litespeed-quic/test/test_common.c

1145 lines
28 KiB
C
Raw Normal View History

/* Copyright (c) 2017 - 2018 LiteSpeed Technologies Inc. See LICENSE. */
#if __GNUC__
2017-09-22 21:00:03 +00:00
#define _GNU_SOURCE /* For struct in6_pktinfo */
#endif
2017-09-22 21:00:03 +00:00
#include <assert.h>
#include <errno.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#if defined(__APPLE__)
# define __APPLE_USE_RFC_3542 1
#endif
#ifndef WIN32
2017-09-22 21:00:03 +00:00
#include <netinet/in.h>
#include <arpa/inet.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <unistd.h>
#else
#include <Windows.h>
#include <WinSock2.h>
#include <MSWSock.h>
#include<io.h>
#pragma warning(disable:4996)//posix name deprecated
#define close closesocket
#endif
#include <sys/stat.h>
#include <sys/queue.h>
2017-09-22 21:00:03 +00:00
#include <fcntl.h>
#include <event2/event.h>
#include "test_common.h"
#include "lsquic.h"
#include "prog.h"
#include "../src/liblsquic/lsquic_logger.h"
#define MAX(a, b) ((a) > (b) ? (a) : (b))
#ifndef WIN32
# define SOCKET_TYPE int
# define CLOSE_SOCKET close
# define CHAR_CAST
#else
/* XXX detect these using cmake? */
# define HAVE_IP_DONTFRAG 1
# define HAVE_IP_MTU_DISCOVER 1
# define SOCKET_TYPE SOCKET
# define CLOSE_SOCKET closesocket
# define CHAR_CAST (char *)
#endif
2017-09-22 21:00:03 +00:00
#if __linux__
# define NDROPPED_SZ CMSG_SPACE(sizeof(uint32_t)) /* SO_RXQ_OVFL */
#else
# define NDROPPED_SZ 0
#endif
#if __linux__ && defined(IP_RECVORIGDSTADDR)
# define DST_MSG_SZ sizeof(struct sockaddr_in)
#elif WIN32
# define DST_MSG_SZ sizeof(struct sockaddr_in)
2017-09-22 21:00:03 +00:00
#elif __linux__
# define DST_MSG_SZ sizeof(struct in_pktinfo)
#else
# define DST_MSG_SZ sizeof(struct sockaddr_in)
#endif
#define MAX_PACKET_SZ 1370
#define CTL_SZ (CMSG_SPACE(MAX(DST_MSG_SZ, \
sizeof(struct in6_pktinfo))) + NDROPPED_SZ)
/* There are `n_alloc' elements in `vecs', `local_addresses', and
* `peer_addresses' arrays. `ctlmsg_data' is n_alloc * CTL_SZ. Each packets
* gets a single `vecs' element that points somewhere into `packet_data'.
*
* `n_alloc' is calculated at run-time based on the socket's receive buffer
* size.
*/
struct packets_in
{
unsigned char *packet_data;
unsigned char *ctlmsg_data;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
struct iovec *vecs;
#else
WSABUF *vecs;
#endif
2017-09-22 21:00:03 +00:00
struct sockaddr_storage *local_addresses,
*peer_addresses;
unsigned n_alloc;
unsigned data_sz;
};
#if WIN32
LPFN_WSARECVMSG pfnWSARecvMsg;
GUID recvGuid = WSAID_WSARECVMSG;
LPFN_WSASENDMSG pfnWSASendMsg;
GUID sendGuid = WSAID_WSASENDMSG;
CRITICAL_SECTION initLock;
LONG initialized = 0;
static void getExtensionPtrs()
{
if (InterlockedCompareExchange(&initialized, 1, 0) == 0)
{
InitializeCriticalSection(&initLock);
}
EnterCriticalSection(&initLock);
if(pfnWSARecvMsg == NULL|| pfnWSASendMsg == NULL)
{
SOCKET sock= socket(PF_INET, SOCK_DGRAM, 0);
DWORD dwBytes;
int rc = 0;
if (pfnWSARecvMsg == NULL)
{
rc = WSAIoctl(sock, SIO_GET_EXTENSION_FUNCTION_POINTER, &recvGuid,
sizeof(recvGuid), &pfnWSARecvMsg, sizeof(pfnWSARecvMsg),
&dwBytes, NULL, NULL);
}
