2005-01-07 01:45:51 +00:00
|
|
|
/*-
|
1995-10-03 16:54:17 +00:00
|
|
|
* Copyright (c) 1982, 1986, 1988, 1990, 1993, 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.
|
|
|
|
|
*
|
1995-10-03 16:54:17 +00:00
|
|
|
* @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
|
2007-10-07 20:44:24 +00:00
|
|
|
#include <sys/cdefs.h>
|
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
|
1998-01-25 04:23:33 +00:00
|
|
|
#include "opt_compat.h"
|
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
|
|
|
#include "opt_inet.h"
|
1999-12-07 17:39:16 +00:00
|
|
|
#include "opt_inet6.h"
|
2000-01-15 14:56:38 +00:00
|
|
|
#include "opt_ipsec.h"
|
1997-09-16 18:36:06 +00:00
|
|
|
#include "opt_tcpdebug.h"
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
|
#include <sys/systm.h>
|
1999-08-30 21:17:07 +00:00
|
|
|
#include <sys/callout.h>
|
2010-12-28 12:13:30 +00:00
|
|
|
#include <sys/hhook.h>
|
1995-11-09 20:23:09 +00:00
|
|
|
#include <sys/kernel.h>
|
2010-12-28 12:13:30 +00:00
|
|
|
#include <sys/khelp.h>
|
1995-11-09 20:23:09 +00:00
|
|
|
#include <sys/sysctl.h>
|
2009-06-17 15:01:01 +00:00
|
|
|
#include <sys/jail.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/malloc.h>
|
|
|
|
|
#include <sys/mbuf.h>
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#include <sys/domain.h>
|
|
|
|
|
#endif
|
2006-11-06 13:42:10 +00:00
|
|
|
#include <sys/priv.h>
|
1999-07-11 18:32:46 +00:00
|
|
|
#include <sys/proc.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/socket.h>
|
|
|
|
|
#include <sys/socketvar.h>
|
|
|
|
|
#include <sys/protosw.h>
|
2001-04-17 18:08:01 +00:00
|
|
|
#include <sys/random.h>
|
1998-03-28 10:18:26 +00:00
|
|
|
|
2002-03-20 05:48:55 +00:00
|
|
|
#include <vm/uma.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
|
#include <net/route.h>
|
|
|
|
|
#include <net/if.h>
|
2009-08-01 19:26:27 +00:00
|
|
|
#include <net/vnet.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2010-11-12 06:41:55 +00:00
|
|
|
#include <netinet/cc.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <netinet/in.h>
|
2011-04-30 11:21:29 +00:00
|
|
|
#include <netinet/in_pcb.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <netinet/in_systm.h>
|
2011-04-30 11:21:29 +00:00
|
|
|
#include <netinet/in_var.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <netinet/ip.h>
|
2011-04-30 11:21:29 +00:00
|
|
|
#include <netinet/ip_icmp.h>
|
|
|
|
|
#include <netinet/ip_var.h>
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#include <netinet/ip6.h>
|
|
|
|
|
#include <netinet6/in6_pcb.h>
|
|
|
|
|
#include <netinet6/ip6_var.h>
|
2005-07-25 12:31:43 +00:00
|
|
|
#include <netinet6/scope6_var.h>
|
2003-11-20 20:07:39 +00:00
|
|
|
#include <netinet6/nd6.h>
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <netinet/tcp_fsm.h>
|
|
|
|
|
#include <netinet/tcp_seq.h>
|
|
|
|
|
#include <netinet/tcp_timer.h>
|
|
|
|
|
#include <netinet/tcp_var.h>
|
2007-07-27 00:57:06 +00:00
|
|
|
#include <netinet/tcp_syncache.h>
|
2007-12-18 22:59:07 +00:00
|
|
|
#include <netinet/tcp_offload.h>
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#include <netinet6/tcp6_var.h>
|
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <netinet/tcpip.h>
|
1995-02-09 23:13:27 +00:00
|
|
|
#ifdef TCPDEBUG
|
|
|
|
|
#include <netinet/tcp_debug.h>
|
|
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET6
|
2000-01-09 19:17:30 +00:00
|
|
|
#include <netinet6/ip6protosw.h>
