freebsd-skq/sys/netinet/tcp_var.h

583 lines
22 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*
* Copyright (c) 1982, 1986, 1993, 1994, 1995
1994-05-24 10:09:53 +00:00
* The Regents of the University of California. 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.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``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 REGENTS OR CONTRIBUTORS 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.
*
* @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
1994-08-21 05:27:42 +00:00
#ifndef _NETINET_TCP_VAR_H_
#define _NETINET_TCP_VAR_H_
#include <netinet/tcp.h>
1994-05-24 10:09:53 +00:00
/*
* Kernel variables for tcp.
*/
extern int tcp_do_rfc1323;
extern int tcp_do_rfc1644;
1994-05-24 10:09:53 +00:00
/* TCP segment queue entry */
struct tseg_qent {
LIST_ENTRY(tseg_qent) tqe_q;
int tqe_len; /* TCP segment data length */
struct tcphdr *tqe_th; /* a pointer to tcp header */
struct mbuf *tqe_m; /* mbuf contains packet */
};
LIST_HEAD(tsegqe_head, tseg_qent);
extern int tcp_reass_qsize;
extern struct uma_zone *tcp_reass_zone;
struct tcptemp {
u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
struct tcphdr tt_t;
};
#define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
1994-05-24 10:09:53 +00:00
/*
* Tcp control block, one per tcp; fields:
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
* Organized for 16 byte cacheline efficiency.
1994-05-24 10:09:53 +00:00
*/
struct tcpcb {
struct tsegqe_head t_segq; /* segment reassembly queue */
int t_segqlen; /* segment reassembly queue length */
int t_dupacks; /* consecutive dup acks recd */
struct tcptemp *unused; /* unused */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
struct callout *tt_rexmt; /* retransmit timer */
struct callout *tt_persist; /* retransmit persistence */
struct callout *tt_keep; /* keepalive */
struct callout *tt_2msl; /* 2*msl TIME_WAIT timer */
struct callout *tt_delack; /* delayed ACK timer */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
struct inpcb *t_inpcb; /* back pointer to internet pcb */
int t_state; /* state of this connection */
u_int t_flags;
#define TF_ACKNOW 0x000001 /* ack peer immediately */
#define TF_DELACK 0x000002 /* ack, but try to delay it */
#define TF_NODELAY 0x000004 /* don't delay packets to coalesce */
#define TF_NOOPT 0x000008 /* don't use tcp options */
#define TF_SENTFIN 0x000010 /* have sent FIN */
#define TF_REQ_SCALE 0x000020 /* have/will request window scaling */
#define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */
#define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */
#define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */
#define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */
#define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */
#define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */
#define TF_NOPUSH 0x001000 /* don't push */
#define TF_REQ_CC 0x002000 /* have/will request CC */
#define TF_RCVD_CC 0x004000 /* a CC was received in SYN */
#define TF_SENDCCNEW 0x008000 /* send CCnew instead of CC in SYN */
#define TF_MORETOCOME 0x010000 /* More data to be appended to sock */
#define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */
#define TF_LASTIDLE 0x040000 /* connection was previously idle */
#define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */
#define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */
#define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
int t_force; /* 1 if forcing out a byte */
1994-05-24 10:09:53 +00:00
tcp_seq snd_una; /* send unacknowledged */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
tcp_seq snd_max; /* highest sequence number sent;
* used to recognize retransmits
*/
1994-05-24 10:09:53 +00:00
tcp_seq snd_nxt; /* send next */
tcp_seq snd_up; /* send urgent pointer */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
1994-05-24 10:09:53 +00:00
tcp_seq snd_wl1; /* window update seg seq number */
tcp_seq snd_wl2; /* window update seg ack number */
tcp_seq iss; /* initial send sequence number */
tcp_seq irs; /* initial receive sequence number */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
tcp_seq rcv_nxt; /* receive next */
1994-05-24 10:09:53 +00:00
tcp_seq rcv_adv; /* advertised window */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
u_long rcv_wnd; /* receive window */
tcp_seq rcv_up; /* receive urgent pointer */
u_long snd_wnd; /* send window */
1994-05-24 10:09:53 +00:00
u_long snd_cwnd; /* congestion-controlled window */
u_long snd_bwnd; /* bandwidth-controlled window */
u_long snd_ssthresh; /* snd_cwnd size threshold for
1994-05-24 10:09:53 +00:00
* for slow start exponential to
* linear switch
*/
u_long snd_bandwidth; /* calculated bandwidth or 0 */
tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
u_int t_maxopd; /* mss plus options */
u_long t_rcvtime; /* inactivity time */
u_long t_starttime; /* time connection was established */
