2018-04-02 19:17:56 +00:00
|
|
|
/* Copyright (c) 2017 - 2018 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>
|
2018-03-12 22:25:01 +00:00
|
|
|
#ifndef WIN32
|
2017-09-22 21:00:03 +00:00
|
|
|
#include <sys/time.h>
|
2018-03-12 22:25:01 +00:00
|
|
|
#endif
|
2017-09-22 21:00:03 +00:00
|
|
|
#include <sys/queue.h>
|
|
|
|
|
|
|
|
#include "lsquic.h"
|
2018-05-02 13:42:32 +00:00
|
|
|
#include "lshpack.h"
|
2017-09-22 21:00:03 +00:00
|
|
|
#include "lsquic_mm.h"
|
|
|
|
#include "lsquic_frame_common.h"
|
|
|
|
#include "lsquic_frame_writer.h"
|
|
|
|
|
|
|
|
|
|
|
|
static struct {
|
|
|
|
size_t sz;
|
|
|
|
size_t max;
|
|
|
|
unsigned char buf[0x1000];
|
|
|
|
} output;
|
|
|
|
|
|
|
|
|
|
|
|
#define reset_output(max_) do { \
|
|
|
|
output.sz = 0; \
|
|
|
|
if (max_) \
|
|
|
|
output.max = max_; \
|
|
|
|
else \
|
|
|
|
output.max = sizeof(output.buf);\
|
|
|
|
} while (0)
|
|
|
|
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
output_write (struct lsquic_stream *stream, const void *buf, size_t sz)
|
|
|
|
{
|
|
|
|
if (output.max - output.sz < sz)
|
|
|
|
{
|
|
|
|
errno = ENOBUFS;
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
memcpy(output.buf + output.sz, buf, sz);
|
|
|
|
output.sz += sz;
|
|
|
|
|
|
|
|
return sz;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
#define IOV(v) { .iov_base = (v), .iov_len = sizeof(v) - 1, }
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_max_frame_size (void)
|
|
|
|
{
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-09-22 21:00:03 +00:00
|
|
|
struct lsquic_mm mm;
|
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
unsigned max_size;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_init(&mm);
|
|
|
|
|
|
|
|
for (max_size = 1; max_size < 6 /* one settings frame */; ++max_size)
|
|
|
|
{
|
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, max_size, &henc,
|
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
|
|
|
output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
assert(!fw);
|
|
|
|
}
|
|
|
|
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, max_size, &henc, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
assert(fw);
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_one_header (void)
|
|
|
|
{
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-09-22 21:00:03 +00:00
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 0x200, &henc, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
reset_output(0);
|
|
|
|
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV(":status"), .value = IOV("302") },
|
|
|
|
};
|
|
|
|
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = 1,
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_headers(fw, 12345, &headers, 0, 100);
|
|
|
|
assert(0 == s);
|
|
|
|
|
|
|
|
struct http_frame_header fh;
|
|
|
|
struct http_prio_frame prio_frame;
|
|
|
|
|
|
|
|
assert(4 + sizeof(struct http_frame_header) + sizeof(struct http_prio_frame) == output.sz);
|
|
|
|
|
|
|
|
memcpy(&fh, output.buf, sizeof(fh));
|
|
|
|
assert(4 + sizeof(struct http_prio_frame) == hfh_get_length(&fh));
|
|
|
|
assert(HTTP_FRAME_HEADERS == fh.hfh_type);
|
|
|
|
assert((HFHF_END_HEADERS|HFHF_PRIORITY) == fh.hfh_flags);
|
|
|
|
assert(fh.hfh_stream_id[0] == 0);
|
|
|
|
assert(fh.hfh_stream_id[1] == 0);
|
|
|
|
assert(fh.hfh_stream_id[2] == 0x30);
|
|
|
|
assert(fh.hfh_stream_id[3] == 0x39);
|
|
|
|
|
|
|
|
memcpy(&prio_frame, output.buf + sizeof(struct http_frame_header),
|
|
|
|
sizeof(struct http_prio_frame));
|
|
|
|
|
|
|
|
