zephyr/subsys/net/ip/tcp2.c

2513 lines
58 KiB
C
Raw Normal View History

/*
* Copyright (c) 2018-2020 Intel Corporation
*
* SPDX-License-Identifier: Apache-2.0
*/
#include <logging/log.h>
LOG_MODULE_REGISTER(net_tcp, CONFIG_NET_TCP_LOG_LEVEL);
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <zephyr.h>
#include <random/rand32.h>
#include <net/net_pkt.h>
#include <net/net_context.h>
#include <net/udp.h>
#include "ipv4.h"
#include "ipv6.h"
#include "connection.h"
#include "net_stats.h"
#include "net_private.h"
#include "tcp2_priv.h"
#define FIN_TIMEOUT_MS MSEC_PER_SEC
#define FIN_TIMEOUT K_MSEC(FIN_TIMEOUT_MS)
static int tcp_rto = CONFIG_NET_TCP_INIT_RETRANSMISSION_TIMEOUT;
static int tcp_retries = CONFIG_NET_TCP_RETRY_COUNT;
static int tcp_window = NET_IPV6_MTU;
static sys_slist_t tcp_conns = SYS_SLIST_STATIC_INIT(&tcp_conns);
static K_MUTEX_DEFINE(tcp_lock);
static K_MEM_SLAB_DEFINE(tcp_conns_slab, sizeof(struct tcp),
CONFIG_NET_MAX_CONTEXTS, 4);
static void tcp_in(struct tcp *conn, struct net_pkt *pkt);
int (*tcp_send_cb)(struct net_pkt *pkt) = NULL;
size_t (*tcp_recv_cb)(struct tcp *conn, struct net_pkt *pkt) = NULL;
static uint32_t tcp_get_seq(struct net_buf *buf)
{
return *(uint32_t *)net_buf_user_data(buf);
}
static void tcp_set_seq(struct net_buf *buf, uint32_t seq)
{
*(uint32_t *)net_buf_user_data(buf) = seq;
}
static int tcp_pkt_linearize(struct net_pkt *pkt, size_t pos, size_t len)
{
struct net_buf *buf, *first = pkt->cursor.buf, *second = first->frags;
int ret = 0;
size_t len1, len2;
if (net_pkt_get_len(pkt) < (pos + len)) {
NET_ERR("Insufficient packet len=%zd (pos+len=%zu)",
net_pkt_get_len(pkt), pos + len);
ret = -EINVAL;
goto out;
}
buf = net_pkt_get_frag(pkt, TCP_PKT_ALLOC_TIMEOUT);
if (!buf || buf->size < len) {
if (buf) {
net_buf_unref(buf);
}
ret = -ENOBUFS;
goto out;
}
net_buf_linearize(buf->data, buf->size, pkt->frags, pos, len);
net_buf_add(buf, len);
len1 = first->len - (pkt->cursor.pos - pkt->cursor.buf->data);
len2 = len - len1;
first->len -= len1;
while (len2) {
size_t pull_len = MIN(second->len, len2);
struct net_buf *next;
len2 -= pull_len;
net_buf_pull(second, pull_len);
next = second->frags;
if (second->len == 0) {
net_buf_unref(second);
}
second = next;
}
buf->frags = second;
first->frags = buf;
out:
return ret;
}
static struct tcphdr *th_get(struct net_pkt *pkt)
{
size_t ip_len = net_pkt_ip_hdr_len(pkt) + net_pkt_ip_opts_len(pkt);
struct tcphdr *th = NULL;
again:
net_pkt_cursor_init(pkt);
net_pkt_set_overwrite(pkt, true);
if (net_pkt_skip(pkt, ip_len) != 0) {
goto out;
}
if (!net_pkt_is_contiguous(pkt, sizeof(*th))) {
if (tcp_pkt_linearize(pkt, ip_len, sizeof(*th)) < 0) {
goto out;
}
goto again;
}
th = net_pkt_cursor_get_pos(pkt);
out:
return th;
}
static size_t tcp_endpoint_len(sa_family_t af)
{
return (af == AF_INET) ? sizeof(struct sockaddr_in) :
sizeof(struct sockaddr_in6);
}
static int tcp_endpoint_set(union tcp_endpoint *ep, struct net_pkt *pkt,
enum pkt_addr src)
{
int ret = 0;
switch (net_pkt_family(pkt)) {
case AF_INET:
if (IS_ENABLED(CONFIG_NET_IPV4)) {
struct net_ipv4_hdr *ip = NET_IPV4_HDR(pkt);
struct tcphdr *th;
th = th_get(pkt);
if (!th) {
return -ENOBUFS;
}
memset(ep, 0, sizeof(*ep));
ep->sin.sin_port = src == TCP_EP_SRC ? th_sport(th) :
th_dport(th);
net_ipaddr_copy(&ep->sin.sin_addr,
src == TCP_EP_SRC ?
&ip->src : &ip->dst);
ep->sa.sa_family = AF_INET;
} else {
ret = -EINVAL;
}
break;
case AF_INET6:
if (IS_ENABLED(CONFIG_NET_IPV6)) {
struct net_ipv6_hdr *ip = NET_IPV6_HDR(pkt);
struct tcphdr *th;
th = th_get(pkt);
if (!th) {
return -ENOBUFS;
}
memset(ep, 0, sizeof(*ep));
ep->sin6.sin6_port = src == TCP_EP_SRC ? th_sport(th) :
th_dport(th);
net_ipaddr_copy(&ep->sin6.sin6_addr,
src == TCP_EP_SRC ?
&ip->src : &ip->dst);
ep->sa.sa_family = AF_INET6;
} else {
ret = -EINVAL;
}
break;
default:
NET_ERR("Unknown address family: %hu", net_pkt_family(pkt));
ret = -EINVAL;
}
return ret;
}
static const char *tcp_flags(uint8_t flags)
{
#define BUF_SIZE 25 /* 6 * 4 + 1 */
static char buf[BUF_SIZE];
int len = 0;
buf[0] = '\0';
if (flags) {
if (flags & SYN) {
len += snprintk(buf + len, BUF_SIZE - len, "SYN,");
}
if (flags & FIN) {
len += snprintk(buf + len, BUF_SIZE - len, "FIN,");
}
if (flags & ACK) {
len += snprintk(buf + len, BUF_SIZE - len, "ACK,");
}
if (flags & PSH) {
len += snprintk(buf + len, BUF_SIZE - len, "PSH,");
}
if (flags & RST) {
len += snprintk(buf + len, BUF_SIZE - len, "RST,");
}
if (flags & URG) {
len += snprintk(buf + len, BUF_SIZE - len, "URG,");
}
buf[len - 1] = '\0'; /* delete the last comma */
}
#undef BUF_SIZE
return buf;
}
static size_t tcp_data_len(struct net_pkt *pkt)
{
struct tcphdr *th = th_get(pkt);
size_t tcp_options_len = (th_off(th) - 5) * 4;
int len = net_pkt_get_len(pkt) - net_pkt_ip_hdr_len(pkt) -
net_pkt_ip_opts_len(pkt) - sizeof(*th) - tcp_options_len;
return len > 0 ? (size_t)len : 0;
}
static const char *tcp_th(struct net_pkt *pkt)
{
#define BUF_SIZE 80
static char buf[BUF_SIZE];
int len = 0;
struct tcphdr *th = th_get(pkt);
buf[0] = '\0';
if (th_off(th) < 5) {
len += snprintk(buf + len, BUF_SIZE - len,
"bogus th_off: %hu", (uint16_t)th_off(th));
goto end;
}
len += snprintk(buf + len, BUF_SIZE - len,
"%s Seq=%u", tcp_flags(th_flags(th)), th_seq(th));
if (th_flags(th) & ACK) {
len += snprintk(buf + len, BUF_SIZE - len,
" Ack=%u", th_ack(th));
}
len += snprintk(buf + len, BUF_SIZE - len,
" Len=%ld", (long)tcp_data_len(pkt));
end:
#undef BUF_SIZE
return buf;
}
#define is_6lo_technology(pkt) \
(IS_ENABLED(CONFIG_NET_IPV6) && net_pkt_family(pkt) == AF_INET6 && \
((IS_ENABLED(CONFIG_NET_L2_BT) && \
net_pkt_lladdr_dst(pkt)->type == NET_LINK_BLUETOOTH) || \
(IS_ENABLED(CONFIG_NET_L2_IEEE802154) && \
net_pkt_lladdr_dst(pkt)->type == NET_LINK_IEEE802154) || \
(IS_ENABLED(CONFIG_NET_L2_CANBUS) && \
net_pkt_lladdr_dst(pkt)->type == NET_LINK_CANBUS)))
static void tcp_send(struct net_pkt *pkt)
{
NET_DBG("%s", log_strdup(tcp_th(pkt)));
tcp_pkt_ref(pkt);
if (tcp_send_cb) {
if (tcp_send_cb(pkt) < 0) {
NET_ERR("net_send_data()");
tcp_pkt_unref(pkt);
}
goto out;
}
/* We must have special handling for some network technologies that
* tweak the IP protocol headers during packet sending. This happens
* with Bluetooth and IEEE 802.15.4 which use IPv6 header compression
* (6lo) and alter the sent network packet. So in order to avoid any
* corruption of the original data buffer, we must copy the sent data.
* For Bluetooth, its fragmentation code will even mangle the data
* part of the message so we need to copy those too.
*/
if (is_6lo_technology(pkt)) {
struct net_pkt *new_pkt;
new_pkt = net_pkt_clone(pkt, TCP_PKT_ALLOC_TIMEOUT);
if (!new_pkt) {
goto out;
}
if (net_send_data(new_pkt) < 0) {
net_pkt_unref(new_pkt);
}
/* We simulate sending of the original pkt and unref it like
* the device driver would do.