int t_rtttime; /* round trip time */
1994-05-24 10:09:53 +00:00
tcp_seq t_rtseq; /* sequence number being timed */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
int t_bw_rtttime; /* used for bandwidth calculation */
tcp_seq t_bw_rtseq; /* used for bandwidth calculation */
int t_rxtcur; /* current retransmit value (ticks) */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
u_int t_maxseg; /* maximum segment size */
int t_srtt; /* smoothed round-trip time */
int t_rttvar; /* variance in round-trip time */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
int t_rxtshift; /* log(2) of rexmt exp. backoff */
u_int t_rttmin; /* minimum rtt allowed */
u_int t_rttbest; /* best rtt we've seen */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
u_long t_rttupdated; /* number of times rtt sampled */
1994-05-24 10:09:53 +00:00
u_long max_sndwnd; /* largest window peer has offered */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
int t_softerror; /* possible error not yet reported */
1994-05-24 10:09:53 +00:00
/* out-of-band data */
char t_oobflags; /* have some */
char t_iobc; /* input character */
#define TCPOOB_HAVEDATA 0x01
#define TCPOOB_HADDATA 0x02
/* RFC 1323 variables */
u_char snd_scale; /* window scaling for send window */
u_char rcv_scale; /* window scaling for recv window */
u_char request_r_scale; /* pending window scaling */
u_char requested_s_scale;
u_long ts_recent; /* timestamp echo data */
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
1994-05-24 10:09:53 +00:00
u_long ts_recent_age; /* when last updated */
tcp_seq last_ack_sent;
/* RFC 1644 variables */
tcp_cc cc_send; /* send connection count */
tcp_cc cc_recv; /* receive connection count */
/* experimental */
u_long snd_cwnd_prev; /* cwnd prior to retransmit */
u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */
tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
u_long t_badrxtwin; /* window for retransmit recovery */
u_char snd_limited; /* segments limited transmitted */
Limiters and sanity checks for TCP MSS (maximum segement size) resource exhaustion attacks. For network link optimization TCP can adjust its MSS and thus packet size according to the observed path MTU. This is done dynamically based on feedback from the remote host and network components along the packet path. This information can be abused to pretend an extremely low path MTU. The resource exhaustion works in two ways: o during tcp connection setup the advertized local MSS is exchanged between the endpoints. The remote endpoint can set this arbitrarily low (except for a minimum MTU of 64 octets enforced in the BSD code). When the local host is sending data it is forced to send many small IP packets instead of a large one. For example instead of the normal TCP payload size of 1448 it forces TCP payload size of 12 (MTU 64) and thus we have a 120 times increase in workload and packets. On fast links this quickly saturates the local CPU and may also hit pps processing limites of network components along the path. This type of attack is particularly effective for servers where the attacker can download large files (WWW and FTP). We mitigate it by enforcing a minimum MTU settable by sysctl net.inet.tcp.minmss defaulting to 256 octets. o the local host is reveiving data on a TCP connection from the remote host. The local host has no control over the packet size the remote host is sending. The remote host may chose to do what is described in the first attack and send the data in packets with an TCP payload of at least one byte. For each packet the tcp_input() function will be entered, the packet is processed and a sowakeup() is signalled to the connected process. For example an attack with 2 Mbit/s gives 4716 packets per second and the same amount of sowakeup()s to the process (and context switches). This type of attack is particularly effective for servers where the attacker can upload large amounts of data. Normally this is the case with WWW server where large POSTs can be made. We mitigate this by calculating the average MSS payload per second. If it goes below 'net.inet.tcp.minmss' and the pps rate is above 'net.inet.tcp.minmssoverload' defaulting to 1000 this particular TCP connection is resetted and dropped. MITRE CVE: CAN-2004-0002 Reviewed by: sam (mentor) MFC after: 1 day
2004-01-08 17:40:07 +00:00
/* anti DoS counters */
u_long rcv_second; /* start of interval second */
u_long rcv_pps; /* received packets per second */
u_long rcv_byps; /* received bytes per second */
1994-05-24 10:09:53 +00:00
};
#define IN_FASTRECOVERY(tp) (tp->t_flags & TF_FASTRECOVERY)
#define ENTER_FASTRECOVERY(tp) tp->t_flags |= TF_FASTRECOVERY
#define EXIT_FASTRECOVERY(tp) tp->t_flags &= ~TF_FASTRECOVERY
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#ifdef TCP_SIGNATURE
/*
* Defines which are needed by the xform_tcp module and tcp_[in|out]put
* for SADB verification and lookup.