if (rc != SOCKET_ERROR)
{
if (pfnWSASendMsg == NULL)
{
rc = WSAIoctl(sock, SIO_GET_EXTENSION_FUNCTION_POINTER,
&sendGuid, sizeof(sendGuid), &pfnWSASendMsg,
sizeof(pfnWSASendMsg), &dwBytes, NULL, NULL);
}
}
if (rc == SOCKET_ERROR)
{
LSQ_ERROR("Can't get extension function pointers: %d",
WSAGetLastError());
}
closesocket(sock);
}
LeaveCriticalSection(&initLock);
}
#endif
2017-09-22 21:00:03 +00:00
static struct packets_in *
allocate_packets_in (SOCKET_TYPE fd)
2017-09-22 21:00:03 +00:00
{
struct packets_in *packs_in;
unsigned n_alloc;
socklen_t opt_len;
int recvsz;
opt_len = sizeof(recvsz);
if (0 != getsockopt(fd, SOL_SOCKET, SO_RCVBUF, (void*)&recvsz, &opt_len))
2017-09-22 21:00:03 +00:00
{
LSQ_ERROR("getsockopt failed: %s", strerror(errno));
return NULL;
}
n_alloc = (unsigned) recvsz / MAX_PACKET_SZ * 2;
LSQ_INFO("socket buffer size: %d bytes; max # packets is set to %u",
recvsz, n_alloc);
packs_in = malloc(sizeof(*packs_in));
packs_in->data_sz = recvsz;
packs_in->n_alloc = n_alloc;
packs_in->packet_data = malloc(recvsz);
packs_in->ctlmsg_data = malloc(n_alloc * CTL_SZ);
packs_in->vecs = malloc(n_alloc * sizeof(packs_in->vecs[0]));
packs_in->local_addresses = malloc(n_alloc * sizeof(packs_in->local_addresses[0]));
packs_in->peer_addresses = malloc(n_alloc * sizeof(packs_in->peer_addresses[0]));
return packs_in;
}
static void
free_packets_in (struct packets_in *packs_in)
{
free(packs_in->peer_addresses);
free(packs_in->local_addresses);
free(packs_in->ctlmsg_data);
free(packs_in->vecs);
free(packs_in->packet_data);
free(packs_in);
}
void
sport_destroy (struct service_port *sport)
{
if (sport->ev)
{
event_del(sport->ev);
event_free(sport->ev);
}
if (sport->fd >= 0)
(void) CLOSE_SOCKET(sport->fd);
2017-09-22 21:00:03 +00:00
if (sport->packs_in)
free_packets_in(sport->packs_in);
free(sport);
}
struct service_port *
sport_new (const char *optarg, struct prog *prog)
{
struct service_port *const sport = malloc(sizeof(*sport));
#if __linux__
sport->n_dropped = 0;
sport->drop_init = 0;
#endif
sport->ev = NULL;
sport->packs_in = NULL;
sport->fd = -1;
char *const addr = strdup(optarg);
#if __linux__
char *if_name;
if_name = strrchr(addr, ',');
if (if_name)
{
strncpy(sport->if_name, if_name + 1, sizeof(sport->if_name) - 1);
sport->if_name[ sizeof(sport->if_name) - 1 ] = '\0';
*if_name = '\0';
}
else
sport->if_name[0] = '\0';
#endif
char *port = strrchr(addr, ':');
if (!port)
goto err;
*port = '\0';
++port;
if ((uintptr_t) port - (uintptr_t) addr > sizeof(sport->host))
goto err;
memcpy(sport->host, addr, port - addr);
struct sockaddr_in *const sa4 = (void *) &sport->sas;
struct sockaddr_in6 *const sa6 = (void *) &sport->sas;
if (inet_pton(AF_INET, addr, &sa4->sin_addr)) {
sa4->sin_family = AF_INET;
sa4->sin_port = htons(atoi(port));
} else if (memset(sa6, 0, sizeof(*sa6)),
inet_pton(AF_INET6, addr, &sa6->sin6_addr)) {
sa6->sin6_family = AF_INET6;
sa6->sin6_port = htons(atoi(port));
} else
goto err;
free(addr);
sport->sp_prog = prog;
return sport;
err:
free(sport);
free(addr);
return NULL;
}
/* Replace IP address part of `sa' with that provided in ancillary messages
* in `msg'.
