litespeed-quic/tests/test_elision.c

420 lines
15 KiB
C
Raw Normal View History

/* Copyright (c) 2017 - 2021 LiteSpeed Technologies Inc. See LICENSE. */
2017-09-22 21:00:03 +00:00
#include <assert.h>
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/queue.h>
#include "lsquic.h"
#include "lsquic_int_types.h"
#include "lsquic_packet_common.h"
#include "lsquic_packet_gquic.h"
2017-09-22 21:00:03 +00:00
#include "lsquic_packet_out.h"
#include "lsquic_parse.h"
#include "lsquic_conn_flow.h"
#include "lsquic_sfcw.h"
#include "lsquic_varint.h"
#include "lsquic_hq.h"
#include "lsquic_hash.h"
2017-09-22 21:00:03 +00:00
#include "lsquic_stream.h"
#include "lsquic_types.h"
#include "lsquic_malo.h"
#include "lsquic_mm.h"
#include "lsquic_engine_public.h"
#include "lsquic_logger.h"
static const struct parse_funcs *const pf = select_pf_by_ver(LSQVER_043);
2017-09-22 21:00:03 +00:00
static struct {
unsigned char buf[0x1000];
2017-09-22 21:00:03 +00:00
size_t bufsz;
uint64_t off;
} stream_contents;
void
setup_stream_contents (uint64_t off, const char *str)
{
stream_contents.bufsz = strlen(str);
stream_contents.off = off;
memcpy(stream_contents.buf, str, stream_contents.bufsz);
}
void
setup_stream_contents_n (uint64_t off, const unsigned char *buf, size_t size)
{
stream_contents.bufsz = size;
stream_contents.off = off;
memcpy(stream_contents.buf, buf, size);
}
2017-09-22 21:00:03 +00:00
int
lsquic_stream_tosend_fin (const lsquic_stream_t *stream)
{
return 0;
}
uint64_t
lsquic_stream_tosend_offset (const lsquic_stream_t *stream)
{
return stream_contents.off;
}
size_t
lsquic_stream_tosend_read (lsquic_stream_t *stream, void *buf, size_t len,
int *reached_fin)
{
if (stream_contents.bufsz < len)
len = stream_contents.bufsz;
memcpy(buf, stream_contents.buf, len);
*reached_fin = lsquic_stream_tosend_fin(stream);
return len;
}
size_t
lsquic_stream_tosend_sz (const lsquic_stream_t *stream)
{
return stream_contents.bufsz;
}
void
lsquic_stream_acked (lsquic_stream_t *stream, enum quic_frame_type frame_type)
2017-09-22 21:00:03 +00:00
{
--stream->n_unacked;
}
static void
elide_single_stream_frame (void)
{
struct packet_out_frec_iter pofi;
2017-09-22 21:00:03 +00:00
struct lsquic_engine_public enpub;
lsquic_stream_t streams[1];
lsquic_packet_out_t *packet_out;
int len, off = 0;
memset(streams, 0, sizeof(streams));
memset(&enpub, 0, sizeof(enpub));
lsquic_mm_init(&enpub.enp_mm);
packet_out = lsquic_mm_get_packet_out(&enpub.enp_mm, NULL, GQUIC_MAX_PAYLOAD_SZ);
2017-09-22 21:00:03 +00:00
setup_stream_contents(123, "Dude, where is my car?");
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[0].id, lsquic_stream_tosend_offset(&streams[0]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[0]),
lsquic_stream_tosend_sz(&streams[0]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[0]);
packet_out->po_data_sz += len;
packet_out->po_frame_types |= (1 << QUIC_FRAME_STREAM);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[0],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, off, len);
2017-09-22 21:00:03 +00:00
assert(1 == streams[0].n_unacked);
assert(lsquic_pofi_first(&pofi, packet_out));
2017-09-22 21:00:03 +00:00
streams[0].stream_flags |= STREAM_RST_SENT;
lsquic_packet_out_elide_reset_stream_frames(packet_out, UINT64_MAX);
2017-09-22 21:00:03 +00:00
assert(0 == streams[0].n_unacked);
assert(0 == packet_out->po_frame_types);
assert(!lsquic_pofi_first(&pofi, packet_out));
2017-09-22 21:00:03 +00:00
lsquic_packet_out_destroy(packet_out, &enpub, NULL);
2017-09-22 21:00:03 +00:00
lsquic_mm_cleanup(&enpub.enp_mm);
}
/* In this test, we check that if the last STREAM frame is moved due to
* elision and PO_STREAM_END is set, the packet size is adjusted. This
* is needed to prevent data corruption for STREAM frames that have
* implicit length.
