litespeed-quic/test/http_client.c

521 lines
16 KiB
C
Raw Normal View History

2017-09-22 21:00:03 +00:00
/* Copyright (c) 2017 LiteSpeed Technologies Inc. See LICENSE. */
/*
* http_client.c -- A simple HTTP/QUIC client
*/
#include <arpa/inet.h>
#include <netinet/in.h>
#include <assert.h>
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/queue.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include "lsquic.h"
#include "test_common.h"
#include "prog.h"
#include "../src/liblsquic/lsquic_logger.h"
/* This is used to exercise generating and sending of priority frames */
static int randomly_reprioritize_streams;
/* If this file descriptor is open, the client will accept server push and
* dump the contents here. See -u flag.
*/
static int promise_fd = -1;
struct lsquic_conn_ctx;
struct path_elem {
TAILQ_ENTRY(path_elem) next_pe;
const char *path;
};
struct http_client_ctx {
TAILQ_HEAD(, lsquic_conn_ctx)
conn_ctxs;
struct service_port *sport;
const char *hostname;
const char *method;
const char *payload;
char payload_size[20];
/* hcc_path_elems holds a list of paths which are to be requested from
* the server. Each new request gets the next path from the list (the
* iterator is stored in hcc_cur_pe); when the end is reached, the
* iterator wraps around.
*/
TAILQ_HEAD(, path_elem) hcc_path_elems;
struct path_elem *hcc_cur_pe;
unsigned hcc_total_n_reqs;
unsigned hcc_reqs_per_conn;
unsigned hcc_concurrency;
unsigned hcc_n_open_conns;
enum {
HCC_DISCARD_RESPONSE = (1 << 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
HCC_SEEN_FIN = (1 << 1),
HCC_ABORT_ON_INCOMPLETE = (1 << 2),
2017-09-22 21:00:03 +00:00
} hcc_flags;
struct prog *prog;
};
struct lsquic_conn_ctx {
TAILQ_ENTRY(lsquic_conn_ctx) next_ch;
lsquic_conn_t *conn;
struct http_client_ctx *client_ctx;
unsigned ch_n_reqs; /* This number gets decremented as streams are closed and
* incremented as push promises are accepted.
*/
};
static void
create_connections (struct http_client_ctx *client_ctx)
{
while (client_ctx->hcc_n_open_conns < client_ctx->hcc_concurrency &&
client_ctx->hcc_total_n_reqs > 0)
if (0 != prog_connect(client_ctx->prog))
{
LSQ_ERROR("connection failed");
exit(EXIT_FAILURE);
}
}
static lsquic_conn_ctx_t *
http_client_on_new_conn (void *stream_if_ctx, lsquic_conn_t *conn)
{
struct http_client_ctx *client_ctx = stream_if_ctx;
lsquic_conn_ctx_t *conn_h = calloc(1, sizeof(*conn_h));
conn_h->conn = conn;
conn_h->client_ctx = client_ctx;
conn_h->ch_n_reqs = client_ctx->hcc_total_n_reqs <
client_ctx->hcc_reqs_per_conn ?