*/
tcp_pkt_unref(pkt);
} else {
if (net_send_data(pkt) < 0) {
NET_ERR("net_send_data()");
tcp_pkt_unref(pkt);
}
}
out:
tcp_pkt_unref(pkt);
}
static void tcp_send_queue_flush(struct tcp *conn)
{
struct net_pkt *pkt;
net: tcp2: fix sysworkq corruption in tcp_conn_unref() Bug description: When in tcp_conn_unref(), in case one of the delayed works is already submitted to sysworkq (after delay period), e.g. send_timer, the check of k_delayed_work_remaining_get() prevents calling k_delayed_work_cancel(). This leads to corrupting sysworkq when zeroing struct tcp* conn. Note that the "next" pointer for the work queue is part of the struct work (in _reserved field). Which is, in this case, a member of struct tcp. Scenario leading to the bug: (1) net_tcp_connect() is called from a work in sysworkq (2) net_tcp_connect() submits conn->send_timer to sysworkq (3) while net_tcp_connect() is waiting on connect_sem, delay period passes (z_timeout) and send_timer enters sysworkq work slist (4) also, some other code (app) submits more works to queue, now pointed by conn->send_timer in sysworkq work list (5) connection fails (no answer to SYN), causing a call to tcp_conn_unref() (6) tcp_conn_unref() is calling tcp_send_queue_flush() (7) checking k_delayed_work_remaining_get(&conn->send_timer) returns 0 due to delay period end, but send_timer is still in sysworkq work slist (sysworkq thread still hasn't handled the work) (8) BUG!: no call to k_delayed_work_cancel(&conn->send_timer) (9) back in tcp_conn_unref(), a call to memset(conn, 0, sizeof(*conn)) zeroes conn->send_timer (10) conn->send_timer is pointed to in sysworkq work slist, but is zeroed, clearing pointer to following works submitted in stage (4) (11) EFFECT! the works in stage (4) are never executed!! NOTES: * k_delayed_work_cancel(), handles both states: (1) delayed work pends on timeout and (2) work already in queue. So there is no need to check k_delayed_work_remaining_get() * This is also relevant for conn->send_data_timer Solution: removing checks of k_delayed_work_remaining_get(), always calling k_delayed_work_cancel() for work in struct tcp, in unref, before memset Signed-off-by: David Komel <a8961713@gmail.com>
2020-09-23 11:49:17 +03:00
k_delayed_work_cancel(&conn->send_timer);
while ((pkt = tcp_slist(&conn->send_queue, get,
struct net_pkt, next))) {
tcp_pkt_unref(pkt);
}
}
static int tcp_conn_unref(struct tcp *conn)
{
int ref_count = atomic_get(&conn->ref_count);
struct net_pkt *pkt;
NET_DBG("conn: %p, ref_count=%d", conn, ref_count);
#if !defined(CONFIG_NET_TEST_PROTOCOL)
if (conn->in_connect) {
NET_DBG("conn: %p is waiting on connect semaphore", conn);
tcp_send_queue_flush(conn);
goto out;
}
#endif /* CONFIG_NET_TEST_PROTOCOL */
ref_count = atomic_dec(&conn->ref_count) - 1;
if (ref_count) {
tp_out(net_context_get_family(conn->context), conn->iface,
"TP_TRACE", "event", "CONN_DELETE");
goto out;
}
k_mutex_lock(&tcp_lock, K_FOREVER);
/* If there is any pending data, pass that to application */
while ((pkt = k_fifo_get(&conn->recv_data, K_NO_WAIT)) != NULL) {
net_context_packet_received(
(struct net_conn *)conn->context->conn_handler,
pkt, NULL, NULL, conn->recv_user_data);
}
if (conn->context->conn_handler) {
net_conn_unregister(conn->context->conn_handler);
conn->context->conn_handler = NULL;
}
if (conn->context->recv_cb) {
conn->context->recv_cb(conn->context, NULL, NULL, NULL,
-ECONNRESET, conn->recv_user_data);
}
conn->context->tcp = NULL;
net_context_unref(conn->context);
tcp_send_queue_flush(conn);
net: tcp2: fix sysworkq corruption in tcp_conn_unref() Bug description: When in tcp_conn_unref(), in case one of the delayed works is already submitted to sysworkq (after delay period), e.g. send_timer, the check of k_delayed_work_remaining_get() prevents calling k_delayed_work_cancel(). This leads to corrupting sysworkq when zeroing struct tcp* conn. Note that the "next" pointer for the work queue is part of the struct work (in _reserved field). Which is, in this case, a member of struct tcp. Scenario leading to the bug: (1) net_tcp_connect() is called from a work in sysworkq (2) net_tcp_connect() submits conn->send_timer to sysworkq (3) while net_tcp_connect() is waiting on connect_sem, delay period passes (z_timeout) and send_timer enters sysworkq work slist (4) also, some other code (app) submits more works to queue, now pointed by conn->send_timer in sysworkq work list (5) connection fails (no answer to SYN), causing a call to tcp_conn_unref() (6) tcp_conn_unref() is calling tcp_send_queue_flush() (7) checking k_delayed_work_remaining_get(&conn->send_timer) returns 0 due to delay period end, but send_timer is still in sysworkq work slist (sysworkq thread still hasn't handled the work) (8) BUG!: no call to k_delayed_work_cancel(&conn->send_timer) (9) back in tcp_conn_unref(), a call to memset(conn, 0, sizeof(*conn)) zeroes conn->send_timer (10) conn->send_timer is pointed to in sysworkq work slist, but is zeroed, clearing pointer to following works submitted in stage (4) (11) EFFECT! the works in stage (4) are never executed!! NOTES: * k_delayed_work_cancel(), handles both states: (1) delayed work pends on timeout and (2) work already in queue. So there is no need to check k_delayed_work_remaining_get() * This is also relevant for conn->send_data_timer Solution: removing checks of k_delayed_work_remaining_get(), always calling k_delayed_work_cancel() for work in struct tcp, in unref, before memset Signed-off-by: David Komel <a8961713@gmail.com>
2020-09-23 11:49:17 +03:00
k_delayed_work_cancel(&conn->send_data_timer);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
tcp_pkt_unref(conn->send_data);
if (CONFIG_NET_TCP_RECV_QUEUE_TIMEOUT) {
tcp_pkt_unref(conn->queue_recv_data);
}
k_delayed_work_cancel(&conn->timewait_timer);
k_delayed_work_cancel(&conn->fin_timer);
sys_slist_find_and_remove(&tcp_conns, &conn->next);
memset(conn, 0, sizeof(*conn));
k_mem_slab_free(&tcp_conns_slab, (void **)&conn);
k_mutex_unlock(&tcp_lock);
out:
return ref_count;
}
int net_tcp_unref(struct net_context *context)
{
int ref_count = 0;
NET_DBG("context: %p, conn: %p", context, context->tcp);
if (context->tcp) {
ref_count = tcp_conn_unref(context->tcp);
}
return ref_count;
}
static bool tcp_send_process_no_lock(struct tcp *conn)
{
bool unref = false;
struct net_pkt *pkt;
pkt = tcp_slist(&conn->send_queue, peek_head,
struct net_pkt, next);
if (!pkt) {
goto out;
}
NET_DBG("%s %s", log_strdup(tcp_th(pkt)), conn->in_retransmission ?