*/
static void
proc_ancillary (
#ifndef WIN32
struct msghdr
#else
WSAMSG
#endif
*msg, struct sockaddr_storage *storage
2017-09-22 21:00:03 +00:00
#if __linux__
, uint32_t *n_dropped
#endif
)
{
const struct in6_pktinfo *in6_pkt;
struct cmsghdr *cmsg;
for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg))
{
if (cmsg->cmsg_level == IPPROTO_IP &&
cmsg->cmsg_type ==
#if __linux__ && defined(IP_RECVORIGDSTADDR)
IP_ORIGDSTADDR
#elif __linux__ || WIN32
2017-09-22 21:00:03 +00:00
IP_PKTINFO
#else
IP_RECVDSTADDR
#endif
)
{
#if __linux__ && defined(IP_RECVORIGDSTADDR)
memcpy(storage, CMSG_DATA(cmsg), sizeof(struct sockaddr_in));
#elif WIN32
const struct in_pktinfo *in_pkt;
in_pkt = (void *) WSA_CMSG_DATA(cmsg);
((struct sockaddr_in *) storage)->sin_addr = in_pkt->ipi_addr;
2017-09-22 21:00:03 +00:00
#elif __linux__
const struct in_pktinfo *in_pkt;
in_pkt = (void *) CMSG_DATA(cmsg);
((struct sockaddr_in *) storage)->sin_addr = in_pkt->ipi_addr;
#else
memcpy(&((struct sockaddr_in *) storage)->sin_addr,
CMSG_DATA(cmsg), sizeof(struct in_addr));
#endif
}
else if (cmsg->cmsg_level == IPPROTO_IPV6 &&
cmsg->cmsg_type == IPV6_PKTINFO)
{
#ifndef WIN32
2017-09-22 21:00:03 +00:00
in6_pkt = (void *) CMSG_DATA(cmsg);
#else
in6_pkt = (void *) WSA_CMSG_DATA(cmsg);
#endif
2017-09-22 21:00:03 +00:00
((struct sockaddr_in6 *) storage)->sin6_addr =
in6_pkt->ipi6_addr;
}
#if __linux__
else if (cmsg->cmsg_level == SOL_SOCKET &&
cmsg->cmsg_type == SO_RXQ_OVFL)
memcpy(n_dropped, CMSG_DATA(cmsg), sizeof(*n_dropped));
#endif
}
}
struct read_iter
{
struct service_port *ri_sport;
unsigned ri_idx; /* Current element */
unsigned ri_off; /* Offset into packet_data */
};
enum rop { ROP_OK, ROP_NOROOM, ROP_ERROR, };
static enum rop
read_one_packet (struct read_iter *iter)
{
unsigned char *ctl_buf;
struct packets_in *packs_in;
#if __linux__
uint32_t n_dropped;
#endif
#ifndef WIN32
2017-09-22 21:00:03 +00:00
ssize_t nread;
#else
DWORD nread;
int socket_ret;
#endif
2017-09-22 21:00:03 +00:00
struct sockaddr_storage *local_addr;
struct service_port *sport;
sport = iter->ri_sport;
packs_in = sport->packs_in;
if (iter->ri_idx >= packs_in->n_alloc ||
iter->ri_off + MAX_PACKET_SZ > packs_in->data_sz)
{
LSQ_DEBUG("out of room in packets_in");
return ROP_NOROOM;
}
#ifndef WIN32
2017-09-22 21:00:03 +00:00
packs_in->vecs[iter->ri_idx].iov_base = packs_in->packet_data + iter->ri_off;
packs_in->vecs[iter->ri_idx].iov_len = MAX_PACKET_SZ;
#else
packs_in->vecs[iter->ri_idx].buf = (char*)packs_in->packet_data + iter->ri_off;
packs_in->vecs[iter->ri_idx].len = MAX_PACKET_SZ;
#endif
2017-09-22 21:00:03 +00:00
ctl_buf = packs_in->ctlmsg_data + iter->ri_idx * CTL_SZ;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
struct msghdr msg = {
.msg_name = &packs_in->peer_addresses[iter->ri_idx],
.msg_namelen = sizeof(packs_in->peer_addresses[iter->ri_idx]),
.msg_iov = &packs_in->vecs[iter->ri_idx],
.msg_iovlen = 1,
.msg_control = ctl_buf,
.msg_controllen = CTL_SZ,
};
nread = recvmsg(sport->fd, &msg, 0);
if (-1 == nread) {
if (!(EAGAIN == errno || EWOULDBLOCK == errno))
LSQ_ERROR("recvmsg: %s", strerror(errno));
return ROP_ERROR;
}
#else
WSAMSG msg = {
.name = (LPSOCKADDR)&packs_in->peer_addresses[iter->ri_idx],
.namelen = sizeof(packs_in->peer_addresses[iter->ri_idx]),
.lpBuffers = &packs_in->vecs[iter->ri_idx],
.dwBufferCount = 1,
.Control = {CTL_SZ,(char*)ctl_buf}
};
socket_ret = pfnWSARecvMsg(sport->fd, &msg, &nread, NULL, NULL);
if (SOCKET_ERROR == socket_ret) {
if (WSAEWOULDBLOCK != WSAGetLastError())
LSQ_ERROR("recvmsg: %d", WSAGetLastError());
return ROP_ERROR;
}
#endif
2017-09-22 21:00:03 +00:00
local_addr = &packs_in->local_addresses[iter->ri_idx];
memcpy(local_addr, &sport->sas, sizeof(*local_addr));
#if __linux__
n_dropped = 0;
#endif
proc_ancillary(&msg, local_addr
