {
		if (pipe2(probe_pipe[i], O_CLOEXEC)) {
			i++;
			break;
		}

		if (fcntl(probe_pipe[i][0], F_SETPIPE_SZ, c->tcp.pipe_size) < 0)
			break;
	}

	for (j = i - 1; j >= 0; j--) {
		close(probe_pipe[j][0]);
		close(probe_pipe[j][1]);
	}

	if (i == TCP_SPLICE_PIPE_POOL_SIZE * 2)
		return;

	if (!(c->tcp.pipe_size /= 2)) {
		c->tcp.pipe_size = MAX_PIPE_SIZE;
		return;
	}

	goto smaller;
}

/**
 * tcp_splice_pipe_refill() - Refill pool of pre-opened pipes
 * @c:		Execution context
 */
static void tcp_splice_pipe_refill(const struct ctx *c)
{
	int i;

	for (i = 0; i < TCP_SPLICE_PIPE_POOL_SIZE; i++) {
		if (splice_pipe_pool[i][0][0] >= 0)
			break;
		if (pipe2(splice_pipe_pool[i][0], O_NONBLOCK | O_CLOEXEC))
			continue;
		if (pipe2(splice_pipe_pool[i][1], O_NONBLOCK | O_CLOEXEC)) {
			close(splice_pipe_pool[i][1][0]);
			close(splice_pipe_pool[i][1][1]);
			continue;
		}

		if (fcntl(splice_pipe_pool[i][0][0], F_SETPIPE_SZ,
			  c->tcp.pipe_size)) {
			trace("TCP (spliced): cannot set a->b pipe size to %lu",
			      c->tcp.pipe_size);
		}

		if (fcntl(splice_pipe_pool[i][1][0], F_SETPIPE_SZ,
			  c->tcp.pipe_size)) {
			trace("TCP (spliced): cannot set b->a pipe size to %lu",
			      c->tcp.pipe_size);
		}
	}
}

/**
 * tcp_splice_init() - Initialise pipe pool and size
 * @c:		Execution context
 */
void tcp_splice_init(struct ctx *c)
{
	memset(splice_pipe_pool, 0xff, sizeof(splice_pipe_pool));
	tcp_set_pipe_size(c);
}

/**
 * tcp_splice_timer() - Timer for spliced connections
 * @c:		Execution context
 */
void tcp_splice_timer(struct ctx *c)
{
	struct tcp_splice_conn *conn;

	for (conn = CONN(c->tcp.splice_conn_count - 1); conn >= tc; conn--) {
		if (conn->flags & CLOSING) {
			tcp_splice_destroy(c, conn);
			return;
		}

		if ( (conn->flags & RCVLOWAT_SET_A) &&
		    !(conn->flags & RCVLOWAT_ACT_A)) {
			if (setsockopt(conn->a, SOL_SOCKET, SO_RCVLOWAT,
				       &((int){ 1 }), sizeof(int))) {
				trace("TCP (spliced): can't set SO_RCVLOWAT on "
				      "%i", conn->a);
			}
			conn_flag(c, conn, ~RCVLOWAT_SET_A);
		}

		if ( (conn->flags & RCVLOWAT_SET_B) &&
		    !(conn->flags & RCVLOWAT_ACT_B)) {
			if (setsockopt(conn->b, SOL_SOCKET, SO_RCVLOWAT,
				       &((int){ 1 }), sizeof(int))) {
				trace("TCP (spliced): can't set SO_RCVLOWAT on "
				      "%i", conn->b);
			}
			conn_flag(c, conn, ~RCVLOWAT_SET_B);
		}

		conn_flag(c, conn, ~RCVLOWAT_ACT_A);
		conn_flag(c, conn, ~RCVLOWAT_ACT_B);
	}

	tcp_splice_pipe_refill(c);
}

/**
 * tcp_splice_defer_handler() - Close connections without timer on file pressure
 * @c:		Execution context
 */
void tcp_splice_defer_handler(struct ctx *c)
{
	int max_conns = c->tcp.conn_count / 100 * TCP_SPLICE_CONN_PRESSURE;
	int max_files = c->nofile / 100 * TCP_SPLICE_FILE_PRESSURE;
	struct tcp_splice_conn *conn;

	if (c->tcp.splice_conn_count < MIN(max_files / 6, max_conns))
		return;

	for (conn = CONN(c->tcp.splice_conn_count - 1); conn >= tc; conn--) {
		if (conn->flags & CLOSING)
			tcp_splice_destroy(c, conn);
	}
}

debug log:

solving 1f26ff4 ...
found 1f26ff4 in https://archives.passt.top/passt-dev/20221104084333.3761760-2-david@gibson.dropbear.id.au/
found 99c5fa7 in https://passt.top/passt
preparing index
index prepared:
100644 99c5fa78f2fef0a47bc4ea3f7f7e0d0b00667273	tcp_splice.c

applying [1/1] https://archives.passt.top/passt-dev/20221104084333.3761760-2-david@gibson.dropbear.id.au/
diff --git a/tcp_splice.c b/tcp_splice.c
index 99c5fa7..1f26ff4 100644

Checking patch tcp_splice.c...
Applied patch tcp_splice.c cleanly.

index at:
100644 1f26ff4c6f54c78e7e09fcdf5205506716395315	tcp_splice.c

Code repositories for project(s) associated with this public inbox

	https://passt.top/passt

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).