*/
#define TCP_SIGLEN 16 /* length of computed digest in bytes */
#define TCP_KEYLEN_MIN 1 /* minimum length of TCP-MD5 key */
#define TCP_KEYLEN_MAX 80 /* maximum length of TCP-MD5 key */
/*
* Only a single SA per host may be specified at this time. An SPI is
* needed in order for the KEY_ALLOCSA() lookup to work.
*/
#define TCP_SIG_SPI 0x1000
#endif /* TCP_SIGNATURE */
/*
* Structure to hold TCP options that are only used during segment
* processing (in tcp_input), but not held in the tcpcb.
* It's basically used to reduce the number of parameters
* to tcp_dooptions.
*/
struct tcpopt {
u_long to_flags; /* which options are present */
#define TOF_TS 0x0001 /* timestamp */
#define TOF_CC 0x0002 /* CC and CCnew are exclusive */
#define TOF_CCNEW 0x0004
#define TOF_CCECHO 0x0008
#define TOF_MSS 0x0010
#define TOF_SCALE 0x0020
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#define TOF_SIGNATURE 0x0040 /* signature option present */
2004-04-20 19:04:24 +00:00
#define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */
u_int32_t to_tsval;
u_int32_t to_tsecr;
tcp_cc to_cc; /* holds CC or CCnew */
tcp_cc to_ccecho;
u_int16_t to_mss;
u_int8_t to_requested_s_scale;
u_int8_t to_pad;
};
#ifdef _NETINET_IN_PCB_H_
struct syncache {
inp_gen_t sc_inp_gencnt; /* pointer check */
struct tcpcb *sc_tp; /* tcb for listening socket */
struct mbuf *sc_ipopts; /* source route */
struct in_conninfo sc_inc; /* addresses */
u_int32_t sc_tsrecent;
tcp_cc sc_cc_send; /* holds CC or CCnew */
tcp_cc sc_cc_recv;
tcp_seq sc_irs; /* seq from peer */
tcp_seq sc_iss; /* our ISS */
u_long sc_rxttime; /* retransmit time */
u_int16_t sc_rxtslot; /* retransmit counter */
u_int16_t sc_peer_mss; /* peer's MSS */
u_int16_t sc_wnd; /* advertised window */
u_int8_t sc_requested_s_scale:4,
sc_request_r_scale:4;
u_int8_t sc_flags;
#define SCF_NOOPT 0x01 /* no TCP options */
#define SCF_WINSCALE 0x02 /* negotiated window scaling */
#define SCF_TIMESTAMP 0x04 /* negotiated timestamps */
#define SCF_CC 0x08 /* negotiated CC */
#define SCF_UNREACH 0x10 /* icmp unreachable received */
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#define SCF_SIGNATURE 0x20 /* send MD5 digests */
TAILQ_ENTRY(syncache) sc_hash;
TAILQ_ENTRY(syncache) sc_timerq;
};
struct syncache_head {
TAILQ_HEAD(, syncache) sch_bucket;
u_int sch_length;
};
#else
struct in_conninfo;
#endif /* _NETINET_IN_PCB_H_ */
struct hc_metrics_lite { /* must stay in sync with hc_metrics */
u_long rmx_mtu; /* MTU for this path */
u_long rmx_ssthresh; /* outbound gateway buffer limit */
u_long rmx_rtt; /* estimated round trip time */
u_long rmx_rttvar; /* estimated rtt variance */
u_long rmx_bandwidth; /* estimated bandwidth */
u_long rmx_cwnd; /* congestion window */
u_long rmx_sendpipe; /* outbound delay-bandwidth product */
u_long rmx_recvpipe; /* inbound delay-bandwidth product */
};
struct tcptw {
struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */
tcp_seq snd_nxt;
tcp_seq rcv_nxt;
tcp_seq iss;
tcp_seq irs;
tcp_cc cc_recv;
tcp_cc cc_send;
u_short last_win; /* cached window value */
u_short tw_so_options; /* copy of so_options */
struct ucred *tw_cred; /* user credentials */
u_long t_recent;
u_long t_starttime;
int tw_time;
LIST_ENTRY(tcptw) tw_2msl;
};
/*
* The TAO cache entry which is stored in the tcp hostcache.
