sergiotarxz/nuttx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
99cde13a11
nuttx/net/tcp/tcp_timer.c
Alexander Lunev 0f080cdeaf net/tcp/sendfile: NET_TCP_WRITE_BUFFERS and NET_SENDFILE were inconsistent with each other: 
tcp_sendfile() reads data directly from a file and does not use NET_TCP_WRITE_BUFFERS data flow
even if CONFIG_NET_TCP_WRITE_BUFFERS option is enabled.
Despite this, tcp_sendfile relied on NET_TCP_WRITE_BUFFERS specific flow control variables that
were idle during sendfile operation. Thus it was a total inconsistency.

E.g. because of the issue, TCP socket used by sendfile() operation never issued
FIN packet on close() command, and the TCP connection hung up.

As a result of the fix, simultaneously enabled CONFIG_NET_TCP_WRITE_BUFFERS and
CONFIG_NET_SENDFILE options can coexist.
2022-01-17 01:42:41 +08:00

533 lines
17 KiB
C
Raw Blame History

/****************************************************************************
* net/tcp/tcp_timer.c
* Poll for the availability of TCP TX data
*
* Copyright (C) 2007-2010, 2015-2016, 2018, 2020 Gregory Nutt. All rights
* reserved.
* Author: Gregory Nutt <gnutt@nuttx.org>
*
* Adapted for NuttX from logic in uIP which also has a BSD-like license:
*
* Original author Adam Dunkels <adam@dunkels.com>
* Copyright () 2001-2003, Adam Dunkels.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote
* products derived from this software without specific prior
* written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS
* OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include <nuttx/config.h>
#if defined(CONFIG_NET) && defined(CONFIG_NET_TCP)
#include <stdint.h>
#include <assert.h>
#include <debug.h>
#include <nuttx/net/netconfig.h>
#include <nuttx/net/net.h>
#include <nuttx/net/netdev.h>
#include <nuttx/net/netstats.h>
#include <nuttx/net/tcp.h>
#include "devif/devif.h"
#include "socket/socket.h"
#include "tcp/tcp.h"
/****************************************************************************
* Pre-processor Definitions
****************************************************************************/
/* Per RFC 1122: "... an ACK should not be excessively delayed; in
* particular, the delay MUST be less than 0.5 seconds ..."
*
* NOTE: We only have 0.5 timing resolution here so the delay will be
* between 0.5 and 1.0 seconds, and may be delayed further, depending on the
* polling rate of the the driver (often 1 second).
*/
#define ACK_DELAY (1)
/****************************************************************************
* Public Functions
****************************************************************************/
/****************************************************************************
* Name: tcp_timer
*
* Description:
* Handle a TCP timer expiration for the provided TCP connection
*
* Input Parameters:
* dev - The device driver structure to use in the send operation
* conn - The TCP "connection" to poll for TX data
* hsec - The polling interval in units of halves of a second
*
* Returned Value:
* None
*
* Assumptions:
* The network is locked.
* dev is not NULL.
* conn is not NULL.
* The connection (conn) is bound to the polling device (dev).
*
****************************************************************************/
void tcp_timer(FAR struct net_driver_s *dev, FAR struct tcp_conn_s *conn,
int hsec)
{
uint16_t result;
uint8_t hdrlen;
/* NOTE: It is important to decrease conn->timer at "hsec" pace,
* not faster. Excessive (false) decrements of conn->timer are not allowed
* here. Otherwise, it breaks TCP timings and leads to TCP spurious
* retransmissions and other issues due to premature timeouts.
*/
DEBUGASSERT(dev != NULL && conn != NULL && dev == conn->dev);
/* Set up for the callback. We can't know in advance if the application
* is going to send a IPv4 or an IPv6 packet, so this setup may not
* actually be used. Furthermore, the TCP logic is required to call
* tcp_ipv4_select() or tcp_ipv6_select() prior to sending any packets.
* We will try to set the correct value here basic on the binding of
* the connection.
*/
#ifdef CONFIG_NET_IPv4
#ifdef CONFIG_NET_IPv6
if (conn->domain == PF_INET)
#endif
{
hdrlen = IPv4TCP_HDRLEN;
tcp_ipv4_select(dev);
}
#endif /* CONFIG_NET_IPv4 */
#ifdef CONFIG_NET_IPv6
#ifdef CONFIG_NET_IPv4
else
#endif
{
hdrlen = IPv6TCP_HDRLEN;
tcp_ipv6_select(dev);
}
#endif /* CONFIG_NET_IPv6 */
/* Increase the TCP sequence number */
tcp_nextsequence();
/* Reset the length variables. */
dev->d_len = 0;
dev->d_sndlen = 0;
if (conn->tcpstateflags == TCP_CLOSED)
{
/* Nothing to be done */
return;
}
/* Check if the connection is in a state in which we simply wait
* for the connection to time out. If so, we increase the
* connection's timer and remove the connection if it times
* out.