#if __linux__
, &n_dropped
#endif
);
#if __linux__
if (sport->drop_init)
{
if (sport->n_dropped < n_dropped)
LSQ_INFO("dropped %u packets", n_dropped - sport->n_dropped);
}
else
sport->drop_init = 1;
sport->n_dropped = n_dropped;
#endif
#ifndef WIN32
2017-09-22 21:00:03 +00:00
packs_in->vecs[iter->ri_idx].iov_len = nread;
#else
packs_in->vecs[iter->ri_idx].len = nread;
#endif
2017-09-22 21:00:03 +00:00
iter->ri_off += nread;
iter->ri_idx += 1;
return ROP_OK;
}
static void
read_handler (evutil_socket_t fd, short flags, void *ctx)
2017-09-22 21:00:03 +00:00
{
struct service_port *sport = ctx;
lsquic_engine_t *const engine = sport->engine;
struct packets_in *packs_in = sport->packs_in;
struct read_iter iter;
unsigned n, n_batches;
enum rop rop;
n_batches = 0;
iter.ri_sport = sport;
do
{
iter.ri_off = 0;
iter.ri_idx = 0;
do
rop = read_one_packet(&iter);
while (ROP_OK == rop);
n_batches += iter.ri_idx > 0;
for (n = 0; n < iter.ri_idx; ++n)
if (0 != lsquic_engine_packet_in(engine,
#ifndef WIN32
2017-09-22 21:00:03 +00:00
packs_in->vecs[n].iov_base,
packs_in->vecs[n].iov_len,
#else
(const unsigned char *) packs_in->vecs[n].buf,
packs_in->vecs[n].len,
#endif
2017-09-22 21:00:03 +00:00
(struct sockaddr *) &packs_in->local_addresses[n],
(struct sockaddr *) &packs_in->peer_addresses[n],
sport))
break;
}
while (ROP_NOROOM == rop);
if (n_batches)
n += packs_in->n_alloc * (n_batches - 1);
if (!prog_is_stopped())
[API Change, OPTIMIZATION] Only process conns that need to be processed The API is simplified: do not expose the user code to several queues. A "connection queue" is now an internal concept. The user processes connections using the single function lsquic_engine_process_conns(). When this function is called, only those connections are processed that need to be processed. A connection needs to be processed when: 1. New incoming packets have been fed to the connection. 2. User wants to read from a stream that is readable. 3. User wants to write to a stream that is writeable. 4. There are buffered packets that can be sent out. (This means that the user wrote to a stream outside of the lsquic library callback.) 5. A control frame (such as BLOCKED) needs to be sent out. 6. A stream needs to be serviced or delayed stream needs to be created. 7. An alarm rings. 8. Pacer timer expires. To achieve this, the library places the connections into two priority queues (min heaps): 1. Tickable Queue; and 2. Advisory Tick Time queue (ATTQ). Each time lsquic_engine_process_conns() is called, the Tickable Queue is emptied. After the connections have been ticked, they are queried again: if a connection is not being closed, it is placed either in the Tickable Queue if it is ready to be ticked again or it is placed in the Advisory Tick Time Queue. It is assumed that a connection always has at least one timer set (the idle alarm). The connections in the Tickable Queue are arranged in the least recently ticked order. This lets connections that have been quiet longer to get their packets scheduled first. This change means that the library no longer needs to be ticked periodically. The user code can query the library when is the next tick event and schedule it exactly. When connections are processed, only the tickable connections are processed, not *all* the connections. When there are no tick events, it means that no timer event is necessary -- only the file descriptor READ event is active. The following are improvements and simplifications that have been triggered: - Queue of connections with incoming packets is gone. - "Pending Read/Write Events" Queue is gone (along with its history and progress checks). This queue has become the Tickable Queue. - The connection hash no longer needs to track the connection insertion order.