client_ctx->hcc_total_n_reqs : client_ctx->hcc_reqs_per_conn;
client_ctx->hcc_total_n_reqs -= conn_h->ch_n_reqs;
TAILQ_INSERT_TAIL(&client_ctx->conn_ctxs, conn_h, next_ch);
++conn_h->client_ctx->hcc_n_open_conns;
lsquic_conn_make_stream(conn);
return conn_h;
}
static void
http_client_on_conn_closed (lsquic_conn_t *conn)
{
lsquic_conn_ctx_t *conn_h = lsquic_conn_get_ctx(conn);
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
enum LSQUIC_CONN_STATUS status;
char errmsg[80];
status = lsquic_conn_status(conn, errmsg, sizeof(errmsg));
LSQ_INFO("Connection closed. Status: %d. Message: %s", status,
errmsg[0] ? errmsg : "<not set>");
#ifndef NDEBUG
if (conn_h->client_ctx->hcc_flags & HCC_ABORT_ON_INCOMPLETE)
assert(conn_h->client_ctx->hcc_flags & HCC_SEEN_FIN);
#endif
2017-09-22 21:00:03 +00:00
TAILQ_REMOVE(&conn_h->client_ctx->conn_ctxs, conn_h, next_ch);
--conn_h->client_ctx->hcc_n_open_conns;
create_connections(conn_h->client_ctx);
if (0 == conn_h->client_ctx->hcc_n_open_conns)
{
LSQ_INFO("All connections are closed: stop engine");
prog_stop(conn_h->client_ctx->prog);
}
free(conn_h);
}
struct lsquic_stream_ctx {
lsquic_stream_t *stream;
struct http_client_ctx *client_ctx;
const char *path;
enum {
HEADERS_SENT = (1 << 0),
} sh_flags;
unsigned count;
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
struct lsquic_reader reader;
2017-09-22 21:00:03 +00:00
};
static lsquic_stream_ctx_t *
http_client_on_new_stream (void *stream_if_ctx, lsquic_stream_t *stream)
{
const int pushed = lsquic_stream_is_pushed(stream);
if (pushed)
{
LSQ_INFO("not accepting server push");
lsquic_stream_refuse_push(stream);
return NULL;
}
lsquic_stream_ctx_t *st_h = calloc(1, sizeof(*st_h));
st_h->stream = stream;
st_h->client_ctx = stream_if_ctx;
if (st_h->client_ctx->hcc_cur_pe)
{
st_h->client_ctx->hcc_cur_pe = TAILQ_NEXT(
st_h->client_ctx->hcc_cur_pe, next_pe);
if (!st_h->client_ctx->hcc_cur_pe) /* Wrap around */
st_h->client_ctx->hcc_cur_pe =
TAILQ_FIRST(&st_h->client_ctx->hcc_path_elems);
}
else
st_h->client_ctx->hcc_cur_pe = TAILQ_FIRST(
&st_h->client_ctx->hcc_path_elems);
st_h->path = st_h->client_ctx->hcc_cur_pe->path;
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
if (st_h->client_ctx->payload)
{
st_h->reader.lsqr_read = test_reader_read;
st_h->reader.lsqr_size = test_reader_size;
st_h->reader.lsqr_ctx = create_lsquic_reader_ctx(st_h->client_ctx->payload);
if (!st_h->reader.lsqr_ctx)
exit(1);
}
else
st_h->reader.lsqr_ctx = NULL;
2017-09-22 21:00:03 +00:00
LSQ_INFO("created new stream, path: %s", st_h->path);
lsquic_stream_wantwrite(stream, 1);
return st_h;
}
static void
send_headers (lsquic_stream_ctx_t *st_h)
{
lsquic_http_header_t headers_arr[] = {
{
.name = { .iov_base = ":method", .iov_len = 7, },
.value = { .iov_base = (void *) st_h->client_ctx->method,
.iov_len = strlen(st_h->client_ctx->method), },
},
{
.name = { .iov_base = ":scheme", .iov_len = 7, },
.value = { .iov_base = "HTTP", .iov_len = 4, }
},
{
.name = { .iov_base = ":path", .iov_len = 5, },
.value = { .iov_base = (void *) st_h->path,
.iov_len = strlen(st_h->path), },
},
{
.name = { ":authority", 10, },
.value = { .iov_base = (void *) st_h->client_ctx->hostname,
.iov_len = strlen(st_h->client_ctx->hostname), },
},
/*
{
.name = { "host", 4 },
.value = { .iov_base = (void *) st_h->client_ctx->hostname,
.iov_len = strlen(st_h->client_ctx->hostname), },
},
*/
{
.name = { .iov_base = "user-agent", .iov_len = 10, },
.value = { .iov_base = (char *) st_h->client_ctx->prog->prog_settings.es_ua,
.iov_len = strlen(st_h->client_ctx->prog->prog_settings.es_ua), },
},
/* The following headers only gets sent if there is request payload: */
{
.name = { .iov_base = "content-type", .iov_len = 12, },