*/
if (conn->tcpstateflags == TCP_TIME_WAIT ||
conn->tcpstateflags == TCP_FIN_WAIT_2)
{
unsigned int newtimer;
/* Increment the connection timer */
newtimer = (unsigned int)conn->timer + hsec;
/* Check if the timer exceeds the timeout value */
if (newtimer >= (TCP_TIME_WAIT_TIMEOUT * HSEC_PER_SEC))
{
/* Set the timer to the maximum value */
conn->timer = TCP_TIME_WAIT_TIMEOUT * HSEC_PER_SEC;
conn->tcpstateflags = TCP_CLOSED;
/* Notify upper layers about the timeout */
tcp_callback(dev, conn, TCP_TIMEDOUT);
ninfo("TCP state: TCP_CLOSED\n");
}
else
{
/* No timeout. Just update the incremented timer */
conn->timer = newtimer;
}
}
else if (conn->tcpstateflags != TCP_CLOSED)
{
/* If the connection has outstanding data, we increase the connection's
* timer and see if it has reached the RTO value in which case we
* retransmit.
*/
if (conn->tx_unacked > 0)
{
/* The connection has outstanding data */
if (conn->timer > hsec)
{
/* Will not yet decrement to zero */
conn->timer -= hsec;
}
else
{
/* Will decrement to zero */
conn->timer = 0;
/* Check for a timeout on connection in the TCP_SYN_RCVD state.
* On such timeouts, we would normally resend the SYNACK until
* the ACK is received, completing the 3-way handshake. But if
* the retry count elapsed, then we must assume that no ACK is
* forthcoming and terminate the attempted connection.
*/
if (conn->tcpstateflags == TCP_SYN_RCVD &&
conn->nrtx >= TCP_MAXSYNRTX)
{
FAR struct tcp_conn_s *listener;
conn->tcpstateflags = TCP_CLOSED;
ninfo("TCP state: TCP_SYN_RCVD->TCP_CLOSED\n");
/* Find the listener for this connection. */
#if defined(CONFIG_NET_IPv4) && defined(CONFIG_NET_IPv6)
listener = tcp_findlistener(&conn->u, conn->lport,
conn->domain);
#else
listener = tcp_findlistener(&conn->u, conn->lport);
#endif
if (listener != NULL)
{
/* We call tcp_callback() for the connection with
* TCP_TIMEDOUT to inform the listener that the
* connection has timed out.
*/
tcp_callback(dev, listener, TCP_TIMEDOUT);
}
/* We also send a reset packet to the remote host. */
tcp_send(dev, conn, TCP_RST | TCP_ACK, hdrlen);
/* Finally, we must free this TCP connection structure */
conn->crefs = 0;
tcp_free(conn);
goto done;
}
/* Otherwise, check for a timeout on an established connection.
* If the retry count is exceeded in this case, we should
* close the connection.
*/
else if (
#ifdef CONFIG_NET_TCP_WRITE_BUFFERS
# ifdef CONFIG_NET_SENDFILE
(!conn->sendfile && conn->expired > 0) ||
(conn->sendfile && conn->nrtx >= TCP_MAXRTX) ||
# else
conn->expired > 0 ||
# endif
#else
conn->nrtx >= TCP_MAXRTX ||
#endif
(conn->tcpstateflags == TCP_SYN_SENT &&
conn->nrtx >= TCP_MAXSYNRTX)
)
{
conn->tcpstateflags = TCP_CLOSED;
ninfo("TCP state: TCP_CLOSED\n");
/* We call tcp_callback() with TCP_TIMEDOUT to
* inform the application that the connection has
* timed out.
*/
tcp_callback(dev, conn, TCP_TIMEDOUT);
/* We also send a reset packet to the remote host. */
tcp_send(dev, conn, TCP_RST | TCP_ACK, hdrlen);
goto done;
}
/* Exponential backoff. */
conn->timer = TCP_RTO << (conn->nrtx > 4 ? 4: conn->nrtx);
(conn->nrtx)++;
/* Ok, so we need to retransmit. We do this differently
* depending on which state we are in. In ESTABLISHED, we
* call upon the application so that it may prepare the
* data for the retransmit. In SYN_RCVD, we resend the
* SYNACK that we sent earlier and in LAST_ACK we have to
* retransmit our FINACK.
*/
#ifdef CONFIG_NET_STATISTICS
g_netstats.tcp.rexmit++;
#endif
switch (conn->tcpstateflags & TCP_STATE_MASK)
{
case TCP_SYN_RCVD:
/* In the SYN_RCVD state, we should retransmit our
* SYNACK.