*/
static void
shrink_packet_post_elision (void)
{
struct packet_out_frec_iter pofi;
struct lsquic_engine_public enpub;
lsquic_stream_t streams[2];
lsquic_packet_out_t *packet_out;
const struct frame_rec *frec;
int len, off = 0;
unsigned char stream2_data[0x1000];
memset(stream2_data, '2', sizeof(stream2_data));
memset(streams, 0, sizeof(streams));
memset(&enpub, 0, sizeof(enpub));
lsquic_mm_init(&enpub.enp_mm);
packet_out = lsquic_mm_get_packet_out(&enpub.enp_mm, NULL, GQUIC_MAX_PAYLOAD_SZ);
setup_stream_contents(123, "Dude, where is my car?");
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[0].id, lsquic_stream_tosend_offset(&streams[0]),
lsquic_stream_tosend_fin(&streams[0]),
lsquic_stream_tosend_sz(&streams[0]),
(gsf_read_f) lsquic_stream_tosend_read,
&streams[0]);
packet_out->po_data_sz += len;
packet_out->po_frame_types |= (1 << QUIC_FRAME_STREAM);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[0],
QUIC_FRAME_STREAM, off, len);
/* We want to fill the packet just right so that PO_STREAM_END gets set */
const int exp = lsquic_packet_out_avail(packet_out);
setup_stream_contents_n(0, stream2_data, exp - 2);
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[1].id, lsquic_stream_tosend_offset(&streams[1]),
lsquic_stream_tosend_fin(&streams[1]),
lsquic_stream_tosend_sz(&streams[1]),
(gsf_read_f) lsquic_stream_tosend_read,
&streams[1]);
assert(len == exp);
packet_out->po_data_sz += len;
packet_out->po_frame_types |= (1 << QUIC_FRAME_STREAM);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[1],
QUIC_FRAME_STREAM, off, len);
assert(0 == lsquic_packet_out_avail(packet_out)); /* Same as len == exp check really */
packet_out->po_flags |= PO_STREAM_END;
assert(1 == streams[0].n_unacked);
assert(1 == streams[1].n_unacked);
assert(lsquic_pofi_first(&pofi, packet_out));
streams[0].stream_flags |= STREAM_RST_SENT;
lsquic_packet_out_elide_reset_stream_frames(packet_out, UINT64_MAX);
assert(0 == streams[0].n_unacked);
assert(QUIC_FTBIT_STREAM == packet_out->po_frame_types);
frec = lsquic_pofi_first(&pofi, packet_out);
assert(frec->fe_stream == &streams[1]);
assert(packet_out->po_data_sz == exp);
lsquic_packet_out_destroy(packet_out, &enpub, NULL);
lsquic_mm_cleanup(&enpub.enp_mm);
}
2017-09-22 21:00:03 +00:00
/* This test is more involved. We will construct the following packet:
*
* | ACK | STREAM A | STREAM B | STREAM C | RST A | STREAM D | STREAM E
*
* and elide STREAM A, STREAM C, and STREAM E to get
*
* | ACK | STREAM B | RST A | STREAM D |
*
* If `chop_regen' is set, ACK is dropped (this tests what happens when
* packet is resent).
*
* This should test most of the corner cases.