"in_retransmission" : "");
if (conn->in_retransmission) {
if (conn->send_retries > 0) {
struct net_pkt *clone = tcp_pkt_clone(pkt);
if (clone) {
tcp_send(clone);
conn->send_retries--;
}
} else {
unref = true;
goto out;
}
} else {
uint8_t fl = th_get(pkt)->th_flags;
bool forget = ACK == fl || PSH == fl || (ACK | PSH) == fl ||
RST & fl;
pkt = forget ? tcp_slist(&conn->send_queue, get, struct net_pkt,
next) : tcp_pkt_clone(pkt);
if (!pkt) {
NET_ERR("net_pkt alloc failure");
goto out;
}
tcp_send(pkt);
if (forget == false && !k_delayed_work_remaining_get(
&conn->send_timer)) {
conn->send_retries = tcp_retries;
conn->in_retransmission = true;
}
}
if (conn->in_retransmission) {
k_delayed_work_submit(&conn->send_timer, K_MSEC(tcp_rto));
}
out:
return unref;
}
static void tcp_send_process(struct k_work *work)
{
struct tcp *conn = CONTAINER_OF(work, struct tcp, send_timer);
bool unref;
k_mutex_lock(&conn->lock, K_FOREVER);
unref = tcp_send_process_no_lock(conn);
k_mutex_unlock(&conn->lock);
if (unref) {
tcp_conn_unref(conn);
}
}
static void tcp_send_timer_cancel(struct tcp *conn)
{
if (conn->in_retransmission == false) {
return;
}
k_delayed_work_cancel(&conn->send_timer);
{
struct net_pkt *pkt = tcp_slist(&conn->send_queue, get,
struct net_pkt, next);
if (pkt) {
NET_DBG("%s", log_strdup(tcp_th(pkt)));
tcp_pkt_unref(pkt);
}
}
if (sys_slist_is_empty(&conn->send_queue)) {
conn->in_retransmission = false;
} else {
conn->send_retries = tcp_retries;
k_delayed_work_submit(&conn->send_timer, K_MSEC(tcp_rto));
}
}
static const char *tcp_state_to_str(enum tcp_state state, bool prefix)
{
const char *s = NULL;
#define _(_x) case _x: do { s = #_x; goto out; } while (0)
switch (state) {
_(TCP_LISTEN);
_(TCP_SYN_SENT);
_(TCP_SYN_RECEIVED);
_(TCP_ESTABLISHED);
_(TCP_FIN_WAIT_1);
_(TCP_FIN_WAIT_2);
_(TCP_CLOSE_WAIT);
_(TCP_CLOSING);
_(TCP_LAST_ACK);
_(TCP_TIME_WAIT);
_(TCP_CLOSED);
}
#undef _
NET_ASSERT(s, "Invalid TCP state: %u", state);
out:
return prefix ? s : (s + 4);
}
static const char *tcp_conn_state(struct tcp *conn, struct net_pkt *pkt)
{
#define BUF_SIZE 160
static char buf[BUF_SIZE];
snprintk(buf, BUF_SIZE, "%s [%s Seq=%u Ack=%u]", pkt ? tcp_th(pkt) : "",
tcp_state_to_str(conn->state, false),
conn->seq, conn->ack);
#undef BUF_SIZE
return buf;
}
static uint8_t *tcp_options_get(struct net_pkt *pkt, int tcp_options_len,
uint8_t *buf, size_t buf_len)
{
struct net_pkt_cursor backup;
int ret;
net_pkt_cursor_backup(pkt, &backup);
net_pkt_cursor_init(pkt);
net_pkt_skip(pkt, net_pkt_ip_hdr_len(pkt) + net_pkt_ip_opts_len(pkt) +
sizeof(struct tcphdr));
ret = net_pkt_read(pkt, buf, MIN(tcp_options_len, buf_len));
if (ret < 0) {
buf = NULL;
}
net_pkt_cursor_restore(pkt, &backup);
return buf;
}
static bool tcp_options_check(struct tcp_options *recv_options,
struct net_pkt *pkt, ssize_t len)
{
uint8_t options_buf[40]; /* TCP header max options size is 40 */
bool result = len > 0 && ((len % 4) == 0) ? true : false;
uint8_t *options = tcp_options_get(pkt, len, options_buf,
sizeof(options_buf));
uint8_t opt, opt_len;
NET_DBG("len=%zd", len);
recv_options->mss_found = false;
recv_options->wnd_found = false;
for ( ; options && len >= 1; options += opt_len, len -= opt_len) {
opt = options[0];
if (opt == TCPOPT_END) {
break;
} else if (opt == TCPOPT_NOP) {
opt_len = 1;
continue;
} else {
if (len < 2) { /* Only END and NOP can have length 1 */
NET_ERR("Illegal option %d with length %zd",
opt, len);
result = false;
break;
}
opt_len = options[1];
}
NET_DBG("opt: %hu, opt_len: %hu", (uint16_t)opt, (uint16_t)opt_len);
if (opt_len < 2 || opt_len > len) {
result = false;
break;
}
switch (opt) {
case TCPOPT_MAXSEG:
if (opt_len != 4) {
result = false;
goto end;
}
recv_options->mss =
ntohs(UNALIGNED_GET((uint16_t *)(options + 2)));
recv_options->mss_found = true;
NET_DBG("MSS=%hu", recv_options->mss);
break;
case TCPOPT_WINDOW:
if (opt_len != 3) {
result = false;
goto end;
}
recv_options->window = opt;
recv_options->wnd_found = true;
break;
default:
continue;
}
}
end:
if (false == result) {
NET_WARN("Invalid TCP options");
}
return result;
}
static size_t tcp_check_pending_data(struct tcp *conn, struct net_pkt *pkt,
size_t len)
{
size_t pending_len = 0;
if (CONFIG_NET_TCP_RECV_QUEUE_TIMEOUT &&
!net_pkt_is_empty(conn->queue_recv_data)) {
struct tcphdr *th = th_get(pkt);
uint32_t expected_seq = th_seq(th) + len;
uint32_t pending_seq;
pending_seq = tcp_get_seq(conn->queue_recv_data->buffer);
if (pending_seq == expected_seq) {
pending_len = net_pkt_get_len(conn->queue_recv_data);
NET_DBG("Found pending data seq %u len %zd",
pending_seq, pending_len);
net_buf_frag_add(pkt->buffer,
conn->queue_recv_data->buffer);
conn->queue_recv_data->buffer = NULL;
k_delayed_work_cancel(&conn->recv_queue_timer);
}
}
return pending_len;
}
static int tcp_data_get(struct tcp *conn, struct net_pkt *pkt, size_t *len)
{
int ret = 0;
if (tcp_recv_cb) {
tcp_recv_cb(conn, pkt);
goto out;
}
if (conn->context->recv_cb) {
struct net_pkt *up = net_pkt_clone(pkt, TCP_PKT_ALLOC_TIMEOUT);
if (!up) {
ret = -ENOBUFS;
goto out;
}
/* If there is any out-of-order pending data, then pass it
* to the application here.
*/
*len += tcp_check_pending_data(conn, up, *len);
net_pkt_cursor_init(up);
net_pkt_set_overwrite(up, true);
net_pkt_skip(up, net_pkt_get_len(up) - *len);
/* Do not pass data to application with TCP conn
* locked as there could be an issue when the app tries
* to send the data and the conn is locked. So the recv
* data is placed in fifo which is flushed in tcp_in()
* after unlocking the conn
*/
k_fifo_put(&conn->recv_data, up);
}
out:
return ret;
}
static int tcp_finalize_pkt(struct net_pkt *pkt)
{
net_pkt_cursor_init(pkt);
if (IS_ENABLED(CONFIG_NET_IPV4) && net_pkt_family(pkt) == AF_INET) {
return net_ipv4_finalize(pkt, IPPROTO_TCP);
}
if (IS_ENABLED(CONFIG_NET_IPV6) && net_pkt_family(pkt) == AF_INET6) {
return net_ipv6_finalize(pkt, IPPROTO_TCP);
}
return -EINVAL;
}
static int tcp_header_add(struct tcp *conn, struct net_pkt *pkt, uint8_t flags,
uint32_t seq)
{
NET_PKT_DATA_ACCESS_DEFINE(tcp_access, struct tcphdr);
struct tcphdr *th;
th = (struct tcphdr *)net_pkt_get_data(pkt, &tcp_access);
if (!th) {
return -ENOBUFS;
}
memset(th, 0, sizeof(struct tcphdr));
UNALIGNED_PUT(conn->src.sin.sin_port, &th->th_sport);
UNALIGNED_PUT(conn->dst.sin.sin_port, &th->th_dport);
th->th_off = 5;
UNALIGNED_PUT(flags, &th->th_flags);
UNALIGNED_PUT(htons(conn->recv_win), &th->th_win);
UNALIGNED_PUT(htonl(seq), &th->th_seq);
if (ACK & flags) {
UNALIGNED_PUT(htonl(conn->ack), &th->th_ack);
}
return net_pkt_set_data(pkt, &tcp_access);
}
static int ip_header_add(struct tcp *conn, struct net_pkt *pkt)
{
if (IS_ENABLED(CONFIG_NET_IPV4) && net_pkt_family(pkt) == AF_INET) {
return net_context_create_ipv4_new(conn->context, pkt,
&conn->src.sin.sin_addr,
&conn->dst.sin.sin_addr);
}
if (IS_ENABLED(CONFIG_NET_IPV6) && net_pkt_family(pkt) == AF_INET6) {
return net_context_create_ipv6_new(conn->context, pkt,
&conn->src.sin6.sin6_addr,
&conn->dst.sin6.sin6_addr);
}
return -EINVAL;
}
static int tcp_out_ext(struct tcp *conn, uint8_t flags, struct net_pkt *data,
uint32_t seq)
{
struct net_pkt *pkt;
int ret = 0;
pkt = tcp_pkt_alloc(conn, sizeof(struct tcphdr));
if (!pkt) {
ret = -ENOBUFS;
goto out;
}
if (data) {
/* Append the data buffer to the pkt */
net_pkt_append_buffer(pkt, data->buffer);
data->buffer = NULL;
}
ret = ip_header_add(conn, pkt);
if (ret < 0) {
tcp_pkt_unref(pkt);
goto out;
}
ret = tcp_header_add(conn, pkt, flags, seq);
if (ret < 0) {
tcp_pkt_unref(pkt);
goto out;
}
ret = tcp_finalize_pkt(pkt);
if (ret < 0) {
tcp_pkt_unref(pkt);
goto out;
}
NET_DBG("%s", log_strdup(tcp_th(pkt)));
if (tcp_send_cb) {
ret = tcp_send_cb(pkt);
goto out;
}
sys_slist_append(&conn->send_queue, &pkt->next);
if (tcp_send_process_no_lock(conn)) {
tcp_conn_unref(conn);
}
out:
return ret;
}
static void tcp_out(struct tcp *conn, uint8_t flags)
{
(void)tcp_out_ext(conn, flags, NULL /* no data */, conn->seq);
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
static int tcp_pkt_pull(struct net_pkt *pkt, size_t len)
{
int total = net_pkt_get_len(pkt);
int ret = 0;
if (len > total) {
ret = -EINVAL;
goto out;
}
net_pkt_cursor_init(pkt);
net_pkt_set_overwrite(pkt, true);
net_pkt_pull(pkt, len);
net_pkt_trim_buffer(pkt);
out:
return ret;
}
static int tcp_pkt_peek(struct net_pkt *to, struct net_pkt *from, size_t pos,
size_t len)
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
{
net_pkt_cursor_init(to);
net_pkt_cursor_init(from);
if (pos) {
net_pkt_set_overwrite(from, true);
net_pkt_skip(from, pos);
}
return net_pkt_copy(to, from, len);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
}
static bool tcp_window_full(struct tcp *conn)
{
bool window_full = !(conn->unacked_len < conn->send_win);
NET_DBG("conn: %p window_full=%hu", conn, window_full);
return window_full;
}
static int tcp_unsent_len(struct tcp *conn)
{
int unsent_len;
if (conn->unacked_len > conn->send_data_total) {
NET_ERR("total=%zu, unacked_len=%d",
conn->send_data_total, conn->unacked_len);
unsent_len = -ERANGE;
goto out;
}
unsent_len = conn->send_data_total - conn->unacked_len;
out:
NET_DBG("unsent_len=%d", unsent_len);
return unsent_len;
}
static int tcp_send_data(struct tcp *conn)
{
int ret = 0;
int pos, len;
struct net_pkt *pkt;
pos = conn->unacked_len;
len = MIN3(conn->send_data_total - conn->unacked_len,
conn->send_win - conn->unacked_len,
conn_mss(conn));
pkt = tcp_pkt_alloc(conn, len);
if (!pkt) {
NET_ERR("conn: %p packet allocation failed, len=%d", conn, len);
ret = -ENOBUFS;
goto out;
}
ret = tcp_pkt_peek(pkt, conn->send_data, pos, len);
if (ret < 0) {
tcp_pkt_unref(pkt);
ret = -ENOBUFS;
goto out;
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
ret = tcp_out_ext(conn, PSH | ACK, pkt, conn->seq + conn->unacked_len);
if (ret == 0) {
conn->unacked_len += len;
if (conn->data_mode == TCP_DATA_MODE_RESEND) {
net_stats_update_tcp_resent(net_pkt_iface(pkt), len);
net_stats_update_tcp_seg_rexmit(conn->iface);
} else {
net_stats_update_tcp_sent(net_pkt_iface(pkt), len);
net_stats_update_tcp_seg_sent(net_pkt_iface(pkt));
}
}
/* The data we want to send, has been moved to the send queue so we
* can unref the head net_pkt. If there was an error, we need to remove
* the packet anyway.