.value = { .iov_base = "application/octet-stream", .iov_len = 24, },
},
{
.name = { .iov_base = "content-length", .iov_len = 14, },
.value = { .iov_base = (void *) st_h->client_ctx->payload_size,
.iov_len = strlen(st_h->client_ctx->payload_size), },
},
};
lsquic_http_headers_t headers = {
.count = sizeof(headers_arr) / sizeof(headers_arr[0]),
.headers = headers_arr,
};
if (!st_h->client_ctx->payload)
headers.count -= 2;
if (0 != lsquic_stream_send_headers(st_h->stream, &headers,
st_h->client_ctx->payload == NULL))
{
LSQ_ERROR("cannot send headers: %s", strerror(errno));
exit(1);
}
}
static void
http_client_on_write (lsquic_stream_t *stream, lsquic_stream_ctx_t *st_h)
{
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
ssize_t nw;
2017-09-22 21:00:03 +00:00
if (st_h->sh_flags & HEADERS_SENT)
{
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
if (st_h->client_ctx->payload && test_reader_size(st_h->reader.lsqr_ctx) > 0)
2017-09-22 21:00:03 +00:00
{
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
nw = lsquic_stream_writef(stream, &st_h->reader);
if (nw < 0)
2017-09-22 21:00:03 +00:00
{
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
LSQ_ERROR("write error: %s", strerror(errno));
2017-09-22 21:00:03 +00:00
exit(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
if (test_reader_size(st_h->reader.lsqr_ctx) > 0)
{
lsquic_stream_wantwrite(stream, 1);
}
else
{
lsquic_stream_shutdown(stream, 1);
lsquic_stream_wantread(stream, 1);
}
}
else
{
lsquic_stream_shutdown(stream, 1);
lsquic_stream_wantread(stream, 1);
2017-09-22 21:00:03 +00:00
}
}
else
{
st_h->sh_flags |= HEADERS_SENT;
send_headers(st_h);
}
}
static void
http_client_on_read (lsquic_stream_t *stream, lsquic_stream_ctx_t *st_h)
{
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
struct http_client_ctx *const client_ctx = st_h->client_ctx;
2017-09-22 21:00:03 +00:00
ssize_t nread;
unsigned old_prio, new_prio;
unsigned char buf[0x200];
unsigned nreads = 0;
do
{
nread = lsquic_stream_read(stream, buf, sizeof(buf));
if (nread > 0)
{
if (!(client_ctx->hcc_flags & HCC_DISCARD_RESPONSE))
write(STDOUT_FILENO, buf, nread);
if (randomly_reprioritize_streams && (st_h->count++ & 0x3F) == 0)
{
old_prio = lsquic_stream_priority(stream);
new_prio = random() & 0xFF;
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
#ifndef NDEBUG
const int s =
#endif
lsquic_stream_set_priority(stream, new_prio);
2017-09-22 21:00:03 +00:00
assert(s == 0);
LSQ_NOTICE("changed stream %u priority from %u to %u",
lsquic_stream_id(stream), old_prio, new_prio);
}
}
else if (0 == nread)
{
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
client_ctx->hcc_flags |= HCC_SEEN_FIN;
2017-09-22 21:00:03 +00:00
lsquic_stream_shutdown(stream, 0);
break;
}
else if (client_ctx->prog->prog_settings.es_rw_once && EWOULDBLOCK == errno)
{
LSQ_NOTICE("emptied the buffer in 'once' mode");
break;
}
else
{
LSQ_ERROR("could not read: %s", strerror(errno));
exit(2);
}
}
while (client_ctx->prog->prog_settings.es_rw_once
&& nreads++ < 3 /* Emulate just a few reads */);
}
static void
http_client_on_close (lsquic_stream_t *stream, lsquic_stream_ctx_t *st_h)
{
const int pushed = lsquic_stream_is_pushed(stream);
if (pushed)
{
assert(NULL == st_h);
return;
}
LSQ_INFO("%s called", __func__);
lsquic_conn_t *conn = lsquic_stream_conn(stream);
lsquic_conn_ctx_t *conn_h;
TAILQ_FOREACH(conn_h, &st_h->client_ctx->conn_ctxs, next_ch)
if (conn_h->conn == conn)
break;
assert(conn_h);
--conn_h->ch_n_reqs;
if (0 == conn_h->ch_n_reqs)
{
LSQ_INFO("all requests completed, closing connection");
lsquic_conn_close(conn_h->conn);
}
else
lsquic_conn_make_stream(conn);
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