*/
static void
elide_three_stream_frames (int chop_regen)
{
struct packet_out_frec_iter pofi;
2017-09-22 21:00:03 +00:00
struct lsquic_engine_public enpub;
lsquic_stream_t streams[5];
lsquic_packet_out_t *packet_out, *ref_out;
struct frame_rec *frec;
2017-09-22 21:00:03 +00:00
unsigned short b_off, d_off;
int len;
memset(streams, 0, sizeof(streams));
memset(&enpub, 0, sizeof(enpub));
lsquic_mm_init(&enpub.enp_mm);
/* First, we construct the reference packet. We will only use it to
* compare payload and sizes:
*/
{
ref_out = lsquic_mm_get_packet_out(&enpub.enp_mm, NULL, GQUIC_MAX_PAYLOAD_SZ);
2017-09-22 21:00:03 +00:00
/* This is fake data for regeneration */
strcpy((char *) ref_out->po_data, "REGEN");
lsquic_packet_out_add_frame(ref_out, &enpub.enp_mm, 0,
QUIC_FRAME_ACK, ref_out->po_data_sz, 5);
2017-09-22 21:00:03 +00:00
ref_out->po_data_sz = ref_out->po_regen_sz = 5;
/* STREAM B */
setup_stream_contents(123, "BBBBBBBBBB");
streams[0].id = 'B';
len = pf->pf_gen_stream_frame(ref_out->po_data + ref_out->po_data_sz,
lsquic_packet_out_avail(ref_out),
streams[0].id, lsquic_stream_tosend_offset(&streams[0]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[0]),
lsquic_stream_tosend_sz(&streams[0]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[0]);
b_off = ref_out->po_data_sz;
ref_out->po_data_sz += len;
len = pf->pf_gen_rst_frame(ref_out->po_data + ref_out->po_data_sz,
lsquic_packet_out_avail(ref_out), 'A', 133, 0);
ref_out->po_data_sz += len;
/* STREAM D */
setup_stream_contents(123, "DDDDDDDDDD");
streams[0].id = 'D';
len = pf->pf_gen_stream_frame(ref_out->po_data + ref_out->po_data_sz,
lsquic_packet_out_avail(ref_out),
streams[0].id, lsquic_stream_tosend_offset(&streams[0]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[0]),
lsquic_stream_tosend_sz(&streams[0]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[0]);
d_off = ref_out->po_data_sz;
ref_out->po_data_sz += len;
}
/* Construct packet from which we will elide streams. Here, we attach
* stream objects to the packet.
*/
{
packet_out = lsquic_mm_get_packet_out(&enpub.enp_mm, NULL, GQUIC_MAX_PAYLOAD_SZ);
2017-09-22 21:00:03 +00:00
/* This is fake data for regeneration */
strcpy((char *) packet_out->po_data, "REGEN");
lsquic_packet_out_add_frame(packet_out, &enpub.enp_mm, 0,
QUIC_FRAME_ACK, packet_out->po_data_sz, 5);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz = packet_out->po_regen_sz = 5;
/* STREAM A */
setup_stream_contents(123, "AAAAAAAAAA");
streams[0].id = 'A';
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[0].id, lsquic_stream_tosend_offset(&streams[0]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[0]),
lsquic_stream_tosend_sz(&streams[0]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[0]);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[0],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
/* STREAM B */
setup_stream_contents(123, "BBBBBBBBBB");
streams[1].id = 'B';
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[1].id, lsquic_stream_tosend_offset(&streams[1]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[1]),
lsquic_stream_tosend_sz(&streams[1]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[1]);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[1],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
/* STREAM C */
setup_stream_contents(123, "CCCCCCCCCC");
streams[2].id = 'C';
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[2].id, lsquic_stream_tosend_offset(&streams[2]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[2]),
lsquic_stream_tosend_sz(&streams[2]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[2]);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[2],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
/* Reset A */
len = pf->pf_gen_rst_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out), 'A', 133, 0);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[0],
QUIC_FRAME_RST_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
/* STREAM D */
setup_stream_contents(123, "DDDDDDDDDD");
streams[3].id = 'D';
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[3].id, lsquic_stream_tosend_offset(&streams[3]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[3]),
lsquic_stream_tosend_sz(&streams[3]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[3]);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[3],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
/* STREAM E */
setup_stream_contents(123, "EEEEEEEEEE");
streams[4].id = 'E';
len = pf->pf_gen_stream_frame(packet_out->po_data + packet_out->po_data_sz,
lsquic_packet_out_avail(packet_out),
streams[4].id, lsquic_stream_tosend_offset(&streams[4]),