*/
tcp_pkt_unref(pkt);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
conn_send_data_dump(conn);
out:
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
return ret;
}
/* Send all queued but unsent data from the send_data packet by packet
* until the receiver's window is full. */
static int tcp_send_queued_data(struct tcp *conn)
{
int ret = 0;
bool subscribe = false;
if (conn->data_mode == TCP_DATA_MODE_RESEND) {
goto out;
}
while (tcp_unsent_len(conn) > 0) {
if (tcp_window_full(conn)) {
subscribe = true;
break;
}
ret = tcp_send_data(conn);
if (ret < 0) {
break;
}
}
if (conn->unacked_len) {
subscribe = true;
}
if (k_delayed_work_remaining_get(&conn->send_data_timer)) {
subscribe = false;
}
/* If we have out-of-bufs case, then do not start retransmit timer
* yet. The socket layer will catch this and resend data if needed.
*/
if (ret == -ENOBUFS) {
NET_DBG("No bufs, cancelling retransmit timer");
k_delayed_work_cancel(&conn->send_data_timer);
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
if (subscribe) {
conn->send_data_retries = 0;
k_delayed_work_submit(&conn->send_data_timer, K_MSEC(tcp_rto));
}
out:
return ret;
}
static void tcp_cleanup_recv_queue(struct k_work *work)
{
struct tcp *conn = CONTAINER_OF(work, struct tcp, recv_queue_timer);
k_mutex_lock(&conn->lock, K_FOREVER);
NET_DBG("Cleanup recv queue conn %p len %zd seq %u", conn,
net_pkt_get_len(conn->queue_recv_data),
tcp_get_seq(conn->queue_recv_data->buffer));
net_buf_unref(conn->queue_recv_data->buffer);
conn->queue_recv_data->buffer = NULL;
k_mutex_unlock(&conn->lock);
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
static void tcp_resend_data(struct k_work *work)
{
struct tcp *conn = CONTAINER_OF(work, struct tcp, send_data_timer);
bool conn_unref = false;
int ret;
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
k_mutex_lock(&conn->lock, K_FOREVER);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
NET_DBG("send_data_retries=%hu", conn->send_data_retries);
if (conn->send_data_retries >= tcp_retries) {
NET_DBG("conn: %p close, data retransmissions exceeded", conn);
conn_unref = true;
goto out;
}
conn->data_mode = TCP_DATA_MODE_RESEND;
conn->unacked_len = 0;
ret = tcp_send_data(conn);
if (ret == 0) {
conn->send_data_retries++;
if (conn->in_close && conn->send_data_total == 0) {
NET_DBG("TCP connection in active close, "
"not disposing yet (waiting %dms)",
FIN_TIMEOUT_MS);
k_delayed_work_submit(&conn->fin_timer, FIN_TIMEOUT);
conn_state(conn, TCP_FIN_WAIT_1);
ret = tcp_out_ext(conn, FIN | ACK, NULL,
conn->seq + conn->unacked_len);
if (ret == 0) {
conn_seq(conn, + 1);
}
goto out;
}
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
k_delayed_work_submit(&conn->send_data_timer, K_MSEC(tcp_rto));
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
out:
k_mutex_unlock(&conn->lock);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
if (conn_unref) {
tcp_conn_unref(conn);
}
}
static void tcp_timewait_timeout(struct k_work *work)
{
struct tcp *conn = CONTAINER_OF(work, struct tcp, timewait_timer);
NET_DBG("conn: %p %s", conn, log_strdup(tcp_conn_state(conn, NULL)));
/* Extra unref from net_tcp_put() */
net_context_unref(conn->context);
}
static void tcp_fin_timeout(struct k_work *work)
{
struct tcp *conn = CONTAINER_OF(work, struct tcp, fin_timer);
NET_DBG("Did not receive FIN in %dms", FIN_TIMEOUT_MS);
NET_DBG("conn: %p %s", conn, log_strdup(tcp_conn_state(conn, NULL)));
/* Extra unref from net_tcp_put() */
net_context_unref(conn->context);
}
static void tcp_conn_ref(struct tcp *conn)
{
int ref_count = atomic_inc(&conn->ref_count) + 1;
NET_DBG("conn: %p, ref_count: %d", conn, ref_count);
}
static struct tcp *tcp_conn_alloc(void)
{
struct tcp *conn = NULL;
int ret;
ret = k_mem_slab_alloc(&tcp_conns_slab, (void **)&conn, K_NO_WAIT);
if (ret) {
goto out;
}
memset(conn, 0, sizeof(*conn));
k_mutex_init(&conn->lock);
k_fifo_init(&conn->recv_data);
conn->state = TCP_LISTEN;
conn->recv_win = tcp_window;
conn->seq = (IS_ENABLED(CONFIG_NET_TEST_PROTOCOL) ||
IS_ENABLED(CONFIG_NET_TEST)) ? 0 : sys_rand32_get();
sys_slist_init(&conn->send_queue);
k_delayed_work_init(&conn->send_timer, tcp_send_process);
k_delayed_work_init(&conn->timewait_timer, tcp_timewait_timeout);
k_delayed_work_init(&conn->fin_timer, tcp_fin_timeout);
if (CONFIG_NET_TCP_RECV_QUEUE_TIMEOUT) {
conn->queue_recv_data = tcp_rx_pkt_alloc(conn, 0);
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
conn->send_data = tcp_pkt_alloc(conn, 0);
k_delayed_work_init(&conn->send_data_timer, tcp_resend_data);
k_delayed_work_init(&conn->recv_queue_timer, tcp_cleanup_recv_queue);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
k_sem_init(&conn->connect_sem, 0, UINT_MAX);
conn->in_connect = false;
tcp_conn_ref(conn);
sys_slist_append(&tcp_conns, &conn->next);
out:
NET_DBG("conn: %p", conn);
return conn;
}
int net_tcp_get(struct net_context *context)
{
int ret = 0;
struct tcp *conn;
k_mutex_lock(&tcp_lock, K_FOREVER);
conn = tcp_conn_alloc();
if (conn == NULL) {
ret = -ENOMEM;
goto out;
}
/* Mutually link the net_context and tcp connection */
conn->context = context;
context->tcp = conn;
out:
k_mutex_unlock(&tcp_lock);
return ret;
}
static bool tcp_endpoint_cmp(union tcp_endpoint *ep, struct net_pkt *pkt,
enum pkt_addr which)
{
union tcp_endpoint ep_tmp;
if (tcp_endpoint_set(&ep_tmp, pkt, which) < 0) {
return false;
}
return !memcmp(ep, &ep_tmp, tcp_endpoint_len(ep->sa.sa_family));
}
static bool tcp_conn_cmp(struct tcp *conn, struct net_pkt *pkt)
{
return tcp_endpoint_cmp(&conn->src, pkt, TCP_EP_DST) &&
tcp_endpoint_cmp(&conn->dst, pkt, TCP_EP_SRC);
}
static struct tcp *tcp_conn_search(struct net_pkt *pkt)
{
bool found = false;
struct tcp *conn;
struct tcp *tmp;
SYS_SLIST_FOR_EACH_CONTAINER_SAFE(&tcp_conns, conn, tmp, next) {
found = tcp_conn_cmp(conn, pkt);
if (found) {
break;
}
}
return found ? conn : NULL;
}
static struct tcp *tcp_conn_new(struct net_pkt *pkt);
static enum net_verdict tcp_recv(struct net_conn *net_conn,
struct net_pkt *pkt,
union net_ip_header *ip,
union net_proto_header *proto,
void *user_data)
{
struct tcp *conn;
struct tcphdr *th;
ARG_UNUSED(net_conn);
ARG_UNUSED(proto);
conn = tcp_conn_search(pkt);
if (conn) {
goto in;
}
th = th_get(pkt);
if (th_flags(th) & SYN && !(th_flags(th) & ACK)) {
struct tcp *conn_old = ((struct net_context *)user_data)->tcp;
conn = tcp_conn_new(pkt);
if (!conn) {
NET_ERR("Cannot allocate a new TCP connection");
goto in;
}
net_ipaddr_copy(&conn_old->context->remote, &conn->dst.sa);
conn->accepted_conn = conn_old;
}
in:
if (conn) {
tcp_in(conn, pkt);
}
return NET_DROP;
}
/* Create a new tcp connection, as a part of it, create and register
* net_context
*/
static struct tcp *tcp_conn_new(struct net_pkt *pkt)
{
struct tcp *conn = NULL;
struct net_context *context = NULL;
sa_family_t af = net_pkt_family(pkt);
struct sockaddr local_addr = { 0 };
int ret;
ret = net_context_get(af, SOCK_STREAM, IPPROTO_TCP, &context);
if (ret < 0) {
NET_ERR("net_context_get(): %d", ret);
goto err;
}
conn = context->tcp;
conn->iface = pkt->iface;
net_context_set_family(conn->context, net_pkt_family(pkt));
if (tcp_endpoint_set(&conn->dst, pkt, TCP_EP_SRC) < 0) {