assert(prio_frame.hpf_stream_id[0] == 0);
|
|
|
|
assert(prio_frame.hpf_stream_id[1] == 0);
|
|
|
|
assert(prio_frame.hpf_stream_id[2] == 0);
|
|
|
|
assert(prio_frame.hpf_stream_id[3] == 0);
|
|
|
|
assert(prio_frame.hpf_weight == 100 - 1);
|
|
|
|
|
|
|
|
assert(0 == memcmp(output.buf + sizeof(struct http_frame_header) +
|
|
|
|
sizeof(struct http_prio_frame), "\x48\x82\x64\x02", 4));
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2017-10-12 15:26:01 +00:00
|
|
|
static void
|
|
|
|
test_oversize_header (void)
|
|
|
|
{
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-10-12 15:26:01 +00:00
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
const size_t big_len = 100 * 1000;
|
|
|
|
char *value;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-10-12 15:26:01 +00:00
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 0x200, &henc, output_write, 0);
|
2017-10-12 15:26:01 +00:00
|
|
|
reset_output(0);
|
|
|
|
|
|
|
|
value = malloc(big_len);
|
|
|
|
memset(value, 'A', big_len);
|
|
|
|
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV(":status"), .value = IOV("302") },
|
|
|
|
{ .name = IOV("some-header"),
|
|
|
|
.value = { .iov_base = value, .iov_len = big_len, } },
|
|
|
|
};
|
|
|
|
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = sizeof(header_arr) / sizeof(header_arr[0]),
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_headers(fw, 12345, &headers, 0, 100);
|
|
|
|
assert(-1 == s);
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-10-12 15:26:01 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
free(value);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2017-09-22 21:00:03 +00:00
|
|
|
static void
|
|
|
|
test_continuations (void)
|
|
|
|
{
|
|
|
|
struct lsquic_frame_writer *fw;
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-09-22 21:00:03 +00:00
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 6, &henc, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
reset_output(0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
perl tools/henc.pl :status 302 x-some-header some-value | hexdump -C
|
|
|
|
00000000 48 82 64 02 40 8a f2 b2 0f 49 56 9c a3 90 b6 7f |H.d.@....IV.....|
|
|
|
|
00000010 87 41 e9 2a dd c7 45 a5 |.A.*..E.|
|
|
|
|
*/
|
|
|
|
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV(":status"), .value = IOV("302") },
|
|
|
|
{. name = IOV("x-some-header"), .value = IOV("some-value") },
|
|
|
|
};
|
|
|
|
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = 2,
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_headers(fw, 12345, &headers, 0, 100);
|
|
|
|
assert(0 == s);
|
|
|
|
|
|
|
|
/* Expected payload is 5 bytes of http_prio_frame and 24 bytes of
|
|
|
|
* compressed headers, split into 4 15-byte chunks (9-byte header
|
|
|
|
* 6-byte payload) and 1 14-byte chunk (9-byte header and 5-byte
|
|
|
|
* payload).
|
|
|
|
*/
|
|
|
|
unsigned char expected_buf[] = {
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06, /* 1 */
|
|
|
|
/* Type: */ HTTP_FRAME_HEADERS,
|
|
|
|
/* Flags: */ HFHF_PRIORITY,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x30, 0x39,
|
|
|
|
/* Payload (priority info): */
|
|
|
|
0x00, 0x00, 0x00, 0x00, 100 - 1,
|
|
|
|
/* Payload (headers): */
|
|
|
|
0x48,
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06, /* 2 */
|
|
|
|
/* Type: */ HTTP_FRAME_CONTINUATION,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x30, 0x39,
|
|
|
|
/* Payload (headers): */
|
|
|
|
0x82, 0x64, 0x02, 0x40, 0x8A, 0xF2,
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06, /* 3 */