Latest changes - [API Change] lsquic_engine_connect() returns pointer to the connection object. - [API Change] Add lsquic_conn_get_engine() to get engine object from connection object. - [API Change] Add lsquic_conn_status() to query connection status. - [API Change] Add add lsquic_conn_set_ctx(). - [API Change] Add new timestamp format, e.g. 2017-03-21 13:43:46.671345 - [OPTIMIZATION] Process handshake STREAM frames as soon as packet arrives. - [OPTIMIZATION] Do not compile expensive send controller sanity check by default. - [OPTIMIZATION] Add fast path to gquic_be_gen_reg_pkt_header. - [OPTIMIZATION] Only make squeeze function call if necessary. - [OPTIMIZATION] Speed up Q039 ACK frame parsing. - [OPTIMIZATION] Fit most used elements of packet_out into first 64 bytes. - [OPTIMIZATION] Keep track of scheduled bytes instead of calculating. - [OPTIMIZATION] Prefetch next unacked packet when processing ACK. - [OPTIMIZATION] Leverage fact that ACK ranges and unacked list are. ordered. - [OPTIMIZATION] Reduce function pointer use for STREAM frame generation - Fix: reset incoming streams that arrive after we send GOAWAY. - Fix: delay client on_new_conn() call until connection is fully set up. - Fixes to buffered packets logic: splitting, STREAM frame elision. - Fix: do not dispatch on_write callback if no packets are available. - Fix WINDOW_UPDATE send and resend logic. - Fix STREAM frame extension code. - Fix: Drop unflushed data when stream is reset. - Switch to tracking CWND using bytes rather than packets. - Fix TCP friendly adjustment in cubic. - Fix: do not generate invalid STOP_WAITING frames during high packet loss. - Pacer fixes.
2018-02-26 21:01:16 +00:00
lsquic_stream_tosend_fin(&streams[4]),
lsquic_stream_tosend_sz(&streams[4]),
2017-09-22 21:00:03 +00:00
(gsf_read_f) lsquic_stream_tosend_read,
&streams[4]);
lsquic_packet_out_add_stream(packet_out, &enpub.enp_mm, &streams[4],
Latest changes - [API Change] Sendfile-like functionality is gone. The stream no longer opens files and deals with file descriptors. (Among other things, this makes the code more portable.) Three writing functions are provided: lsquic_stream_write lsquic_stream_writev lsquic_stream_writef (NEW) lsquic_stream_writef() is given an abstract reader that has function pointers for size() and read() functions which the user can implement. This is the most flexible way. lsquic_stream_write() and lsquic_stream_writev() are now both implemented as wrappers around lsquic_stream_writef(). - [OPTIMIZATION] When writing to stream, be it within or without the on_write() callback, place data directly into packet buffer, bypassing auxiliary data structures. This reduces amount of memory required, for the amount of data that can be written is limited by the congestion window. To support writes outside the on_write() callback, we keep N outgoing packet buffers per connection which can be written to by any stream. One half of these are reserved for the highest priority stream(s), the other half for all other streams. This way, low-priority streams cannot write instead of high-priority streams and, on the other hand, low-priority streams get a chance to send their packets out. The algorithm is as follows: - When user writes to stream outside of the callback: - If this is the highest priority stream, place it onto the reserved N/2 queue or fail. (The actual size of this queue is dynamic -- MAX(N/2, CWND) -- rather than N/2, allowing high-priority streams to write as much as can be sent.) - If the stream is not the highest priority, try to place the data onto the reserved N/2 queue or fail. - When tick occurs *and* more packets can be scheduled: - Transfer packets from the high N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for highest-priority streams, placing resulting packets directly onto the scheduled queue. - If more scheduling is allowed: - Transfer packets from the low N/2 queue to the scheduled queue. - If more scheduling is allowed: - Call on_write callbacks for non-highest-priority streams, placing resulting packets directly onto the scheduled queue The number N is currently 20, but it could be varied based on resource usage. - If stream is created due to incoming headers, make headers readable from on_new. - Outgoing packets are no longer marked non-writeable to prevent placing more than one STREAM frame from the same stream into a single packet. This property is maintained via code flow and an explicit check. Packets for stream data are allocated using a special function. - STREAM frame elision is cheaper, as we only perform it if a reset stream has outgoing packets referencing it. - lsquic_packet_out_t is smaller, as stream_rec elements are now inside a union.