net_context_unref(context);
conn = NULL;
goto err;
}
if (tcp_endpoint_set(&conn->src, pkt, TCP_EP_DST) < 0) {
net_context_unref(context);
conn = NULL;
goto err;
}
NET_DBG("conn: src: %s, dst: %s",
log_strdup(net_sprint_addr(conn->src.sa.sa_family,
(const void *)&conn->src.sin.sin_addr)),
log_strdup(net_sprint_addr(conn->dst.sa.sa_family,
(const void *)&conn->dst.sin.sin_addr)));
memcpy(&context->remote, &conn->dst, sizeof(context->remote));
context->flags |= NET_CONTEXT_REMOTE_ADDR_SET;
net_sin_ptr(&context->local)->sin_family = af;
local_addr.sa_family = net_context_get_family(context);
if (IS_ENABLED(CONFIG_NET_IPV6) &&
net_context_get_family(context) == AF_INET6) {
if (net_sin6_ptr(&context->local)->sin6_addr) {
net_ipaddr_copy(&net_sin6(&local_addr)->sin6_addr,
net_sin6_ptr(&context->local)->sin6_addr);
}
} else if (IS_ENABLED(CONFIG_NET_IPV4) &&
net_context_get_family(context) == AF_INET) {
if (net_sin_ptr(&context->local)->sin_addr) {
net_ipaddr_copy(&net_sin(&local_addr)->sin_addr,
net_sin_ptr(&context->local)->sin_addr);
}
}
ret = net_context_bind(context, &local_addr, sizeof(local_addr));
if (ret < 0) {
NET_DBG("Cannot bind accepted context, connection reset");
net_context_unref(context);
conn = NULL;
goto err;
}
NET_DBG("context: local: %s, remote: %s",
log_strdup(net_sprint_addr(
local_addr.sa_family,
(const void *)&net_sin(&local_addr)->sin_addr)),
log_strdup(net_sprint_addr(
context->remote.sa_family,
(const void *)&net_sin(&context->remote)->sin_addr)));
ret = net_conn_register(IPPROTO_TCP, af,
&context->remote, &local_addr,
ntohs(conn->dst.sin.sin_port),/* local port */
ntohs(conn->src.sin.sin_port),/* remote port */
tcp_recv, context,
&context->conn_handler);
if (ret < 0) {
NET_ERR("net_conn_register(): %d", ret);
net_context_unref(context);
conn = NULL;
goto err;
}
err:
if (!conn) {
net_stats_update_tcp_seg_conndrop(net_pkt_iface(pkt));
}
return conn;
}
static bool tcp_validate_seq(struct tcp *conn, struct tcphdr *hdr)
{
return (net_tcp_seq_cmp(th_seq(hdr), conn->ack) >= 0) &&
(net_tcp_seq_cmp(th_seq(hdr), conn->ack + conn->recv_win) < 0);
}
static void print_seq_list(struct net_buf *buf)
{
struct net_buf *tmp = buf;
uint32_t seq;
while (tmp) {
seq = tcp_get_seq(tmp);
NET_DBG("buf %p seq %u len %d", tmp, seq, tmp->len);
tmp = tmp->frags;
}
}
static void tcp_queue_recv_data(struct tcp *conn, struct net_pkt *pkt,
size_t len, uint32_t seq)
{
uint32_t seq_start = seq;
bool inserted = false;
struct net_buf *tmp;
NET_DBG("conn: %p len %zd seq %u ack %u", conn, len, seq, conn->ack);
tmp = pkt->buffer;
tcp_set_seq(tmp, seq);
seq += tmp->len;
tmp = tmp->frags;
while (tmp) {
tcp_set_seq(tmp, seq);
seq += tmp->len;
tmp = tmp->frags;
}
if (IS_ENABLED(CONFIG_NET_TCP_LOG_LEVEL_DBG)) {
NET_DBG("Queuing data: conn %p", conn);
print_seq_list(pkt->buffer);
}
if (!net_pkt_is_empty(conn->queue_recv_data)) {
/* Place the data to correct place in the list. If the data
* would not be sequential, then drop this packet.
*/
uint32_t pending_seq;
pending_seq = tcp_get_seq(conn->queue_recv_data->buffer);
if (pending_seq == seq) {
/* Put new data before the pending data */
net_buf_frag_add(pkt->buffer,
conn->queue_recv_data->buffer);
conn->queue_recv_data->buffer = pkt->buffer;
inserted = true;
} else {
struct net_buf *last;
last = net_buf_frag_last(conn->queue_recv_data->buffer);
pending_seq = tcp_get_seq(last);
if ((pending_seq + last->len) == seq_start) {
/* Put new data after pending data */
last->frags = pkt->buffer;
inserted = true;
}
}
if (IS_ENABLED(CONFIG_NET_TCP_LOG_LEVEL_DBG)) {
if (inserted) {
NET_DBG("All pending data: conn %p", conn);
print_seq_list(conn->queue_recv_data->buffer);
} else {
NET_DBG("Cannot add new data to queue");
}
}
} else {
net_pkt_append_buffer(conn->queue_recv_data, pkt->buffer);
inserted = true;
}
if (inserted) {
/* We need to keep the received data but free the pkt */
pkt->buffer = NULL;
if (!k_delayed_work_pending(&conn->recv_queue_timer)) {
k_delayed_work_submit(&conn->recv_queue_timer,
K_MSEC(CONFIG_NET_TCP_RECV_QUEUE_TIMEOUT));
}
}
}
static bool tcp_data_received(struct tcp *conn, struct net_pkt *pkt,
size_t *len)
{
if (tcp_data_get(conn, pkt, len) < 0) {
return false;
}
net_stats_update_tcp_seg_recv(conn->iface);
conn_ack(conn, *len);
tcp_out(conn, ACK);
return true;
}
static void tcp_out_of_order_data(struct tcp *conn, struct net_pkt *pkt,
size_t data_len, uint32_t seq)
{
size_t headers_len;
headers_len = net_pkt_get_len(pkt) - data_len;
/* Get rid of protocol headers from the data */
if (tcp_pkt_pull(pkt, headers_len) < 0) {
return;
}
/* We received out-of-order data. Try to queue it.
*/
tcp_queue_recv_data(conn, pkt, data_len, seq);
}
/* TCP state machine, everything happens here */
static void tcp_in(struct tcp *conn, struct net_pkt *pkt)
{
struct tcphdr *th = pkt ? th_get(pkt) : NULL;
uint8_t next = 0, fl = 0;
bool do_close = false;
size_t tcp_options_len = th ? (th_off(th) - 5) * 4 : 0;
struct net_conn *conn_handler = NULL;
struct net_pkt *recv_pkt;
void *recv_user_data;
struct k_fifo *recv_data_fifo;
size_t len;
int ret;
if (th) {
/* Currently we ignore ECN and CWR flags */
fl = th_flags(th) & ~(ECN | CWR);
}
k_mutex_lock(&conn->lock, K_FOREVER);
NET_DBG("%s", log_strdup(tcp_conn_state(conn, pkt)));
if (th && th_off(th) < 5) {
tcp_out(conn, RST);
conn_state(conn, TCP_CLOSED);
goto next_state;
}
if (FL(&fl, &, RST)) {
/* We only accept RST packet that has valid seq field. */
if (!tcp_validate_seq(conn, th)) {
net_stats_update_tcp_seg_rsterr(net_pkt_iface(pkt));
k_mutex_unlock(&conn->lock);
return;
}
net_stats_update_tcp_seg_rst(net_pkt_iface(pkt));
conn_state(conn, TCP_CLOSED);
goto next_state;
}
if (tcp_options_len && !tcp_options_check(&conn->recv_options, pkt,
tcp_options_len)) {
NET_DBG("DROP: Invalid TCP option list");
tcp_out(conn, RST);
conn_state(conn, TCP_CLOSED);
goto next_state;
}
if (th) {
size_t max_win;
conn->send_win = ntohs(th_win(th));
#if defined(CONFIG_NET_TCP_MAX_SEND_WINDOW_SIZE)
if (CONFIG_NET_TCP_MAX_SEND_WINDOW_SIZE) {
max_win = CONFIG_NET_TCP_MAX_SEND_WINDOW_SIZE;
} else
#endif
{
/* Adjust the window so that we do not run out of bufs
* while waiting acks.
*/
max_win = (CONFIG_NET_BUF_TX_COUNT *
CONFIG_NET_BUF_DATA_SIZE) / 3;
}
max_win = MAX(max_win, NET_IPV6_MTU);
if ((size_t)conn->send_win > max_win) {
NET_DBG("Lowering send window from %zd to %zd",
(size_t)conn->send_win, max_win);
conn->send_win = max_win;
}
}
next_state:
len = pkt ? tcp_data_len(pkt) : 0;
switch (conn->state) {
case TCP_LISTEN:
if (FL(&fl, ==, SYN)) {
conn_ack(conn, th_seq(th) + 1); /* capture peer's isn */
tcp_out(conn, SYN | ACK);
conn_seq(conn, + 1);
next = TCP_SYN_RECEIVED;
} else {
tcp_out(conn, SYN);
conn_seq(conn, + 1);
next = TCP_SYN_SENT;
}
break;
case TCP_SYN_RECEIVED:
if (FL(&fl, &, ACK, th_ack(th) == conn->seq &&
th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
next = TCP_ESTABLISHED;
net_context_set_state(conn->context,
NET_CONTEXT_CONNECTED);
if (conn->accepted_conn) {
conn->accepted_conn->accept_cb(
conn->context,
&conn->accepted_conn->context->remote,
sizeof(struct sockaddr), 0,
conn->accepted_conn->context);
/* Make sure the accept_cb is only called once.