2018-04-09 13:39:38 +00:00
prog_process_conns(sport->sp_prog);
2017-09-22 21:00:03 +00:00
LSQ_DEBUG("read %u packet%.*s in %u batch%s", n, n != 1, "s", n_batches, n_batches != 1 ? "es" : "");
}
static int
add_to_event_loop (struct service_port *sport, struct event_base *eb)
{
sport->ev = event_new(eb, sport->fd, EV_READ|EV_PERSIST, read_handler,
sport);
if (sport->ev)
{
event_add(sport->ev, NULL);
return 0;
}
else
return -1;
}
int
sport_init_client (struct service_port *sport, struct lsquic_engine *engine,
struct event_base *eb)
{
const struct sockaddr *sa_peer = (struct sockaddr *) &sport->sas;
int saved_errno, s;
#ifndef WIN32
int flags;
#endif
SOCKET_TYPE sockfd;
2017-09-22 21:00:03 +00:00
socklen_t socklen;
union {
struct sockaddr_in sin;
struct sockaddr_in6 sin6;
} u;
struct sockaddr *sa_local = (struct sockaddr *) &u;
char addr_str[0x20];
switch (sa_peer->sa_family)
{
case AF_INET:
socklen = sizeof(struct sockaddr_in);
u.sin.sin_family = AF_INET;
u.sin.sin_addr.s_addr = INADDR_ANY;
u.sin.sin_port = 0;
break;
case AF_INET6:
socklen = sizeof(struct sockaddr_in6);
memset(&u.sin6, 0, sizeof(u.sin6));
u.sin6.sin6_family = AF_INET6;
break;
default:
errno = EINVAL;
return -1;
}
#if WIN32
getExtensionPtrs();
#endif
2017-09-22 21:00:03 +00:00
sockfd = socket(sa_peer->sa_family, SOCK_DGRAM, 0);
if (-1 == sockfd)
return -1;
if (0 != bind(sockfd, sa_local, socklen)) {
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
/* Make socket non-blocking */
#ifndef WIN32
2017-09-22 21:00:03 +00:00
flags = fcntl(sockfd, F_GETFL);
if (-1 == flags) {
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
flags |= O_NONBLOCK;
if (0 != fcntl(sockfd, F_SETFL, flags)) {
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
#else
{
u_long on = 1;
ioctlsocket(sockfd, FIONBIO, &on);
}
#endif
2017-09-22 21:00:03 +00:00
#if LSQUIC_DONTFRAG_SUPPORTED
2017-09-22 21:00:03 +00:00
if (sport->sp_flags & SPORT_DONT_FRAGMENT)
{
if (AF_INET == sa_local->sa_family)
{
int on;
2017-09-22 21:00:03 +00:00
#if __linux__
on = IP_PMTUDISC_DO;
s = setsockopt(sockfd, IPPROTO_IP, IP_MTU_DISCOVER, &on,
sizeof(on));
#elif WIN32
on = 1;
s = setsockopt(sockfd, IPPROTO_IP, IP_DONTFRAGMENT, (char*)&on, sizeof(on));
2017-09-22 21:00:03 +00:00
#else
on = 1;
s = setsockopt(sockfd, IPPROTO_IP, IP_DONTFRAG, &on, sizeof(on));
#endif
if (0 != s)
{
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
}
}
#endif
2017-09-22 21:00:03 +00:00
if (sport->sp_flags & SPORT_SET_SNDBUF)
{
s = setsockopt(sockfd, SOL_SOCKET, SO_SNDBUF,
CHAR_CAST &sport->sp_sndbuf, sizeof(sport->sp_sndbuf));
2017-09-22 21:00:03 +00:00
if (0 != s)
{
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
}
if (sport->sp_flags & SPORT_SET_RCVBUF)
{
s = setsockopt(sockfd, SOL_SOCKET, SO_RCVBUF,
CHAR_CAST &sport->sp_rcvbuf, sizeof(sport->sp_rcvbuf));
2017-09-22 21:00:03 +00:00
if (0 != s)
{
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
}
if (0 != getsockname(sockfd, sa_local, &socklen))
{
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
sport->packs_in = allocate_packets_in(sockfd);
if (!sport->packs_in)
{
saved_errno = errno;
CLOSE_SOCKET(sockfd);
2017-09-22 21:00:03 +00:00
errno = saved_errno;
return -1;
}
switch (sa_local->sa_family) {
case AF_INET:
LSQ_DEBUG("local address: %s:%d",