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2000-01-09 19:17:30 +00:00
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#ifdef IPSEC
|
2002-10-16 02:25:05 +00:00
|
|
|
#include <netipsec/ipsec.h>
|
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
|
|
|
#include <netipsec/xform.h>
|
2002-10-16 02:25:05 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#include <netipsec/ipsec6.h>
|
|
|
|
|
#endif
|
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
|
|
|
#include <netipsec/key.h>
|
2008-09-13 17:26:46 +00:00
|
|
|
#include <sys/syslog.h>
|
2007-07-03 12:13:45 +00:00
|
|
|
#endif /*IPSEC*/
|
2002-10-16 02:25:05 +00:00
|
|
|
|
2000-03-27 19:14:27 +00:00
|
|
|
#include <machine/in_cksum.h>
|
2001-08-22 00:58:16 +00:00
|
|
|
#include <sys/md5.h>
|
2000-03-27 19:14:27 +00:00
|
|
|
|
2006-10-22 11:52:19 +00:00
|
|
|
#include <security/mac/mac_framework.h>
|
|
|
|
|
|
2010-04-29 11:52:42 +00:00
|
|
|
VNET_DEFINE(int, tcp_mssdflt) = TCP_MSS;
|
1999-11-05 14:41:39 +00:00
|
|
|
#ifdef INET6
|
2010-04-29 11:52:42 +00:00
|
|
|
VNET_DEFINE(int, tcp_v6mssdflt) = TCP6_MSS;
|
2008-09-07 14:44:55 +00:00
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
sysctl_net_inet_tcp_mss_check(SYSCTL_HANDLER_ARGS)
|
|
|
|
|
{
|
|
|
|
|
int error, new;
|
|
|
|
|
|
2008-09-07 15:20:21 +00:00
|
|
|
new = V_tcp_mssdflt;
|
2008-09-07 14:44:55 +00:00
|
|
|
error = sysctl_handle_int(oidp, &new, 0, req);
|
|
|
|
|
if (error == 0 && req->newptr) {
|
|
|
|
|
if (new < TCP_MINMSS)
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
else
|
2008-09-07 15:20:21 +00:00
|
|
|
V_tcp_mssdflt = new;
|
2008-09-07 14:44:55 +00:00
|
|
|
}
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_PROC(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt,
|
|
|
|
|
CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(tcp_mssdflt), 0,
|
2008-11-26 22:32:07 +00:00
|
|
|
&sysctl_net_inet_tcp_mss_check, "I",
|
|
|
|
|
"Default TCP Maximum Segment Size");
|
2008-09-07 14:44:55 +00:00
|
|
|
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
static int
|
|
|
|
|
sysctl_net_inet_tcp_mss_v6_check(SYSCTL_HANDLER_ARGS)
|
|
|
|
|
{
|
|
|
|
|
int error, new;
|
|
|
|
|
|
2008-09-07 15:20:21 +00:00
|
|
|
new = V_tcp_v6mssdflt;
|
2008-09-07 14:44:55 +00:00
|
|
|
error = sysctl_handle_int(oidp, &new, 0, req);
|
|
|
|
|
if (error == 0 && req->newptr) {
|
|
|
|
|
if (new < TCP_MINMSS)
|
|
|
|
|
error = EINVAL;
|
|
|
|
|
else
|
2008-09-07 15:20:21 +00:00
|
|
|
V_tcp_v6mssdflt = new;
|
2008-09-07 14:44:55 +00:00
|
|
|
}
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_PROC(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt,
|
|
|
|
|
CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(tcp_v6mssdflt), 0,
|
2008-11-26 22:32:07 +00:00
|
|
|
&sysctl_net_inet_tcp_mss_v6_check, "I",
|
|
|
|
|
"Default TCP Maximum Segment Size for IPv6");
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET6 */
|
1999-11-05 14:41:39 +00:00
|
|
|
|
2004-01-08 17:40:07 +00:00
|
|
|
/*
|
|
|
|
|
* Minimum MSS we accept and use. This prevents DoS attacks where
|
|
|
|
|
* we are forced to a ridiculous low MSS like 20 and send hundreds
|
|
|
|
|
* of packets instead of one. The effect scales with the available
|
|
|
|
|
* bandwidth and quickly saturates the CPU and network interface
|
|
|
|
|
* with packet generation and sending. Set to zero to disable MINMSS
|
|
|
|
|
* checking. This setting prevents us from sending too small packets.