if (st_h->reader.lsqr_ctx)
destroy_lsquic_reader_ctx(st_h->reader.lsqr_ctx);
2017-09-22 21:00:03 +00:00
free(st_h);
}
const struct lsquic_stream_if http_client_if = {
.on_new_conn = http_client_on_new_conn,
.on_conn_closed = http_client_on_conn_closed,
.on_new_stream = http_client_on_new_stream,
.on_read = http_client_on_read,
.on_write = http_client_on_write,
.on_close = http_client_on_close,
};
static void
usage (const char *prog)
{
const char *const slash = strrchr(prog, '/');
if (slash)
prog = slash + 1;
printf(
"Usage: %s [opts]\n"
"\n"
"Options:\n"
" -p PATH Path to request. May be specified more than once.\n"
" -n CONNS Number of concurrent connections. Defaults to 1.\n"
" -r NREQS Total number of requests to send. Defaults to 1.\n"
" -R NREQS Maximum number of requests per single connection. Some\n"
" connections will have fewer requests than this.\n"
" -m METHOD Method. Defaults to GET.\n"
" -P PAYLOAD Name of the file that contains payload to be used in the\n"
" request. This adds two more headers to the request:\n"
" content-type: application/octet-stream and\n"
" content-length\n"
" -K Discard server response\n"
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
" -I Abort on incomplete reponse from server\n"
2017-09-22 21:00:03 +00:00
, prog);
}
int
main (int argc, char **argv)
{
int opt, s;
struct http_client_ctx client_ctx;
struct stat st;
struct path_elem *pe;
struct sport_head sports;
struct prog prog;
TAILQ_INIT(&sports);
memset(&client_ctx, 0, sizeof(client_ctx));
client_ctx.hcc_concurrency = 1;
TAILQ_INIT(&client_ctx.hcc_path_elems);
TAILQ_INIT(&client_ctx.conn_ctxs);
client_ctx.hostname = "localhost";
client_ctx.method = "GET";
client_ctx.hcc_concurrency = 1;
client_ctx.hcc_reqs_per_conn = 1;
client_ctx.hcc_total_n_reqs = 1;
client_ctx.prog = &prog;
prog_init(&prog, LSENG_HTTP, &sports, &http_client_if, &client_ctx);
while (-1 != (opt = getopt(argc, argv, PROG_OPTS "r:R:Ku:EP:M:n:H:p:h")))
{
switch (opt) {
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
case 'I':
client_ctx.hcc_flags |= HCC_ABORT_ON_INCOMPLETE;
break;
2017-09-22 21:00:03 +00:00
case 'K':
client_ctx.hcc_flags |= HCC_DISCARD_RESPONSE;
break;
case 'u': /* Accept p<U>sh promise */
promise_fd = open(optarg, O_WRONLY|O_CREAT|O_TRUNC, 0644);
if (promise_fd < 0)
{
perror("open");
exit(1);
}
prog.prog_settings.es_support_push = 1; /* Pokes into prog */
break;
case 'E': /* E: randomly reprioritize str<E>ams. Now, that's
* pretty random. :)
*/
randomly_reprioritize_streams = 1;
break;
case 'n':
client_ctx.hcc_concurrency = atoi(optarg);
break;
case 'P':
client_ctx.payload = optarg;
if (0 != stat(optarg, &st))
{
perror("stat");
exit(2);
}
sprintf(client_ctx.payload_size, "%jd", (intmax_t) st.st_size);
break;
case 'M':
client_ctx.method = optarg;
break;
case 'r':
client_ctx.hcc_total_n_reqs = atoi(optarg);
break;
case 'R':
client_ctx.hcc_reqs_per_conn = atoi(optarg);
break;
case 'H':
client_ctx.hostname = optarg;
prog.prog_hostname = optarg; /* Pokes into prog */
break;
case 'p':
pe = calloc(1, sizeof(*pe));
pe->path = optarg;
TAILQ_INSERT_TAIL(&client_ctx.hcc_path_elems, pe, next_pe);
break;
case 'h':
usage(argv[0]);
prog_print_common_options(&prog, stdout);
exit(0);
default:
if (0 != prog_set_opt(&prog, opt, optarg))
exit(1);
}
}
if (TAILQ_EMPTY(&client_ctx.hcc_path_elems))
{
fprintf(stderr, "Specify at least one path using -p option\n");
exit(1);
}
if (0 != prog_prep(&prog))
{
LSQ_ERROR("could not prep");
exit(EXIT_FAILURE);
}
create_connections(&client_ctx);
LSQ_DEBUG("entering event loop");
s = prog_run(&prog);
prog_cleanup(&prog);
if (promise_fd >= 0)
(void) close(promise_fd);
exit(0 == s ? EXIT_SUCCESS : EXIT_FAILURE);
}