*/
#if !defined(CONFIG_NET_TCP_WRITE_BUFFERS)
tcp_setsequence(conn->sndseq, conn->rexmit_seq);
#else
/* REVISIT for the buffered mode */
#endif
tcp_synack(dev, conn, TCP_ACK | TCP_SYN);
goto done;
case TCP_SYN_SENT:
/* In the SYN_SENT state, we retransmit out SYN. */
#if !defined(CONFIG_NET_TCP_WRITE_BUFFERS)
tcp_setsequence(conn->sndseq, conn->rexmit_seq);
#else
/* REVISIT for the buffered mode */
#endif
tcp_synack(dev, conn, TCP_SYN);
goto done;
case TCP_ESTABLISHED:
/* In the ESTABLISHED state, we call upon the application
* to do the actual retransmit after which we jump into
* the code for sending out the packet.
*/
result = tcp_callback(dev, conn, TCP_REXMIT);
tcp_rexmit(dev, conn, result);
goto done;
case TCP_FIN_WAIT_1:
case TCP_CLOSING:
case TCP_LAST_ACK:
/* In all these states we should retransmit a FINACK. */
#if !defined(CONFIG_NET_TCP_WRITE_BUFFERS)
tcp_setsequence(conn->sndseq, conn->rexmit_seq);
#else
/* REVISIT for the buffered mode */
#endif
tcp_send(dev, conn, TCP_FIN | TCP_ACK, hdrlen);
goto done;
}
}
}
/* The connection does not have outstanding data. Check if the TCP
* connection has been established.
*/
else if ((conn->tcpstateflags & TCP_STATE_MASK) == TCP_ESTABLISHED)
{
#ifdef CONFIG_NET_TCP_KEEPALIVE
/* Is this an established connected with KeepAlive enabled? */
if (conn->keepalive)
{
socktimeo_t timeo;
uint32_t saveseq;
/* If this is the first probe, then the keepstart time is
* the time that the last ACK or data was received from the
* remote.
*
* On subsequent retries, keepstart is the time that the
* last probe was sent.
*/
if (conn->keepretries > 0)
{
timeo = (socktimeo_t)conn->keepintvl;
}
else
{
timeo = (socktimeo_t)conn->keepidle;
}
/* Yes... has the idle period elapsed with no data or ACK
* received from the remote peer?
*/
if (net_timeo(conn->keeptime, timeo))
{
/* Yes.. Has the retry count expired? */
if (conn->keepretries >= conn->keepcnt)
{
/* Yes... stop the network monitor, closing the
* connection and all sockets associated with the
* connection.
*/
tcp_stop_monitor(conn, TCP_ABORT);
}
else
{
unsigned int tcpiplen;
/* No.. we need to send another probe.
* Get the size of the IP and TCP header.
*/
#ifdef CONFIG_NET_IPv4
#ifdef CONFIG_NET_IPv6
if (conn->domain == PF_INET)
#endif
{
tcpiplen = IPv4_HDRLEN + TCP_HDRLEN;
}
#endif
#ifdef CONFIG_NET_IPv6
#ifdef CONFIG_NET_IPv4
else
#endif
{
tcpiplen = IPv6_HDRLEN + TCP_HDRLEN;
}
#endif
/* And send the probe.
* The packet we send must have these properties:
*
* - TCP_ACK flag (only) is set.
* - Sequence number is the sequence number of
* previously ACKed data, i.e., the expected
* sequence number minus one.
*
* tcp_send() will send the TCP sequence number as
* conn->sndseq. Rather than creating a new
* interface, we spoof tcp_end() here:
*/
saveseq = tcp_getsequence(conn->sndseq);
tcp_setsequence(conn->sndseq, saveseq - 1);
tcp_send(dev, conn, TCP_ACK, tcpiplen);
tcp_setsequence(conn->sndseq, saveseq);
#ifdef CONFIG_NET_TCP_WRITE_BUFFERS
/* Increment the un-ACKed sequence number */
conn->sndseq_max++;
#endif
/* Update for the next probe */
conn->keeptime = clock_systime_ticks();
conn->keepretries++;
}
goto done;
}
}
#endif
#ifdef CONFIG_NET_TCP_DELAYED_ACK
/* Handle delayed acknowledgments. Is there a segment with a
* delayed acknowledgment?
*/
if (conn->rx_unackseg > 0)
{
/* Increment the ACK delay. */
conn->rx_acktimer += hsec;
/* Per RFC 1122: "...an ACK should not be excessively
* delayed; in particular, the delay must be less than
* 0.5 seconds..."
*/
if (conn->rx_acktimer >= ACK_DELAY)
{
/* Reset the delayed ACK state and send the ACK
* packet.
*/
conn->rx_unackseg = 0;
conn->rx_acktimer = 0;
tcp_synack(dev, conn, TCP_ACK);
goto done;
}
}
#endif
/* There was no need for a retransmission and there was no
* need to probe the remote peer and there was no need to
* send a delayed ACK. We poll the application for new
* outgoing data.
*/
result = tcp_callback(dev, conn, TCP_POLL);
tcp_appsend(dev, conn, result);
goto done;
}
}
/* Nothing to be done */
dev->d_len = 0;
done:
return;
}
#endif /* CONFIG_NET && CONFIG_NET_TCP */
Reference in New Issue View Git Blame Copy Permalink