*/
conn->accepted_conn = NULL;
}
if (len) {
if (tcp_data_get(conn, pkt, &len) < 0) {
break;
}
conn_ack(conn, + len);
tcp_out(conn, ACK);
}
}
break;
case TCP_SYN_SENT:
/* if we are in SYN SENT and receive only a SYN without an
* ACK , shouldn't we go to SYN RECEIVED state? See Figure
* 6 of RFC 793
*/
if (FL(&fl, &, SYN | ACK, th && th_ack(th) == conn->seq)) {
tcp_send_timer_cancel(conn);
conn_ack(conn, th_seq(th) + 1);
if (len) {
if (tcp_data_get(conn, pkt, &len) < 0) {
break;
}
conn_ack(conn, + len);
}
k_sem_give(&conn->connect_sem);
next = TCP_ESTABLISHED;
net_context_set_state(conn->context,
NET_CONTEXT_CONNECTED);
tcp_out(conn, ACK);
}
break;
case TCP_ESTABLISHED:
/* full-close */
if (th && FL(&fl, ==, (FIN | ACK), th_seq(th) == conn->ack)) {
if (net_tcp_seq_cmp(th_ack(th), conn->seq) > 0) {
uint32_t len_acked = th_ack(th) - conn->seq;
conn_seq(conn, + len_acked);
}
conn_ack(conn, + 1);
tcp_out(conn, FIN | ACK);
next = TCP_LAST_ACK;
break;
} else if (th && FL(&fl, ==, FIN, th_seq(th) == conn->ack)) {
conn_ack(conn, + 1);
tcp_out(conn, ACK);
next = TCP_CLOSE_WAIT;
break;
} else if (th && FL(&fl, ==, (FIN | ACK | PSH),
th_seq(th) == conn->ack)) {
if (len) {
if (tcp_data_get(conn, pkt, &len) < 0) {
break;
}
}
conn_ack(conn, + len + 1);
tcp_out(conn, FIN | ACK);
next = TCP_LAST_ACK;
break;
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
if (th && net_tcp_seq_cmp(th_ack(th), conn->seq) > 0) {
uint32_t len_acked = th_ack(th) - conn->seq;
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
NET_DBG("conn: %p len_acked=%u", conn, len_acked);
if ((conn->send_data_total < len_acked) ||
(tcp_pkt_pull(conn->send_data,
len_acked) < 0)) {
NET_ERR("conn: %p, Invalid len_acked=%u "
"(total=%zu)", conn, len_acked,
conn->send_data_total);
net_stats_update_tcp_seg_drop(conn->iface);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
tcp_out(conn, RST);
conn_state(conn, TCP_CLOSED);
break;
}
conn->send_data_total -= len_acked;
conn->unacked_len -= len_acked;
conn_seq(conn, + len_acked);
net_stats_update_tcp_seg_recv(conn->iface);
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
conn_send_data_dump(conn);
if (!k_delayed_work_remaining_get(&conn->send_data_timer)) {
NET_DBG("conn: %p, Missing a subscription "
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
"of the send_data queue timer", conn);
break;
}
conn->send_data_retries = 0;
k_delayed_work_cancel(&conn->send_data_timer);
if (conn->data_mode == TCP_DATA_MODE_RESEND) {
conn->unacked_len = 0;
}
conn->data_mode = TCP_DATA_MODE_SEND;
/* We are closing the connection, send a FIN to peer */
if (conn->in_close && conn->send_data_total == 0) {
tcp_send_timer_cancel(conn);
next = TCP_FIN_WAIT_1;
tcp_out(conn, FIN | ACK);
conn_seq(conn, + 1);
break;
}
ret = tcp_send_queued_data(conn);
if (ret < 0 && ret != -ENOBUFS) {
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
tcp_out(conn, RST);
conn_state(conn, TCP_CLOSED);
break;
}
}
if (th && len) {
if (th_seq(th) == conn->ack) {
if (!tcp_data_received(conn, pkt, &len)) {
break;
}
} else if (net_tcp_seq_greater(conn->ack, th_seq(th))) {
tcp_out(conn, ACK); /* peer has resent */
net_stats_update_tcp_seg_ackerr(conn->iface);
} else if (CONFIG_NET_TCP_RECV_QUEUE_TIMEOUT) {
tcp_out_of_order_data(conn, pkt, len,
th_seq(th));
}
}
break;
case TCP_CLOSE_WAIT:
tcp_out(conn, FIN);
next = TCP_LAST_ACK;
break;
case TCP_LAST_ACK:
if (th && FL(&fl, ==, ACK, th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
next = TCP_CLOSED;
}
break;
case TCP_CLOSED:
do_close = true;
break;
case TCP_FIN_WAIT_1:
if (th && FL(&fl, ==, (FIN | ACK), th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
conn_ack(conn, + 1);
tcp_out(conn, ACK);
next = TCP_TIME_WAIT;
} else if (th && FL(&fl, ==, FIN, th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
conn_ack(conn, + 1);
tcp_out(conn, ACK);
next = TCP_CLOSING;
} else if (th && FL(&fl, ==, ACK, th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
next = TCP_FIN_WAIT_2;
}
break;
case TCP_FIN_WAIT_2:
if (th && (FL(&fl, ==, FIN, th_seq(th) == conn->ack) ||
FL(&fl, ==, FIN | ACK, th_seq(th) == conn->ack))) {
/* Received FIN on FIN_WAIT_2, so cancel the timer */
k_delayed_work_cancel(&conn->fin_timer);
conn_ack(conn, + 1);
tcp_out(conn, ACK);
next = TCP_TIME_WAIT;
}
break;
case TCP_CLOSING:
if (th && FL(&fl, ==, ACK, th_seq(th) == conn->ack)) {
tcp_send_timer_cancel(conn);
next = TCP_TIME_WAIT;
}
break;
case TCP_TIME_WAIT:
k_delayed_work_submit(&conn->timewait_timer,
K_MSEC(CONFIG_NET_TCP_TIME_WAIT_DELAY));
break;
default:
NET_ASSERT(false, "%s is unimplemented",
tcp_state_to_str(conn->state, true));
}
if (next) {
pkt = NULL;
th = NULL;
conn_state(conn, next);
next = 0;
goto next_state;
}
/* If the conn->context is not set, then the connection was already
* closed.
*/
if (conn->context) {
conn_handler = (struct net_conn *)conn->context->conn_handler;
}
recv_user_data = conn->recv_user_data;
recv_data_fifo = &conn->recv_data;
k_mutex_unlock(&conn->lock);
/* Pass all the received data stored in recv fifo to the application.
* This is done like this so that we do not have any connection lock
* held.
*/
while (conn_handler && atomic_get(&conn->ref_count) > 0 &&
(recv_pkt = k_fifo_get(recv_data_fifo, K_NO_WAIT)) != NULL) {
net_context_packet_received(conn_handler, recv_pkt, NULL, NULL,
recv_user_data);
}
/* We must not try to unref the connection while having a connection
* lock because the unref will try to acquire net_context lock and the
* application might have that lock held already, and that might lead
* to a deadlock.
*/
if (do_close) {
tcp_conn_unref(conn);
}
}
/* Active connection close: send FIN and go to FIN_WAIT_1 state */
int net_tcp_put(struct net_context *context)
{
struct tcp *conn = context->tcp;
if (!conn) {
return -ENOENT;
}
k_mutex_lock(&conn->lock, K_FOREVER);
NET_DBG("%s", conn ? log_strdup(tcp_conn_state(conn, NULL)) : "");
NET_DBG("context %p %s", context,
log_strdup(({ const char *state = net_context_state(context);
state ? state : "<unknown>"; })));
if (conn && conn->state == TCP_ESTABLISHED) {
/* Send all remaining data if possible. */
if (conn->send_data_total > 0) {
NET_DBG("conn %p pending %zu bytes", conn,
conn->send_data_total);
conn->in_close = true;
/* How long to wait until all the data has been sent?
*/
k_delayed_work_submit(&conn->send_data_timer,
K_MSEC(tcp_rto));
} else {
int ret;
NET_DBG("TCP connection in active close, not "
"disposing yet (waiting %dms)", FIN_TIMEOUT_MS);
k_delayed_work_submit(&conn->fin_timer, FIN_TIMEOUT);
ret = tcp_out_ext(conn, FIN | ACK, NULL,
conn->seq + conn->unacked_len);
if (ret == 0) {
conn_seq(conn, + 1);
}
conn_state(conn, TCP_FIN_WAIT_1);
}
/* Make sure we do not delete the connection yet until we have
* sent the final ACK.
*/
net_context_ref(context);
}
k_mutex_unlock(&conn->lock);
net_context_unref(context);
return 0;
}
int net_tcp_listen(struct net_context *context)
{
/* when created, tcp connections are in state TCP_LISTEN */
net_context_set_state(context, NET_CONTEXT_LISTENING);
return 0;
}
int net_tcp_update_recv_wnd(struct net_context *context, int32_t delta)
{
ARG_UNUSED(context);
ARG_UNUSED(delta);
return -EPROTONOSUPPORT;
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
/* net_context queues the outgoing data for the TCP connection */
int net_tcp_queue_data(struct net_context *context, struct net_pkt *pkt)
{
struct tcp *conn = context->tcp;
struct net_buf *orig_buf = NULL;
int ret = 0;
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
size_t len;
if (!conn || conn->state != TCP_ESTABLISHED) {
return -ENOTCONN;
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
k_mutex_lock(&conn->lock, K_FOREVER);
if (tcp_window_full(conn)) {
/* Trigger resend if the timer is not active */
if (!k_delayed_work_remaining_get(&conn->send_data_timer)) {
NET_DBG("Window full, trigger resend");
tcp_resend_data(&conn->send_data_timer.work);
}
ret = -EAGAIN;
goto out;
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
len = net_pkt_get_len(pkt);
if (conn->send_data->buffer) {
orig_buf = net_buf_frag_last(conn->send_data->buffer);
}
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
net_pkt_append_buffer(conn->send_data, pkt->buffer);
conn->send_data_total += len;
NET_DBG("conn: %p Queued %zu bytes (total %zu)", conn, len,
conn->send_data_total);
pkt->buffer = NULL;
ret = tcp_send_queued_data(conn);
if (ret < 0 && ret != -ENOBUFS) {
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
tcp_conn_unref(conn);
goto out;
}
if (ret == -ENOBUFS) {
/* Restore the original data so that we do not resend the pkt
* data multiple times.