|
|
|
|
/* Type: */ HTTP_FRAME_CONTINUATION,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x30, 0x39,
|
|
|
|
/* Payload (headers): */
|
|
|
|
0xb2, 0x0f, 0x49, 0x56, 0x9c, 0xa3,
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06, /* 4 */
|
|
|
|
/* Type: */ HTTP_FRAME_CONTINUATION,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x30, 0x39,
|
|
|
|
/* Payload (headers): */
|
|
|
|
0x90, 0xb6, 0x7f, 0x87, 0x41, 0xe9,
|
|
|
|
/* Length: */ 0x00, 0x00, 0x05, /* 5 */
|
|
|
|
/* Type: */ HTTP_FRAME_CONTINUATION,
|
|
|
|
/* Flags: */ HFHF_END_HEADERS,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x30, 0x39,
|
|
|
|
/* Payload (headers): */
|
|
|
|
0x2a, 0xdd, 0xc7, 0x45, 0xa5,
|
|
|
|
};
|
|
|
|
|
|
|
|
assert(sizeof(expected_buf) == output.sz);
|
|
|
|
|
|
|
|
assert(0 == memcmp(output.buf + 0, expected_buf + 0, 15));
|
|
|
|
assert(0 == memcmp(output.buf + 15, expected_buf + 15, 15));
|
|
|
|
assert(0 == memcmp(output.buf + 30, expected_buf + 30, 15));
|
|
|
|
assert(0 == memcmp(output.buf + 45, expected_buf + 45, 15));
|
|
|
|
assert(0 == memcmp(output.buf + 60, expected_buf + 60, 14));
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_settings_short (void)
|
|
|
|
{
|
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 7, NULL, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
struct lsquic_http2_setting settings[] = { { 1, 2, }, { 3, 4, } };
|
|
|
|
s = lsquic_frame_writer_write_settings(fw, settings, 2);
|
|
|
|
assert(0 == s);
|
|
|
|
const unsigned char exp_buf[] = {
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06,
|
|
|
|
/* Type: */ HTTP_FRAME_SETTINGS,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x00, 0x00,
|
|
|
|
/* Payload: */ 0x00, 0x01, 0x00, 0x00, 0x00, 0x02,
|
|
|
|
/* Length: */ 0x00, 0x00, 0x06,
|
|
|
|
/* Type: */ HTTP_FRAME_SETTINGS,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x00, 0x00,
|
|
|
|
/* Payload: */ 0x00, 0x03, 0x00, 0x00, 0x00, 0x04,
|
|
|
|
};
|
|
|
|
assert(output.sz == sizeof(exp_buf));
|
|
|
|
assert(0 == memcmp(output.buf, exp_buf, sizeof(exp_buf)));
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
struct lsquic_http2_setting settings[] = { { 1, 2, }, { 3, 4, } };
|
|
|
|
s = lsquic_frame_writer_write_settings(fw, settings, 0);
|
|
|
|
assert(-1 == s);
|
|
|
|
assert(EINVAL == errno);
|
|
|
|
}
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_settings_normal (void)
|
|
|
|
{
|
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 0, NULL, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
struct lsquic_http2_setting settings[] = { { 1, 2, }, { 3, 4, } };
|
|
|
|
s = lsquic_frame_writer_write_settings(fw, settings, 2);
|
|
|
|
assert(0 == s);
|
|
|
|
const unsigned char exp_buf[] = {
|
|
|
|
/* Length: */ 0x00, 0x00, 0x0C,
|
|
|
|
/* Type: */ HTTP_FRAME_SETTINGS,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x00, 0x00,
|
|
|
|
/* Payload: */ 0x00, 0x01, 0x00, 0x00, 0x00, 0x02,
|
|
|
|
/* Payload: */ 0x00, 0x03, 0x00, 0x00, 0x00, 0x04,
|
|
|
|
};
|
|
|
|
assert(output.sz == sizeof(exp_buf));
|
|
|
|
assert(0 == memcmp(output.buf, exp_buf, sizeof(exp_buf)));
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
struct lsquic_http2_setting settings[] = { { 1, 2, }, { 3, 4, } };
|
|
|
|
s = lsquic_frame_writer_write_settings(fw, settings, 0);
|
|
|
|
assert(-1 == s);
|
|
|
|
assert(EINVAL == errno);
|
|
|
|
}
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Gotta override these so that LSQUIC_LOG_CONN_ID in lsquic_frame_writer.c
|
|
|
|
* works.