2017-10-31 13:35:58 +00:00
QUIC_FRAME_STREAM, packet_out->po_data_sz, len);
2017-09-22 21:00:03 +00:00
packet_out->po_data_sz += len;
packet_out->po_frame_types = (1 << QUIC_FRAME_STREAM) | (1 << QUIC_FRAME_RST_STREAM);
}
/* Reset streams A, C, and E: */
streams[0].stream_flags |= STREAM_RST_SENT;
streams[2].stream_flags |= STREAM_RST_SENT;
streams[4].stream_flags |= STREAM_RST_SENT;
if (chop_regen)
lsquic_packet_out_chop_regen(packet_out);
lsquic_packet_out_elide_reset_stream_frames(packet_out, UINT64_MAX);
2017-09-22 21:00:03 +00:00
assert(ref_out->po_data_sz == packet_out->po_data_sz + (chop_regen ? 5 : 0));
assert(ref_out->po_regen_sz == packet_out->po_regen_sz + (chop_regen ? 5 : 0));
if (chop_regen)
assert(0 == memcmp(ref_out->po_data + 5, packet_out->po_data, packet_out->po_data_sz));
else
assert(0 == memcmp(ref_out->po_data, packet_out->po_data, packet_out->po_data_sz));
assert(1 == streams[0].n_unacked); /* Still has RST outstanding */
assert(1 == streams[1].n_unacked);
assert(0 == streams[2].n_unacked);
assert(1 == streams[3].n_unacked);
assert(0 == streams[4].n_unacked);
assert(packet_out->po_frame_types == ((1 << QUIC_FRAME_STREAM) | (1 << QUIC_FRAME_RST_STREAM)));
frec = lsquic_pofi_first(&pofi, packet_out);
if (!chop_regen)
frec = lsquic_pofi_next(&pofi);
assert(frec->fe_stream == &streams[1]);
assert(frec->fe_frame_type == QUIC_FRAME_STREAM);
assert(frec->fe_off == b_off - (chop_regen ? 5 : 0));
2017-09-22 21:00:03 +00:00
frec = lsquic_pofi_next(&pofi);
assert(frec->fe_stream == &streams[0]);
assert(frec->fe_frame_type == QUIC_FRAME_RST_STREAM);
frec = lsquic_pofi_next(&pofi);
assert(frec->fe_stream == &streams[3]);
assert(frec->fe_frame_type == QUIC_FRAME_STREAM);
assert(frec->fe_off == d_off - (chop_regen ? 5 : 0));
2017-09-22 21:00:03 +00:00
frec = lsquic_pofi_next(&pofi);
assert(!frec);
2017-09-22 21:00:03 +00:00
lsquic_packet_out_destroy(packet_out, &enpub, NULL);
lsquic_packet_out_destroy(ref_out, &enpub, NULL);
2017-09-22 21:00:03 +00:00
lsquic_mm_cleanup(&enpub.enp_mm);
}
int
main (void)
{
/* TODO-ENDIAN: test with every PF */
elide_single_stream_frame();
shrink_packet_post_elision();
2017-09-22 21:00:03 +00:00
elide_three_stream_frames(0);
elide_three_stream_frames(1);
return 0;
}