*/
struct rmxp_tao {
tcp_cc tao_cc; /* latest CC in valid SYN */
tcp_cc tao_ccsent; /* latest CC sent to peer */
u_short tao_mssopt; /* peer's cached MSS */
#ifdef notyet
u_short tao_flags; /* cache status flags */
#define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
#define TAOF_OK 0x0002 /* peer does understand rfc1644 */
#define TAOF_UNDEF 0 /* we don't know yet */
#endif /* notyet */
};
1994-05-24 10:09:53 +00:00
#define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
#define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb)
1994-05-24 10:09:53 +00:00
#define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
/*
* The smoothed round-trip time and estimated variance
* are stored as fixed point numbers scaled by the values below.
* For convenience, these scales are also used in smoothing the average
* (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
* With these scales, srtt has 3 bits to the right of the binary point,
* and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
* binary point, and is smoothed with an ALPHA of 0.75.
*/
#define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
#define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
#define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
#define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
#define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
1994-05-24 10:09:53 +00:00
/*
* The initial retransmission should happen at rtt + 4 * rttvar.
* Because of the way we do the smoothing, srtt and rttvar
* will each average +1/2 tick of bias. When we compute
* the retransmit timer, we want 1/2 tick of rounding and
* 1 extra tick because of +-1/2 tick uncertainty in the
* firing of the timer. The bias will give us exactly the
* 1.5 tick we need. But, because the bias is
* statistical, we have to test that we don't drop below
* the minimum feasible timer (which is 2 ticks).
* This version of the macro adapted from a paper by Lawrence
* Brakmo and Larry Peterson which outlines a problem caused
* by insufficient precision in the original implementation,
* which results in inappropriately large RTO values for very
* fast networks.
1994-05-24 10:09:53 +00:00
*/
#define TCP_REXMTVAL(tp) \
max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
+ (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
1994-05-24 10:09:53 +00:00
/*
* TCP statistics.
* Many of these should be kept per connection,
* but that's inconvenient at the moment.
*/
struct tcpstat {
u_long tcps_connattempt; /* connections initiated */
u_long tcps_accepts; /* connections accepted */
u_long tcps_connects; /* connections established */
u_long tcps_drops; /* connections dropped */
u_long tcps_conndrops; /* embryonic connections dropped */
Limiters and sanity checks for TCP MSS (maximum segement size) resource exhaustion attacks. For network link optimization TCP can adjust its MSS and thus packet size according to the observed path MTU. This is done dynamically based on feedback from the remote host and network components along the packet path. This information can be abused to pretend an extremely low path MTU. The resource exhaustion works in two ways: o during tcp connection setup the advertized local MSS is exchanged between the endpoints. The remote endpoint can set this arbitrarily low (except for a minimum MTU of 64 octets enforced in the BSD code). When the local host is sending data it is forced to send many small IP packets instead of a large one. For example instead of the normal TCP payload size of 1448 it forces TCP payload size of 12 (MTU 64) and thus we have a 120 times increase in workload and packets. On fast links this quickly saturates the local CPU and may also hit pps processing limites of network components along the path. This type of attack is particularly effective for servers where the attacker can download large files (WWW and FTP). We mitigate it by enforcing a minimum MTU settable by sysctl net.inet.tcp.minmss defaulting to 256 octets. o the local host is reveiving data on a TCP connection from the remote host. The local host has no control over the packet size the remote host is sending. The remote host may chose to do what is described in the first attack and send the data in packets with an TCP payload of at least one byte. For each packet the tcp_input() function will be entered, the packet is processed and a sowakeup() is signalled to the connected process. For example an attack with 2 Mbit/s gives 4716 packets per second and the same amount of sowakeup()s to the process (and context switches). This type of attack is particularly effective for servers where the attacker can upload large amounts of data. Normally this is the case with WWW server where large POSTs can be made. We mitigate this by calculating the average MSS payload per second. If it goes below 'net.inet.tcp.minmss' and the pps rate is above 'net.inet.tcp.minmssoverload' defaulting to 1000 this particular TCP connection is resetted and dropped. MITRE CVE: CAN-2004-0002 Reviewed by: sam (mentor) MFC after: 1 day