inet_ntop(AF_INET, &u.sin.sin_addr, addr_str, sizeof(addr_str)),
ntohs(u.sin.sin_port));
break;
}
sport->engine = engine;
sport->fd = sockfd;
return add_to_event_loop(sport, eb);
}
static void
setup_control_msg (
#ifndef WIN32
struct msghdr
#else
WSAMSG
#endif
*msg, const struct lsquic_out_spec *spec,
2017-09-22 21:00:03 +00:00
unsigned char *buf, size_t bufsz)
{
struct cmsghdr *cmsg;
struct sockaddr_in *local_sa;
struct sockaddr_in6 *local_sa6;
#if __linux__ || __APPLE__ || WIN32
2017-09-22 21:00:03 +00:00
struct in_pktinfo info;
#endif
struct in6_pktinfo info6;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
msg->msg_control = buf;
msg->msg_controllen = bufsz;
#else
msg->Control.buf = (char*)buf;
msg->Control.len = bufsz;
#endif
2017-09-22 21:00:03 +00:00
cmsg = CMSG_FIRSTHDR(msg);
if (AF_INET == spec->dest_sa->sa_family)
{
local_sa = (struct sockaddr_in *) spec->local_sa;
#if __linux__ || __APPLE__
2017-09-22 21:00:03 +00:00
memset(&info, 0, sizeof(info));
info.ipi_spec_dst = local_sa->sin_addr;
cmsg->cmsg_level = IPPROTO_IP;
cmsg->cmsg_type = IP_PKTINFO;
cmsg->cmsg_len = CMSG_LEN(sizeof(info));
memcpy(CMSG_DATA(cmsg), &info, sizeof(info));
#elif WIN32
memset(&info, 0, sizeof(info));
info.ipi_addr = local_sa->sin_addr;
cmsg->cmsg_level = IPPROTO_IP;
cmsg->cmsg_type = IP_PKTINFO;
cmsg->cmsg_len = CMSG_LEN(sizeof(info));
memcpy(WSA_CMSG_DATA(cmsg), &info, sizeof(info));
2017-09-22 21:00:03 +00:00
#else
cmsg->cmsg_level = IPPROTO_IP;
cmsg->cmsg_type = IP_SENDSRCADDR;
cmsg->cmsg_len = CMSG_LEN(sizeof(local_sa->sin_addr));
memcpy(CMSG_DATA(cmsg), &local_sa->sin_addr,
sizeof(local_sa->sin_addr));
#endif
}
else
{
local_sa6 = (struct sockaddr_in6 *) spec->local_sa;
memset(&info6, 0, sizeof(info6));
info6.ipi6_addr = local_sa6->sin6_addr;
cmsg->cmsg_level = IPPROTO_IPV6;
cmsg->cmsg_type = IPV6_PKTINFO;
cmsg->cmsg_len = CMSG_LEN(sizeof(info6));
#ifndef WIN32
2017-09-22 21:00:03 +00:00
memcpy(CMSG_DATA(cmsg), &info6, sizeof(info6));
#else
memcpy(WSA_CMSG_DATA(cmsg), &info6, sizeof(info6));
#endif
2017-09-22 21:00:03 +00:00
}
#ifndef WIN32
2017-09-22 21:00:03 +00:00
msg->msg_controllen = cmsg->cmsg_len;
#else
msg->Control.len = cmsg->cmsg_len;
#endif
2017-09-22 21:00:03 +00:00
}
static int
send_packets_one_by_one (const struct lsquic_out_spec *specs, unsigned count)
{
const struct service_port *sport;
unsigned n;
int s = 0;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
struct msghdr msg;
#else
DWORD bytes;
WSAMSG msg;
#endif
2017-09-22 21:00:03 +00:00
union {
/* cmsg(3) recommends union for proper alignment */
#if __linux__ || WIN32
# define SIZE1 sizeof(struct in_pktinfo)
2017-09-22 21:00:03 +00:00
#else
# define SIZE1 sizeof(struct in_addr)
2017-09-22 21:00:03 +00:00
#endif
unsigned char buf[
CMSG_SPACE(MAX(SIZE1, sizeof(struct in6_pktinfo)))];
2017-09-22 21:00:03 +00:00
struct cmsghdr cmsg;
} ancil;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
struct iovec iov;
#else
WSABUF iov;
#endif
2017-09-22 21:00:03 +00:00
if (0 == count)
return 0;
for (n = 0; n < count; ++n)
{
sport = specs[n].peer_ctx;
#ifndef WIN32
2017-09-22 21:00:03 +00:00
iov.iov_base = (void *) specs[n].buf;
iov.iov_len = specs[n].sz;
msg.msg_name = (void *) specs[n].dest_sa;
msg.msg_namelen = (AF_INET == specs[n].dest_sa->sa_family ?