*/
conn->send_data_total -= len;
if (orig_buf) {
pkt->buffer = orig_buf->frags;
orig_buf->frags = NULL;
} else {
pkt->buffer = conn->send_data->buffer;
conn->send_data->buffer = NULL;
}
} else {
/* We should not free the pkt if there was an error. It will be
* freed in net_context.c:context_sendto()
*/
tcp_pkt_unref(pkt);
}
out:
net: tcp2: Implement the retransmission for the outgoing data In order to support the retransmission for the outgoing data: 1. The outgoing data packet is appended to the send_data queue in net_tcp_queue_data(). 2. tcp_send_queued_data() is called and will use tcp_send_data() to sends queued but unsent data packet by packet until there's an unsent data and the receiver's window isn't full. tcp_send_queued_data() subscribes send_data_timer that will handle retrasmissions with tcp_resend_data(). 3. tcp_send_data() peeks a single chunk of data from the send_data queue that will not exceed the maximum segment size until the the receiver's window is full. tcp_send_data() uses conn->seq and conn->unack_len as the sequence number for the TCP packet. conn->unacked_len is advanced on each send. 4. On data acknowledgment: - acknowledged amount of data is removed from the beginning of the send_data queue - conn->seq is advanced by the acknowledged amount - conn->unacked_len is decremented by the acknowledged amount - send_data_timer is cancelled - tcp_send_queued_data() is called to send queued but prevoiusly unsent data 5. On timeout, tcp_resend_data() will reset conn->unack_len, peek one packet from the beginning of the send_queue and resend, terminating the connection on retries exceeded. Meanwhile the outgoing data tcp_send_queued_data() is just appended to the send_data but not sent. In case of the acknowledgement, tcp_send_queued_data() will start sending multiple packets until the receiver's window is full. Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
2020-05-15 18:17:37 +03:00
k_mutex_unlock(&conn->lock);
return ret;
}
/* net context is about to send out queued data - inform caller only */
int net_tcp_send_data(struct net_context *context, net_context_send_cb_t cb,
void *user_data)
{
if (cb) {
cb(context, 0, user_data);
}
return 0;
}
/* When connect() is called on a TCP socket, register the socket for incoming
* traffic with net context and give the TCP packet receiving function, which
* in turn will call tcp_in() to deliver the TCP packet to the stack
*/
int net_tcp_connect(struct net_context *context,
const struct sockaddr *remote_addr,
struct sockaddr *local_addr,
uint16_t remote_port, uint16_t local_port,
k_timeout_t timeout, net_context_connect_cb_t cb,
void *user_data)
{
struct tcp *conn;
int ret = 0;
NET_DBG("context: %p, local: %s, remote: %s", context,
log_strdup(net_sprint_addr(
local_addr->sa_family,
(const void *)&net_sin(local_addr)->sin_addr)),
log_strdup(net_sprint_addr(
remote_addr->sa_family,
(const void *)&net_sin(remote_addr)->sin_addr)));
conn = context->tcp;
conn->iface = net_context_get_iface(context);
switch (net_context_get_family(context)) {
const struct in_addr *ip4;
const struct in6_addr *ip6;
case AF_INET:
memset(&conn->src, 0, sizeof(struct sockaddr_in));
memset(&conn->dst, 0, sizeof(struct sockaddr_in));
conn->src.sa.sa_family = AF_INET;
conn->dst.sa.sa_family = AF_INET;
conn->dst.sin.sin_port = remote_port;
conn->src.sin.sin_port = local_port;
/* we have to select the source address here as
* net_context_create_ipv4_new() is not called in the packet
* output chain
*/
ip4 = net_if_ipv4_select_src_addr(
net_context_get_iface(context),
&net_sin(remote_addr)->sin_addr);
conn->src.sin.sin_addr = *ip4;
net_ipaddr_copy(&conn->dst.sin.sin_addr,
&net_sin(remote_addr)->sin_addr);
break;
case AF_INET6:
memset(&conn->src, 0, sizeof(struct sockaddr_in6));
memset(&conn->dst, 0, sizeof(struct sockaddr_in6));
conn->src.sin6.sin6_family = AF_INET6;
conn->dst.sin6.sin6_family = AF_INET6;
conn->dst.sin6.sin6_port = remote_port;
conn->src.sin6.sin6_port = local_port;
ip6 = net_if_ipv6_select_src_addr(
net_context_get_iface(context),
&net_sin6(remote_addr)->sin6_addr);
conn->src.sin6.sin6_addr = *ip6;
net_ipaddr_copy(&conn->dst.sin6.sin6_addr,
&net_sin6(remote_addr)->sin6_addr);
break;
default:
ret = -EPROTONOSUPPORT;
}
NET_DBG("conn: %p src: %s, dst: %s", conn,
log_strdup(net_sprint_addr(conn->src.sa.sa_family,
(const void *)&conn->src.sin.sin_addr)),
log_strdup(net_sprint_addr(conn->dst.sa.sa_family,
(const void *)&conn->dst.sin.sin_addr)));
net_context_set_state(context, NET_CONTEXT_CONNECTING);
ret = net_conn_register(net_context_get_ip_proto(context),
net_context_get_family(context),
remote_addr, local_addr,
ntohs(remote_port), ntohs(local_port),
tcp_recv, context,
&context->conn_handler);
if (ret < 0) {
goto out;
}
/* Input of a (nonexistent) packet with no flags set will cause
* a TCP connection to be established
*/
tcp_in(conn, NULL);
if (!IS_ENABLED(CONFIG_NET_TEST_PROTOCOL)) {
conn->in_connect = true;
if (k_sem_take(&conn->connect_sem, timeout) != 0 &&
conn->state != TCP_ESTABLISHED) {
conn->in_connect = false;
tcp_conn_unref(conn);
ret = -ETIMEDOUT;
goto out;
}
conn->in_connect = false;
}
out:
NET_DBG("conn: %p, ret=%d", conn, ret);
return ret;
}
int net_tcp_accept(struct net_context *context, net_tcp_accept_cb_t cb,
void *user_data)
{
struct tcp *conn = context->tcp;
struct sockaddr local_addr = { };
uint16_t local_port, remote_port;
if (!conn) {
return -EINVAL;
}
NET_DBG("context: %p, tcp: %p, cb: %p", context, conn, cb);
if (conn->state != TCP_LISTEN) {
return -EINVAL;
}
conn->accept_cb = cb;
local_addr.sa_family = net_context_get_family(context);
switch (local_addr.sa_family) {
struct sockaddr_in *in;
struct sockaddr_in6 *in6;
case AF_INET:
in = (struct sockaddr_in *)&local_addr;
if (net_sin_ptr(&context->local)->sin_addr) {
net_ipaddr_copy(&in->sin_addr,
net_sin_ptr(&context->local)->sin_addr);
}
in->sin_port =
net_sin((struct sockaddr *)&context->local)->sin_port;
local_port = ntohs(in->sin_port);
remote_port = ntohs(net_sin(&context->remote)->sin_port);
break;
case AF_INET6:
in6 = (struct sockaddr_in6 *)&local_addr;
if (net_sin6_ptr(&context->local)->sin6_addr) {
net_ipaddr_copy(&in6->sin6_addr,
net_sin6_ptr(&context->local)->sin6_addr);
}
in6->sin6_port =
net_sin6((struct sockaddr *)&context->local)->sin6_port;
local_port = ntohs(in6->sin6_port);
remote_port = ntohs(net_sin6(&context->remote)->sin6_port);
break;
default:
return -EINVAL;
}
context->user_data = user_data;
/* Remove the temporary connection handler and register
* a proper now as we have an established connection.
*/
net_conn_unregister(context->conn_handler);
return net_conn_register(net_context_get_ip_proto(context),
local_addr.sa_family,
context->flags & NET_CONTEXT_REMOTE_ADDR_SET ?