2004-01-08 17:40:07 +00:00
u_long tcps_minmssdrops; /* average minmss too low drops */
1994-05-24 10:09:53 +00:00
u_long tcps_closed; /* conn. closed (includes drops) */
u_long tcps_segstimed; /* segs where we tried to get rtt */
u_long tcps_rttupdated; /* times we succeeded */
u_long tcps_delack; /* delayed acks sent */
u_long tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
u_long tcps_rexmttimeo; /* retransmit timeouts */
u_long tcps_persisttimeo; /* persist timeouts */
u_long tcps_keeptimeo; /* keepalive timeouts */
u_long tcps_keepprobe; /* keepalive probes sent */
u_long tcps_keepdrops; /* connections dropped in keepalive */
u_long tcps_sndtotal; /* total packets sent */
u_long tcps_sndpack; /* data packets sent */
u_long tcps_sndbyte; /* data bytes sent */
u_long tcps_sndrexmitpack; /* data packets retransmitted */
u_long tcps_sndrexmitbyte; /* data bytes retransmitted */
u_long tcps_sndrexmitbad; /* unnecessary packet retransmissions */
1994-05-24 10:09:53 +00:00
u_long tcps_sndacks; /* ack-only packets sent */
u_long tcps_sndprobe; /* window probes sent */
u_long tcps_sndurg; /* packets sent with URG only */
u_long tcps_sndwinup; /* window update-only packets sent */
u_long tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
u_long tcps_rcvtotal; /* total packets received */
u_long tcps_rcvpack; /* packets received in sequence */
u_long tcps_rcvbyte; /* bytes received in sequence */
u_long tcps_rcvbadsum; /* packets received with ccksum errs */
u_long tcps_rcvbadoff; /* packets received with bad offset */
u_long tcps_rcvmemdrop; /* packets dropped for lack of memory */
1994-05-24 10:09:53 +00:00
u_long tcps_rcvshort; /* packets received too short */
u_long tcps_rcvduppack; /* duplicate-only packets received */
u_long tcps_rcvdupbyte; /* duplicate-only bytes received */
u_long tcps_rcvpartduppack; /* packets with some duplicate data */
u_long tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
u_long tcps_rcvoopack; /* out-of-order packets received */
u_long tcps_rcvoobyte; /* out-of-order bytes received */
u_long tcps_rcvpackafterwin; /* packets with data after window */
u_long tcps_rcvbyteafterwin; /* bytes rcvd after window */
u_long tcps_rcvafterclose; /* packets rcvd after "close" */
u_long tcps_rcvwinprobe; /* rcvd window probe packets */
u_long tcps_rcvdupack; /* rcvd duplicate acks */
u_long tcps_rcvacktoomuch; /* rcvd acks for unsent data */
u_long tcps_rcvackpack; /* rcvd ack packets */
u_long tcps_rcvackbyte; /* bytes acked by rcvd acks */
u_long tcps_rcvwinupd; /* rcvd window update packets */
u_long tcps_pawsdrop; /* segments dropped due to PAWS */
u_long tcps_predack; /* times hdr predict ok for acks */
u_long tcps_preddat; /* times hdr predict ok for data pkts */
u_long tcps_pcbcachemiss;
u_long tcps_cachedrtt; /* times cached RTT in route updated */
u_long tcps_cachedrttvar; /* times cached rttvar updated */
u_long tcps_cachedssthresh; /* times cached ssthresh updated */
u_long tcps_usedrtt; /* times RTT initialized from route */
u_long tcps_usedrttvar; /* times RTTVAR initialized from rt */
u_long tcps_usedssthresh; /* times ssthresh initialized from rt*/
u_long tcps_persistdrop; /* timeout in persist state */
u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */
u_long tcps_mturesent; /* resends due to MTU discovery */
u_long tcps_listendrop; /* listen queue overflows */
u_long tcps_sc_added; /* entry added to syncache */
u_long tcps_sc_retransmitted; /* syncache entry was retransmitted */
u_long tcps_sc_dupsyn; /* duplicate SYN packet */
u_long tcps_sc_dropped; /* could not reply to packet */
u_long tcps_sc_completed; /* successful extraction of entry */
u_long tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
u_long tcps_sc_cacheoverflow; /* syncache cache limit hit */
u_long tcps_sc_reset; /* RST removed entry from syncache */
u_long tcps_sc_stale; /* timed out or listen socket gone */
u_long tcps_sc_aborted; /* syncache entry aborted */
u_long tcps_sc_badack; /* removed due to bad ACK */
u_long tcps_sc_unreach; /* ICMP unreachable received */
u_long tcps_sc_zonefail; /* zalloc() failed */
u_long tcps_sc_sendcookie; /* SYN cookie sent */
u_long tcps_sc_recvcookie; /* SYN cookie received */
u_long tcps_hc_added; /* entry added to hostcache */
u_long tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */
1994-05-24 10:09:53 +00:00
};
/*
* TCB structure exported to user-land via sysctl(3).
* Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
* included. Not all of our clients do.
*/
#if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
struct xtcpcb {
size_t xt_len;
struct inpcb xt_inp;
struct tcpcb xt_tp;
struct xsocket xt_socket;
u_quad_t xt_alignment_hack;
};
#endif
/*
* Names for TCP sysctl objects
*/
#define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
#define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
#define TCPCTL_MSSDFLT 3 /* MSS default */
#define TCPCTL_STATS 4 /* statistics (read-only) */
#define TCPCTL_RTTDFLT 5 /* default RTT estimate */
#define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
#define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
#define TCPCTL_SENDSPACE 8 /* send buffer space */
#define TCPCTL_RECVSPACE 9 /* receive buffer space */
#define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
#define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
#define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
#define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
#define TCPCTL_MAXID 14
#define TCPCTL_NAMES { \
{ 0, 0 }, \
{ "rfc1323", CTLTYPE_INT }, \
{ "rfc1644", CTLTYPE_INT }, \
{ "mssdflt", CTLTYPE_INT }, \
{ "stats", CTLTYPE_STRUCT }, \
{ "rttdflt", CTLTYPE_INT }, \
{ "keepidle", CTLTYPE_INT }, \
{ "keepintvl", CTLTYPE_INT }, \
{ "sendspace", CTLTYPE_INT }, \
{ "recvspace", CTLTYPE_INT }, \
{ "keepinit", CTLTYPE_INT }, \
{ "pcblist", CTLTYPE_STRUCT }, \
{ "delacktime", CTLTYPE_INT }, \
{ "v6mssdflt", CTLTYPE_INT }, \
{ "maxid", CTLTYPE_INT }, \
}
#ifdef _KERNEL
#ifdef SYSCTL_DECL
SYSCTL_DECL(_net_inet_tcp);
#endif
extern struct inpcbhead tcb; /* head of queue of active tcpcb's */
extern struct inpcbinfo tcbinfo;
extern struct tcpstat tcpstat; /* tcp statistics */
extern int tcp_mssdflt; /* XXX */
Limiters and sanity checks for TCP MSS (maximum segement size) resource exhaustion attacks. For network link optimization TCP can adjust its MSS and thus packet size according to the observed path MTU. This is done dynamically based on feedback from the remote host and network components along the packet path. This information can be abused to pretend an extremely low path MTU. The resource exhaustion works in two ways: o during tcp connection setup the advertized local MSS is exchanged between the endpoints. The remote endpoint can set this arbitrarily low (except for a minimum MTU of 64 octets enforced in the BSD code). When the local host is sending data it is forced to send many small IP packets instead of a large one. For example instead of the normal TCP payload size of 1448 it forces TCP payload size of 12 (MTU 64) and thus we have a 120 times increase in workload and packets. On fast links this quickly saturates the local CPU and may also hit pps processing limites of network components along the path. This type of attack is particularly effective for servers where the attacker can download large files (WWW and FTP). We mitigate it by enforcing a minimum MTU settable by sysctl net.inet.tcp.minmss defaulting to 256 octets. o the local host is reveiving data on a TCP connection from the remote host. The local host has no control over the packet size the remote host is sending. The remote host may chose to do what is described in the first attack and send the data in packets with an TCP payload of at least one byte. For each packet the tcp_input() function will be entered, the packet is processed and a sowakeup() is signalled to the connected process. For example an attack with 2 Mbit/s gives 4716 packets per second and the same amount of sowakeup()s to the process (and context switches). This type of attack is particularly effective for servers where the attacker can upload large amounts of data. Normally this is the case with WWW server where large POSTs can be made. We mitigate this by calculating the average MSS payload per second. If it goes below 'net.inet.tcp.minmss' and the pps rate is above 'net.inet.tcp.minmssoverload' defaulting to 1000 this particular TCP connection is resetted and dropped. MITRE CVE: CAN-2004-0002 Reviewed by: sam (mentor) MFC after: 1 day