|
|
|
|
|
*/
|
2010-04-29 11:52:42 +00:00
|
|
|
VNET_DEFINE(int, tcp_minmss) = TCP_MINMSS;
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW,
|
|
|
|
|
&VNET_NAME(tcp_minmss), 0,
|
|
|
|
|
"Minmum TCP Maximum Segment Size");
|
2004-01-08 17:40:07 +00:00
|
|
|
|
2010-04-29 11:52:42 +00:00
|
|
|
VNET_DEFINE(int, tcp_do_rfc1323) = 1;
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW,
|
|
|
|
|
&VNET_NAME(tcp_do_rfc1323), 0,
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
"Enable rfc1323 (high performance TCP) extensions");
|
1995-11-14 20:34:56 +00:00
|
|
|
|
2007-10-05 22:39:44 +00:00
|
|
|
static int tcp_log_debug = 0;
|
2007-07-28 12:20:39 +00:00
|
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_debug, CTLFLAG_RW,
|
|
|
|
|
&tcp_log_debug, 0, "Log errors caused by incoming TCP segments");
|
|
|
|
|
|
1999-08-26 19:52:17 +00:00
|
|
|
static int tcp_tcbhashsize = 0;
|
2003-10-21 18:28:36 +00:00
|
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN,
|
2007-03-19 19:00:51 +00:00
|
|
|
&tcp_tcbhashsize, 0, "Size of TCP control-block hashtable");
|
1999-08-26 19:52:17 +00:00
|
|
|
|
1999-12-28 23:18:33 +00:00
|
|
|
static int do_tcpdrain = 1;
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0,
|
2007-03-19 19:00:51 +00:00
|
|
|
"Enable tcp_drain routine for extra help when low on mbufs");
|
1999-12-28 23:18:33 +00:00
|
|
|
|
2011-01-12 19:53:50 +00:00
|
|
|
SYSCTL_VNET_UINT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD,
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
&VNET_NAME(tcbinfo.ipi_count), 0, "Number of active PCBs");
|
1998-05-15 20:11:40 +00:00
|
|
|
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(int, icmp_may_rst) = 1;
|
2010-04-29 11:52:42 +00:00
|
|
|
#define V_icmp_may_rst VNET(icmp_may_rst)
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW,
|
|
|
|
|
&VNET_NAME(icmp_may_rst), 0,
|
2001-02-23 20:51:46 +00:00
|
|
|
"Certain ICMP unreachable messages may abort connections in SYN_SENT");
|
We currently does not react to ICMP administratively prohibited
messages send by routers when they deny our traffic, this causes
a timeout when trying to connect to TCP ports/services on a remote
host, which is blocked by routers or firewalls.
rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually
requi re that we treat such a message for a TCP session, that we
treat it like if we had recieved a RST.
quote begin.
A Destination Unreachable message that is received MUST be
reported to the transport layer. The transport layer SHOULD
use the information appropriately; for example, see Sections
4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol
that has its own mechanism for notifying the sender that a
port is unreachable (e.g., TCP, which sends RST segments)
MUST nevertheless accept an ICMP Port Unreachable for the
same purpose.
quote end.
I've written a small extension that implement this, it also create
a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if
this new behaviour is activated.
When it's activated (set to 1) we'll treat a ICMP administratively
prohibited message (icmp type 3 code 9, 10 and 13) for a TCP
sessions, as if we recived a TCP RST, but only if the TCP session
is in SYN_SENT state.
The reason for only reacting when in SYN_SENT state, is that this
will solve the problem, and at the same time minimize the risk of
this being abused.
I suggest that we enable this new behaviour by default, but it
would be a change of current behaviour, so if people prefer to
leave it disabled by default, at least for now, this would be ok
for me, the attached diff actually have the sysctl set to 0 by
default.