|
|
|
|
*/
|
|
|
|
lsquic_cid_t
|
|
|
|
lsquic_conn_id (const lsquic_conn_t *lconn)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
lsquic_conn_t *
|
|
|
|
lsquic_stream_conn (const lsquic_stream_t *stream)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_priority (void)
|
|
|
|
{
|
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 6, NULL, output_write, 0);
|
2017-09-22 21:00:03 +00:00
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_priority(fw, 3, 0, 1UL << 31, 256);
|
|
|
|
assert(s < 0); /* Invalid dependency stream ID */
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_priority(fw, 3, 0, 1, 0);
|
|
|
|
assert(s < 0); /* Invalid priority stream ID */
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_priority(fw, 3, 0, 1, 257);
|
|
|
|
assert(s < 0); /* Invalid priority stream ID */
|
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
s = lsquic_frame_writer_write_priority(fw, 3, 0, 1, 256);
|
|
|
|
assert(0 == s);
|
|
|
|
const unsigned char exp_buf[] = {
|
|
|
|
/* Length: */ 0x00, 0x00, 5,
|
|
|
|
/* Type: */ HTTP_FRAME_PRIORITY,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x00, 0x03,
|
|
|
|
/* Dep stream Id: */0x00, 0x00, 0x00, 0x01,
|
|
|
|
/* Weight: */ 0xFF,
|
|
|
|
};
|
|
|
|
assert(output.sz == sizeof(exp_buf));
|
|
|
|
assert(0 == memcmp(output.buf, exp_buf, sizeof(exp_buf)));
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
reset_output(0);
|
|
|
|
s = lsquic_frame_writer_write_priority(fw, 20, 1, 100, 256);
|
|
|
|
assert(0 == s);
|
|
|
|
const unsigned char exp_buf[] = {
|
|
|
|
/* Length: */ 0x00, 0x00, 5,
|
|
|
|
/* Type: */ HTTP_FRAME_PRIORITY,
|
|
|
|
/* Flags: */ 0x00,
|
|
|
|
/* Stream Id: */ 0x00, 0x00, 0x00, 0x14,
|
|
|
|
/* Dep stream Id: */0x80, 0x00, 0x00, 0x64,
|
|
|
|
/* Weight: */ 0xFF,
|
|
|
|
};
|
|
|
|
assert(output.sz == sizeof(exp_buf));
|
|
|
|
assert(0 == memcmp(output.buf, exp_buf, sizeof(exp_buf)));
|
|
|
|
}
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_errors (void)
|
|
|
|
{
|
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
struct lsquic_mm mm;
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-09-22 21:00:03 +00:00
|
|
|
int s;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 0x200, &henc, output_write, 1);
|
2017-09-22 21:00:03 +00:00
|
|
|
reset_output(0);
|
|
|
|
|
|
|
|
{
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV(":status"), .value = IOV("200") },
|
|
|
|
{ .name = IOV("Content-type"), .value = IOV("text/html") },
|
|
|
|
};
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = 2,
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
s = lsquic_frame_writer_write_headers(fw, 12345, &headers, 0, 80);
|
|
|
|
assert(-1 == s);
|
|
|
|
assert(EINVAL == errno);
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV(":status"), .value = IOV("200") },
|
|
|
|
{ .name = IOV("content-type"), .value = IOV("text/html") },
|
|
|
|
};
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = 2,
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
lsquic_frame_writer_max_header_list_size(fw, 40);
|
|
|
|
s = lsquic_frame_writer_write_headers(fw, 12345, &headers, 0, 80);
|
|
|
|
assert(-1 == s);
|
|
|
|
assert(EMSGSIZE == errno);
|
|
|
|
}
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
test_push_promise (void)
|
|
|
|
{
|
2018-05-02 13:42:32 +00:00
|
|
|
struct lshpack_enc henc;
|
2017-09-22 21:00:03 +00:00
|
|
|
struct lsquic_frame_writer *fw;