2004-01-08 17:40:07 +00:00
extern int tcp_minmss;
extern int tcp_minmssoverload;
extern int tcp_delack_enabled;
extern int tcp_do_newreno;
extern int path_mtu_discovery;
extern int ss_fltsz;
extern int ss_fltsz_local;
1994-05-24 10:09:53 +00:00
2002-03-19 21:25:46 +00:00
void tcp_canceltimers(struct tcpcb *);
1994-05-24 10:09:53 +00:00
struct tcpcb *
2002-03-19 21:25:46 +00:00
tcp_close(struct tcpcb *);
void tcp_twstart(struct tcpcb *);
int tcp_twrecycleable(struct tcptw *tw);
struct tcptw *
tcp_twclose(struct tcptw *_tw, int _reuse);
2002-03-19 21:25:46 +00:00
void tcp_ctlinput(int, struct sockaddr *, void *);
int tcp_ctloutput(struct socket *, struct sockopt *);
1994-05-24 10:09:53 +00:00
struct tcpcb *
2002-03-19 21:25:46 +00:00
tcp_drop(struct tcpcb *, int);
void tcp_drain(void);
void tcp_fasttimo(void);
void tcp_init(void);
void tcp_fini(void *);
void tcp_reass_init(void);
2002-03-19 21:25:46 +00:00
void tcp_input(struct mbuf *, int);
u_long tcp_maxmtu(struct in_conninfo *);
u_long tcp_maxmtu6(struct in_conninfo *);
2002-03-19 21:25:46 +00:00
void tcp_mss(struct tcpcb *, int);
int tcp_mssopt(struct in_conninfo *);
struct inpcb *
tcp_drop_syn_sent(struct inpcb *, int);
struct inpcb *
tcp_mtudisc(struct inpcb *, int);
1994-05-24 10:09:53 +00:00
struct tcpcb *
2002-03-19 21:25:46 +00:00
tcp_newtcpcb(struct inpcb *);
int tcp_output(struct tcpcb *);
struct inpcb *
tcp_quench(struct inpcb *, int);
2002-03-19 21:25:46 +00:00
void tcp_respond(struct tcpcb *, void *,
struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
int tcp_twrespond(struct tcptw *, int);
2002-03-19 21:25:46 +00:00
void tcp_setpersist(struct tcpcb *);
2004-02-13 18:21:45 +00:00
#ifdef TCP_SIGNATURE
int tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int);
#endif
2002-03-19 21:25:46 +00:00
void tcp_slowtimo(void);
struct tcptemp *
tcpip_maketemplate(struct inpcb *);
void tcpip_fillheaders(struct inpcb *, void *, void *);
1994-05-24 10:09:53 +00:00
struct tcpcb *
2002-03-19 21:25:46 +00:00
tcp_timers(struct tcpcb *, int);
void tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int);
void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
void syncache_init(void);
void syncache_unreach(struct in_conninfo *, struct tcphdr *);
int syncache_expand(struct in_conninfo *, struct tcphdr *,
struct socket **, struct mbuf *);
int syncache_add(struct in_conninfo *, struct tcpopt *,
struct tcphdr *, struct socket **, struct mbuf *);
void syncache_chkrst(struct in_conninfo *, struct tcphdr *);
void syncache_badack(struct in_conninfo *);
/*
* All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
*/
void tcp_hc_init(void);
void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *);
u_long tcp_hc_getmtu(struct in_conninfo *);
void tcp_hc_gettao(struct in_conninfo *, struct rmxp_tao *);
void tcp_hc_updatemtu(struct in_conninfo *, u_long);
void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *);
void tcp_hc_updatetao(struct in_conninfo *, int, tcp_cc, u_short);
/* update which tao field */
#define TCP_HC_TAO_CC 0x1
#define TCP_HC_TAO_CCSENT 0x2
#define TCP_HC_TAO_MSSOPT 0x3
extern struct pr_usrreqs tcp_usrreqs;
extern u_long tcp_sendspace;
extern u_long tcp_recvspace;
2002-03-19 21:25:46 +00:00
tcp_seq tcp_new_isn(struct tcpcb *);
#endif /* _KERNEL */
1994-08-21 05:27:42 +00:00
#endif /* _NETINET_TCP_VAR_H_ */