PR: 23086
Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
|
|
|
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(int, tcp_isn_reseed_interval) = 0;
|
2010-04-29 11:52:42 +00:00
|
|
|
#define V_tcp_isn_reseed_interval VNET(tcp_isn_reseed_interval)
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW,
|
|
|
|
|
&VNET_NAME(tcp_isn_reseed_interval), 0,
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
"Seconds between reseeding of ISN secret");
|
2001-07-08 02:20:47 +00:00
|
|
|
|
2009-09-15 22:23:45 +00:00
|
|
|
static int tcp_soreceive_stream = 0;
|
|
|
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, soreceive_stream, CTLFLAG_RDTUN,
|
|
|
|
|
&tcp_soreceive_stream, 0, "Using soreceive_stream for TCP sockets");
|
|
|
|
|
|
2011-04-25 17:13:40 +00:00
|
|
|
#ifdef TCP_SIGNATURE
|
|
|
|
|
static int tcp_sig_checksigs = 1;
|
|
|
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, signature_verify_input, CTLFLAG_RW,
|
|
|
|
|
&tcp_sig_checksigs, 0, "Verify RFC2385 digests on inbound traffic");
|
|
|
|
|
#endif
|
|
|
|
|
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
VNET_DEFINE(uma_zone_t, sack_hole_zone);
|
2009-07-16 21:13:04 +00:00
|
|
|
#define V_sack_hole_zone VNET(sack_hole_zone)
|
2004-06-23 21:04:37 +00:00
|
|
|
|
2010-12-28 12:13:30 +00:00
|
|
|
VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]);
|
|
|
|
|
|
2002-06-14 08:35:21 +00:00
|
|
|
static struct inpcb *tcp_notify(struct inpcb *, int);
|
2010-08-18 17:39:47 +00:00
|
|
|
static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th,
|
|
|
|
|
void *ip4hdr, const void *ip6hdr);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-04-09 01:29:31 +00:00
|
|
|
/*
|
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
|
|
|
* Target size of TCP PCB hash tables. Must be a power of two.
|
1999-02-03 08:59:30 +00:00
|
|
|
*
|
|
|
|
|
* Note that this can be overridden by the kernel environment
|
|
|
|
|
* variable net.inet.tcp.tcbhashsize
|
1995-04-09 01:29:31 +00:00
|
|
|
*/
|
|
|
|
|
#ifndef TCBHASHSIZE
|
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
|
|
|
#define TCBHASHSIZE 512
|
1995-04-09 01:29:31 +00:00
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1998-03-24 18:06:34 +00:00
|
|
|
/*
|
2003-02-19 22:32:43 +00:00
|
|
|
* XXX
|
|
|
|
|
* Callouts should be moved into struct tcp directly. They are currently
|
2003-12-17 16:12:01 +00:00
|
|
|
* separate because the tcpcb structure is exported to userland for sysctl
|
2003-02-19 22:32:43 +00:00
|
|
|
* parsing purposes, which do not know about callouts.
|
1998-03-24 18:06:34 +00:00
|
|
|
*/
|
2007-04-11 09:45:16 +00:00
|
|
|
struct tcpcb_mem {
|
|
|
|
|
struct tcpcb tcb;
|
2007-09-24 05:26:24 +00:00
|
|
|
struct tcp_timer tt;
|
2010-12-28 01:38:52 +00:00
|
|
|
struct cc_var ccv;
|
2010-12-28 12:13:30 +00:00
|
|
|
struct osd osd;
|
2003-02-19 22:32:43 +00:00
|
|
|
};
|
|
|
|
|
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(uma_zone_t, tcpcb_zone);
|
2009-07-16 21:13:04 +00:00
|
|
|
#define V_tcpcb_zone VNET(tcpcb_zone)
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
MALLOC_DEFINE(M_TCPLOG, "tcplog", "TCP address and flags print buffers");
|
2006-04-22 19:23:24 +00:00
|
|
|
static struct mtx isn_mtx;
|
1998-03-24 18:06:34 +00:00
|
|
|
|
2006-04-23 12:27:42 +00:00
|
|
|
#define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF)
|
|
|
|
|
#define ISN_LOCK() mtx_lock(&isn_mtx)
|
|
|
|
|
#define ISN_UNLOCK() mtx_unlock(&isn_mtx)
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
2006-04-03 12:59:27 +00:00
|
|
|
* TCP initialization.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2006-04-21 09:25:40 +00:00
|
|
|
static void
|
|
|
|
|
tcp_zone_change(void *tag)
|
|
|
|
|
{
|
|
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
uma_zone_set_max(V_tcbinfo.ipi_zone, maxsockets);
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
uma_zone_set_max(V_tcpcb_zone, maxsockets);
|
2007-05-13 22:16:13 +00:00
|
|
|
tcp_tw_zone_change();
|
2006-04-21 09:25:40 +00:00
|
|
|
}
|
|
|
|
|
|
2006-07-18 22:34:27 +00:00
|
|
|
static int
|
|
|
|
|
tcp_inpcb_init(void *mem, int size, int flags)
|
|
|
|
|
{
|
2006-12-30 17:53:28 +00:00
|
|
|
struct inpcb *inp = mem;
|
|
|
|
|
|
2006-07-18 22:34:27 +00:00
|
|
|
INP_LOCK_INIT(inp, "inp", "tcpinp");
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_init(void)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2008-11-19 09:39:34 +00:00