&context->remote : NULL,
&local_addr,
remote_port, local_port,
tcp_recv, context,
&context->conn_handler);
}
int net_tcp_recv(struct net_context *context, net_context_recv_cb_t cb,
void *user_data)
{
struct tcp *conn = context->tcp;
NET_DBG("context: %p, cb: %p, user_data: %p", context, cb, user_data);
context->recv_cb = cb;
if (conn) {
conn->recv_user_data = user_data;
}
return 0;
}
int net_tcp_finalize(struct net_pkt *pkt)
{
NET_PKT_DATA_ACCESS_DEFINE(tcp_access, struct net_tcp_hdr);
struct net_tcp_hdr *tcp_hdr;
tcp_hdr = (struct net_tcp_hdr *)net_pkt_get_data(pkt, &tcp_access);
if (!tcp_hdr) {
return -ENOBUFS;
}
tcp_hdr->chksum = 0U;
if (net_if_need_calc_tx_checksum(net_pkt_iface(pkt))) {
tcp_hdr->chksum = net_calc_chksum_tcp(pkt);
}
return net_pkt_set_data(pkt, &tcp_access);
}
struct net_tcp_hdr *net_tcp_input(struct net_pkt *pkt,
struct net_pkt_data_access *tcp_access)
{
struct net_tcp_hdr *tcp_hdr;
if (IS_ENABLED(CONFIG_NET_TCP_CHECKSUM) &&
net_if_need_calc_rx_checksum(net_pkt_iface(pkt)) &&
net_calc_chksum_tcp(pkt) != 0U) {
NET_DBG("DROP: checksum mismatch");
goto drop;
}
tcp_hdr = (struct net_tcp_hdr *)net_pkt_get_data(pkt, tcp_access);
if (tcp_hdr && !net_pkt_set_data(pkt, tcp_access)) {
return tcp_hdr;
}
drop:
net_stats_update_tcp_seg_chkerr(net_pkt_iface(pkt));
return NULL;
}
#if defined(CONFIG_NET_TEST_PROTOCOL)
static enum net_verdict tcp_input(struct net_conn *net_conn,
struct net_pkt *pkt,
union net_ip_header *ip,
union net_proto_header *proto,
void *user_data)
{
struct tcphdr *th = th_get(pkt);
if (th) {
struct tcp *conn = tcp_conn_search(pkt);
if (conn == NULL && SYN == th_flags(th)) {
struct net_context *context =
tcp_calloc(1, sizeof(struct net_context));
net_tcp_get(context);
net_context_set_family(context, net_pkt_family(pkt));
conn = context->tcp;
tcp_endpoint_set(&conn->dst, pkt, TCP_EP_SRC);
tcp_endpoint_set(&conn->src, pkt, TCP_EP_DST);
/* Make an extra reference, the sanity check suite
* will delete the connection explicitly
*/
tcp_conn_ref(conn);
}
if (conn) {
conn->iface = pkt->iface;
tcp_in(conn, pkt);
}
}
return NET_DROP;
}
static size_t tp_tcp_recv_cb(struct tcp *conn, struct net_pkt *pkt)
{
ssize_t len = tcp_data_len(pkt);
struct net_pkt *up = tcp_pkt_clone(pkt);
NET_DBG("pkt: %p, len: %zu", pkt, net_pkt_get_len(pkt));
net_pkt_cursor_init(up);
net_pkt_set_overwrite(up, true);
net_pkt_pull(up, net_pkt_get_len(up) - len);
net_tcp_queue_data(conn->context, up);
return len;
}
static ssize_t tp_tcp_recv(int fd, void *buf, size_t len, int flags)
{
return 0;
}
static void tp_init(struct tcp *conn, struct tp *tp)
{
struct tp out = {
.msg = "",
.status = "",
.state = tcp_state_to_str(conn->state, true),
.seq = conn->seq,
.ack = conn->ack,
.rcv = "",
.data = "",
.op = "",
};
*tp = out;
}
static void tcp_to_json(struct tcp *conn, void *data, size_t *data_len)
{
struct tp tp;
tp_init(conn, &tp);
tp_encode(&tp, data, data_len);
}
enum net_verdict tp_input(struct net_conn *net_conn,
struct net_pkt *pkt,
union net_ip_header *ip_hdr,
union net_proto_header *proto,
void *user_data)
{
struct net_udp_hdr *uh = net_udp_get_hdr(pkt, NULL);
size_t data_len = ntohs(uh->len) - sizeof(*uh);
struct tcp *conn = tcp_conn_search(pkt);
size_t json_len = 0;
struct tp *tp;
struct tp_new *tp_new;
enum tp_type type;
bool responded = false;
static char buf[512];
net_pkt_cursor_init(pkt);
net_pkt_set_overwrite(pkt, true);
net_pkt_skip(pkt, net_pkt_ip_hdr_len(pkt) +
net_pkt_ip_opts_len(pkt) + sizeof(*uh));
net_pkt_read(pkt, buf, data_len);
buf[data_len] = '\0';
data_len += 1;
type = json_decode_msg(buf, data_len);
data_len = ntohs(uh->len) - sizeof(*uh);
net_pkt_cursor_init(pkt);
net_pkt_set_overwrite(pkt, true);
net_pkt_skip(pkt, net_pkt_ip_hdr_len(pkt) +
net_pkt_ip_opts_len(pkt) + sizeof(*uh));
net_pkt_read(pkt, buf, data_len);
buf[data_len] = '\0';
data_len += 1;
switch (type) {
case TP_CONFIG_REQUEST:
tp_new = json_to_tp_new(buf, data_len);
break;
default:
tp = json_to_tp(buf, data_len);
break;
}
switch (type) {
case TP_COMMAND:
if (is("CONNECT", tp->op)) {
tp_output(pkt->family, pkt->iface, buf, 1);
responded = true;
{
struct net_context *context = tcp_calloc(1,
sizeof(struct net_context));
net_tcp_get(context);
net_context_set_family(context,
net_pkt_family(pkt));
conn = context->tcp;
tcp_endpoint_set(&conn->dst, pkt, TCP_EP_SRC);
tcp_endpoint_set(&conn->src, pkt, TCP_EP_DST);
conn->iface = pkt->iface;
tcp_conn_ref(conn);
}
conn->seq = tp->seq;
tcp_in(conn, NULL);
}
if (is("CLOSE", tp->op)) {
tp_trace = false;
{
struct net_context *context;
conn = (void *)sys_slist_peek_head(&tcp_conns);
context = conn->context;
while (tcp_conn_unref(conn))
;
tcp_free(context);
}
tp_mem_stat();
tp_nbuf_stat();
tp_pkt_stat();
tp_seq_stat();
}
if (is("CLOSE2", tp->op)) {
struct tcp *conn =
(void *)sys_slist_peek_head(&tcp_conns);
net_tcp_put(conn->context);
}
if (is("RECV", tp->op)) {
#define HEXSTR_SIZE 64
char hexstr[HEXSTR_SIZE];
ssize_t len = tp_tcp_recv(0, buf, sizeof(buf), 0);
tp_init(conn, tp);
bin2hex(buf, len, hexstr, HEXSTR_SIZE);
tp->data = hexstr;
NET_DBG("%zd = tcp_recv(\"%s\")", len, tp->data);
json_len = sizeof(buf);
tp_encode(tp, buf, &json_len);
}
if (is("SEND", tp->op)) {
ssize_t len = tp_str_to_hex(buf, sizeof(buf), tp->data);
struct tcp *conn =
(void *)sys_slist_peek_head(&tcp_conns);
tp_output(pkt->family, pkt->iface, buf, 1);
responded = true;
NET_DBG("tcp_send(\"%s\")", tp->data);
{
struct net_pkt *data_pkt;
data_pkt = tcp_pkt_alloc(conn, len);
net_pkt_write(data_pkt, buf, len);
net_pkt_cursor_init(data_pkt);
net_tcp_queue_data(conn->context, data_pkt);
}
}
break;
case TP_CONFIG_REQUEST:
tp_new_find_and_apply(tp_new, "tcp_rto", &tcp_rto, TP_INT);
tp_new_find_and_apply(tp_new, "tcp_retries", &tcp_retries,
TP_INT);
tp_new_find_and_apply(tp_new, "tcp_window", &tcp_window,
TP_INT);
tp_new_find_and_apply(tp_new, "tp_trace", &tp_trace, TP_BOOL);
break;
case TP_INTROSPECT_REQUEST:
json_len = sizeof(buf);
conn = (void *)sys_slist_peek_head(&tcp_conns);
tcp_to_json(conn, buf, &json_len);
break;
case TP_DEBUG_STOP: case TP_DEBUG_CONTINUE:
tp_state = tp->type;
break;
default:
NET_ASSERT(false, "Unimplemented tp command: %s", tp->msg);
}
if (json_len) {
tp_output(pkt->family, pkt->iface, buf, json_len);
} else if ((TP_CONFIG_REQUEST == type || TP_COMMAND == type)
&& responded == false) {
tp_output(pkt->family, pkt->iface, buf, 1);
}
return NET_DROP;
}
static void test_cb_register(sa_family_t family, uint8_t proto, uint16_t remote_port,
uint16_t local_port, net_conn_cb_t cb)
{
struct net_conn_handle *conn_handle = NULL;
const struct sockaddr addr = { .sa_family = family, };
int ret = net_conn_register(proto,
family,
&addr, /* remote address */
&addr, /* local address */
local_port,
remote_port,
cb,
NULL, /* user_data */
&conn_handle);
if (ret < 0) {
NET_ERR("net_conn_register(): %d", ret);
}
}
#endif /* CONFIG_NET_TEST_PROTOCOL */
void net_tcp_foreach(net_tcp_cb_t cb, void *user_data)
{
struct tcp *conn;
struct tcp *tmp;
k_mutex_lock(&tcp_lock, K_FOREVER);
SYS_SLIST_FOR_EACH_CONTAINER_SAFE(&tcp_conns, conn, tmp, next) {
if (atomic_get(&conn->ref_count) > 0) {
k_mutex_unlock(&tcp_lock);
cb(conn, user_data);
k_mutex_lock(&tcp_lock, K_FOREVER);
}
}
k_mutex_unlock(&tcp_lock);
}
uint16_t net_tcp_get_recv_mss(const struct tcp *conn)
{
sa_family_t family = net_context_get_family(conn->context);
if (family == AF_INET) {
#if defined(CONFIG_NET_IPV4)
struct net_if *iface = net_context_get_iface(conn->context);
if (iface && net_if_get_mtu(iface) >= NET_IPV4TCPH_LEN) {
/* Detect MSS based on interface MTU minus "TCP,IP
* header size"
*/
return net_if_get_mtu(iface) - NET_IPV4TCPH_LEN;
}
#else
return 0;
#endif /* CONFIG_NET_IPV4 */
}
#if defined(CONFIG_NET_IPV6)
else if (family == AF_INET6) {
struct net_if *iface = net_context_get_iface(conn->context);
int mss = 0;
if (iface && net_if_get_mtu(iface) >= NET_IPV6TCPH_LEN) {
/* Detect MSS based on interface MTU minus "TCP,IP
* header size"
*/
mss = net_if_get_mtu(iface) - NET_IPV6TCPH_LEN;
}
if (mss < NET_IPV6_MTU) {
mss = NET_IPV6_MTU;
}
return mss;
}
#endif /* CONFIG_NET_IPV6 */
return 0;
}
const char *net_tcp_state_str(enum tcp_state state)
{
return tcp_state_to_str(state, false);
}
void net_tcp_init(void)
{
#if defined(CONFIG_NET_TEST_PROTOCOL)
/* Register inputs for TTCN-3 based TCP2 sanity check */
test_cb_register(AF_INET, IPPROTO_TCP, 4242, 4242, tcp_input);
test_cb_register(AF_INET6, IPPROTO_TCP, 4242, 4242, tcp_input);
test_cb_register(AF_INET, IPPROTO_UDP, 4242, 4242, tp_input);
test_cb_register(AF_INET6, IPPROTO_UDP, 4242, 4242, tp_input);
tcp_recv_cb = tp_tcp_recv_cb;
#endif
}