|
|
|
|
int s;
|
|
|
|
struct lsquic_mm mm;
|
|
|
|
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_init(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_init(&mm);
|
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
|
|
|
fw = lsquic_frame_writer_new(&mm, NULL, 0x200, &henc, output_write, 1);
|
2017-09-22 21:00:03 +00:00
|
|
|
reset_output(0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
perl tools/hpack.pl :method GET :path /index.html :authority www.example.com :scheme https x-some-header some-value| hexdump -C
|
|
|
|
00000000 82 85 41 8c f1 e3 c2 e5 f2 3a 6b a0 ab 90 f4 ff |..A......:k.....|
|
|
|
|
00000010 87 40 8a f2 b2 0f 49 56 9c a3 90 b6 7f 87 41 e9 |.@....IV......A.|
|
|
|
|
00000020 2a dd c7 45 a5 |*..E.|
|
|
|
|
*/
|
|
|
|
|
|
|
|
const unsigned char exp_headers[] = {
|
|
|
|
0x82, 0x85, 0x41, 0x8c, 0xf1, 0xe3, 0xc2, 0xe5, 0xf2, 0x3a,
|
|
|
|
0x6b, 0xa0, 0xab, 0x90, 0xf4, 0xff, 0x87, 0x40, 0x8a, 0xf2,
|
|
|
|
0xb2, 0x0f, 0x49, 0x56, 0x9c, 0xa3, 0x90, 0xb6, 0x7f, 0x87,
|
|
|
|
0x41, 0xe9, 0x2a, 0xdd, 0xc7, 0x45, 0xa5,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct iovec path = IOV("/index.html");
|
|
|
|
struct iovec host = IOV("www.example.com");
|
|
|
|
|
|
|
|
struct lsquic_http_header header_arr[] =
|
|
|
|
{
|
|
|
|
{ .name = IOV("x-some-header"), .value = IOV("some-value") },
|
|
|
|
};
|
|
|
|
|
|
|
|
struct lsquic_http_headers headers = {
|
|
|
|
.count = 1,
|
|
|
|
.headers = header_arr,
|
|
|
|
};
|
|
|
|
|
|
|
|
s = lsquic_frame_writer_write_promise(fw, 12345, 0xEEEE, &path, &host,
|
|
|
|
&headers);
|
|
|
|
assert(0 == s);
|
|
|
|
|
|
|
|
struct http_frame_header fh;
|
|
|
|
struct http_push_promise_frame push_frame;
|
|
|
|
|
|
|
|
assert(sizeof(exp_headers) + sizeof(struct http_frame_header) +
|
|
|
|
sizeof(struct http_push_promise_frame) == output.sz);
|
|
|
|
|
|
|
|
memcpy(&fh, output.buf, sizeof(fh));
|
|
|
|
assert(sizeof(exp_headers) + sizeof(struct http_push_promise_frame) == hfh_get_length(&fh));
|
|
|
|
assert(HTTP_FRAME_PUSH_PROMISE == fh.hfh_type);
|
|
|
|
assert(HFHF_END_HEADERS == fh.hfh_flags);
|
|
|
|
assert(fh.hfh_stream_id[0] == 0);
|
|
|
|
assert(fh.hfh_stream_id[1] == 0);
|
|
|
|
assert(fh.hfh_stream_id[2] == 0x30);
|
|
|
|
assert(fh.hfh_stream_id[3] == 0x39);
|
|
|
|
|
|
|
|
memcpy(&push_frame, output.buf + sizeof(struct http_frame_header),
|
|
|
|
sizeof(struct http_push_promise_frame));
|
|
|
|
|
|
|
|
assert(push_frame.hppf_promised_id[0] == 0);
|
|
|
|
assert(push_frame.hppf_promised_id[1] == 0);
|
|
|
|
assert(push_frame.hppf_promised_id[2] == 0xEE);
|
|
|
|
assert(push_frame.hppf_promised_id[3] == 0xEE);
|
|
|
|
|
|
|
|
assert(0 == memcmp(output.buf + sizeof(struct http_frame_header) +
|
|
|
|
sizeof(struct http_push_promise_frame), exp_headers,
|
|
|
|
sizeof(exp_headers)));
|
|
|
|
|
|
|
|
lsquic_frame_writer_destroy(fw);
|
2018-05-02 13:42:32 +00:00
|
|
|
lshpack_enc_cleanup(&henc);
|
2017-09-22 21:00:03 +00:00
|
|
|
lsquic_mm_cleanup(&mm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
int
|
|
|
|
main (void)
|
|
|
|
{
|
|
|
|
test_one_header();
|
2017-10-12 15:26:01 +00:00
|
|
|
test_oversize_header();
|
2017-09-22 21:00:03 +00:00
|
|
|
test_continuations();
|
|
|
|
test_settings_normal();
|
|
|
|
test_settings_short();
|
|
|
|
test_priority();
|
|
|
|
test_push_promise();
|
|
|
|
test_errors();
|
|
|
|
test_max_frame_size();
|
|
|
|
return 0;
|
|
|
|
}
|