sizeof(struct sockaddr_in) :
sizeof(struct sockaddr_in6)),
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_flags = 0;
#else
iov.buf = (void *) specs[n].buf;
iov.len = specs[n].sz;
msg.name = (void *) specs[n].dest_sa;
msg.namelen = (AF_INET == specs[n].dest_sa->sa_family ?
sizeof(struct sockaddr_in) :
sizeof(struct sockaddr_in6)),
msg.lpBuffers = &iov;
msg.dwBufferCount = 1;
msg.dwFlags = 0;
#endif
2017-09-22 21:00:03 +00:00
if (sport->sp_flags & SPORT_SERVER)
setup_control_msg(&msg, &specs[n], ancil.buf, sizeof(ancil.buf));
else
{
#ifndef WIN32
2017-09-22 21:00:03 +00:00
msg.msg_control = NULL;
msg.msg_controllen = 0;
#else
msg.Control.buf = NULL;
msg.Control.len = 0;
#endif
2017-09-22 21:00:03 +00:00
}
#ifndef WIN32
2017-09-22 21:00:03 +00:00
s = sendmsg(sport->fd, &msg, 0);
#else
s = pfnWSASendMsg(sport->fd, &msg, 0, &bytes, NULL, NULL);
#endif
2017-09-22 21:00:03 +00:00
if (s < 0)
{
#ifndef WIN32
2017-09-22 21:00:03 +00:00
LSQ_INFO("sendto failed: %s", strerror(errno));
#else
LSQ_INFO("sendto failed: %s", WSAGetLastError());
#endif
2017-09-22 21:00:03 +00:00
break;
}
}
if (n > 0)
return n;
else if (s < 0)
return -1;
else
return 0;
}
int
sport_packets_out (void *ctx, const struct lsquic_out_spec *specs,
unsigned count)
{
return send_packets_one_by_one(specs, count);
}
int
set_engine_option (struct lsquic_engine_settings *settings,
int *version_cleared, const char *name)
{
int len;
const char *val = strchr(name, '=');
if (!val)
return -1;
len = val - name;
++val;
switch (len)
{
case 2:
if (0 == strncmp(name, "ua", 2))
{
settings->es_ua = val;
return 0;
}
break;
case 4:
if (0 == strncmp(name, "cfcw", 4))
{
settings->es_cfcw = atoi(val);
return 0;
}
if (0 == strncmp(name, "sfcw", 4))
{
settings->es_sfcw = atoi(val);
return 0;
}
if (0 == strncmp(name, "srej", 4))
{
settings->es_support_srej = atoi(val);
return 0;
}
break;
case 7:
if (0 == strncmp(name, "version", 7))
{
if (!*version_cleared)
{
*version_cleared = 1;
settings->es_versions = 0;
}
const enum lsquic_version ver = lsquic_str2ver(val, strlen(val));
if (ver < N_LSQVER)
2017-09-22 21:00:03 +00:00
{
settings->es_versions |= 1 << ver;
2017-09-22 21:00:03 +00:00
return 0;
}
}
else if (0 == strncmp(name, "rw_once", 7))
{
settings->es_rw_once = atoi(val);
return 0;
}
break;
case 8:
if (0 == strncmp(name, "max_cfcw", 8))
{
settings->es_max_cfcw = atoi(val);
return 0;
}
if (0 == strncmp(name, "max_sfcw", 8))
{
settings->es_max_sfcw = atoi(val);
return 0;
}
break;
case 10:
if (0 == strncmp(name, "honor_prst", 10))
{
settings->es_honor_prst = atoi(val);
return 0;
}
break;
case 12:
if (0 == strncmp(name, "idle_conn_to", 12))
{
settings->es_idle_conn_to = atoi(val);
return 0;
}
if (0 == strncmp(name, "silent_close", 12))
{
settings->es_silent_close = atoi(val);
return 0;
}
if (0 == strncmp(name, "support_nstp", 12))
{
settings->es_support_nstp = atoi(val);
return 0;
}
if (0 == strncmp(name, "pace_packets", 12))
{
settings->es_pace_packets = atoi(val);
return 0;
}
break;
case 13:
if (0 == strncmp(name, "support_tcid0", 13))
{
settings->es_support_tcid0 = atoi(val);
return 0;
}
break;
case 14:
if (0 == strncmp(name, "max_streams_in", 14))
{
settings->es_max_streams_in = atoi(val);
return 0;
}
if (0 == strncmp(name, "progress_check", 14))
{
settings->es_progress_check = atoi(val);
return 0;
}
break;
case 16:
if (0 == strncmp(name, "proc_time_thresh", 16))
{
settings->es_proc_time_thresh = atoi(val);
return 0;
}
break;
case 20:
if (0 == strncmp(name, "max_header_list_size", 20))
{
settings->es_max_header_list_size = atoi(val);
return 0;
}
break;
}
return -1;
}
#define MAX_PACKOUT_BUF_SZ MAX_PACKET_SZ
struct packout_buf
{
SLIST_ENTRY(packout_buf) next_free_pb;
};
void
pba_init (struct packout_buf_allocator *pba, unsigned max)
{
SLIST_INIT(&pba->free_packout_bufs);
pba->max = max;
pba->n_out = 0;
}
void *
pba_allocate (void *packout_buf_allocator, size_t size)
{
struct packout_buf_allocator *const pba = packout_buf_allocator;
struct packout_buf *pb;
if (size > MAX_PACKOUT_BUF_SZ)
{
fprintf(stderr, "packout buf size too large: %zd", size);
abort();
}
if (pba->max && pba->n_out >= pba->max)
{
LSQ_DEBUG("# outstanding packout bufs reached the limit of %u, "
"returning NULL", pba->max);
return NULL;
}
pb = SLIST_FIRST(&pba->free_packout_bufs);
if (pb)
SLIST_REMOVE_HEAD(&pba->free_packout_bufs, next_free_pb);
else
pb = malloc(MAX_PACKOUT_BUF_SZ);
if (pb)
++pba->n_out;
return pb;
}
void
pba_release (void *packout_buf_allocator, void *obj)
{
struct packout_buf_allocator *const pba = packout_buf_allocator;
struct packout_buf *const pb = obj;
SLIST_INSERT_HEAD(&pba->free_packout_bufs, pb, next_free_pb);
--pba->n_out;
}
void
pba_cleanup (struct packout_buf_allocator *pba)
{
unsigned n = 0;
struct packout_buf *pb;
if (pba->n_out)
LSQ_WARN("%u packout bufs outstanding at deinit", pba->n_out);
while ((pb = SLIST_FIRST(&pba->free_packout_bufs)))
{
SLIST_REMOVE_HEAD(&pba->free_packout_bufs, next_free_pb);
free(pb);
++n;
}
LSQ_INFO("pba deinitialized, freed %u packout bufs", n);
}
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 reader_ctx
{
size_t file_size;
size_t nread;
int fd;
};
size_t
test_reader_size (void *void_ctx)
{
struct reader_ctx *const ctx = void_ctx;
return ctx->file_size - ctx->nread;
}
size_t
test_reader_read (void *void_ctx, void *buf, size_t count)
{
struct reader_ctx *const ctx = void_ctx;
ssize_t nread;
if (count > test_reader_size(ctx))
count = test_reader_size(ctx);
#ifndef WIN32
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
nread = read(ctx->fd, buf, count);
#else
nread = _read(ctx->fd, buf, count);
#endif
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 (nread >= 0)
{
ctx->nread += nread;
return nread;
}
else
{
LSQ_WARN("%s: error reading from file: %s", __func__, strerror(errno));
ctx->nread = ctx->file_size = 0;
return 0;
}
}
struct reader_ctx *
create_lsquic_reader_ctx (const char *filename)
{
int fd;
struct stat st;
#ifndef WIN32
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
fd = open(filename, O_RDONLY);
#else
fd = _open(filename, _O_RDONLY);
#endif
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 (fd < 0)
{
LSQ_ERROR("cannot open %s for reading: %s", filename, strerror(errno));
return NULL;
}
if (0 != fstat(fd, &st))
{
LSQ_ERROR("cannot fstat(%s) failed: %s", filename, strerror(errno));
(void) close(fd);
return NULL;
}
struct reader_ctx *ctx = malloc(sizeof(*ctx));
ctx->file_size = st.st_size;
ctx->nread = 0;
ctx->fd = fd;
return ctx;
}
void
destroy_lsquic_reader_ctx (struct reader_ctx *ctx)
{
(void) close(ctx->fd);
free(ctx);
}