|
|
|
int hashsize;
|
|
|
|
|
|
2010-12-28 12:13:30 +00:00
|
|
|
if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN,
|
|
|
|
|
&V_tcp_hhh[HHOOK_TCP_EST_IN], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0)
|
|
|
|
|
printf("%s: WARNING: unable to register helper hook\n", __func__);
|
|
|
|
|
if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT,
|
|
|
|
|
&V_tcp_hhh[HHOOK_TCP_EST_OUT], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0)
|
|
|
|
|
printf("%s: WARNING: unable to register helper hook\n", __func__);
|
|
|
|
|
|
2008-11-19 09:39:34 +00:00
|
|
|
hashsize = TCBHASHSIZE;
|
2001-06-08 05:24:21 +00:00
|
|
|
TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize);
|
1999-02-04 03:27:43 +00:00
|
|
|
if (!powerof2(hashsize)) {
|
1999-02-03 08:59:30 +00:00
|
|
|
printf("WARNING: TCB hash size not a power of 2\n");
|
|
|
|
|
hashsize = 512; /* safe default */
|
|
|
|
|
}
|
2010-03-14 18:59:11 +00:00
|
|
|
in_pcbinfo_init(&V_tcbinfo, "tcp", &V_tcb, hashsize, hashsize,
|
Implement a CPU-affine TCP and UDP connection lookup data structure,
struct inpcbgroup. pcbgroups, or "connection groups", supplement the
existing inpcbinfo connection hash table, which when pcbgroups are
enabled, might now be thought of more usefully as a per-protocol
4-tuple reservation table.
Connections are assigned to connection groups base on a hash of their
4-tuple; wildcard sockets require special handling, and are members
of all connection groups. During a connection lookup, a
per-connection group lock is employed rather than the global pcbinfo
lock. By aligning connection groups with input path processing,
connection groups take on an effective CPU affinity, especially when
aligned with RSS work placement (see a forthcoming commit for
details). This eliminates cache line migration associated with
global, protocol-layer data structures in steady state TCP and UDP
processing (with the exception of protocol-layer statistics; further
commit to follow).
Elements of this approach were inspired by Willman, Rixner, and Cox's
2006 USENIX paper, "An Evaluation of Network Stack Parallelization
Strategies in Modern Operating Systems". However, there are also
significant differences: we maintain the inpcb lock, rather than using
the connection group lock for per-connection state.
Likewise, the focus of this implementation is alignment with NIC
packet distribution strategies such as RSS, rather than pure software
strategies. Despite that focus, software distribution is supported
through the parallel netisr implementation, and works well in
configurations where the number of hardware threads is greater than
the number of NIC input queues, such as in the RMI XLR threaded MIPS
architecture.
Another important difference is the continued maintenance of existing
hash tables as "reservation tables" -- these are useful both to
distinguish the resource allocation aspect of protocol name management
and the more common-case lookup aspect. In configurations where
connection tables are aligned with hardware hashes, it is desirable to
use the traditional lookup tables for loopback or encapsulated traffic
rather than take the expense of hardware hashes that are hard to
implement efficiently in software (such as RSS Toeplitz).
Connection group support is enabled by compiling "options PCBGROUP"
into your kernel configuration; for the time being, this is an
experimental feature, and hence is not enabled by default.
Subject to the limited MFCability of change dependencies in inpcb,
and its change to the inpcbinfo init function signature, this change
in principle could be merged to FreeBSD 8.x.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-06-06 12:55:02 +00:00
|
|
|
"tcp_inpcb", tcp_inpcb_init, NULL, UMA_ZONE_NOFREE,
|
|
|
|
|
IPI_HASHFIELDS_4TUPLE);
|
2010-03-14 18:59:11 +00:00
|
|
|
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
/*
|
|
|
|
|
* These have to be type stable for the benefit of the timers.
|
|
|
|
|
*/
|
|
|
|
|
V_tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem),
|
|
|
|
|
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
|
|
|
|
|
uma_zone_set_max(V_tcpcb_zone, maxsockets);
|
2010-04-29 11:52:42 +00:00
|
|
|
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
tcp_tw_init();
|
|
|
|
|
syncache_init();
|
|
|
|
|
tcp_hc_init();
|
|
|
|
|
tcp_reass_init();
|
2010-04-29 11:52:42 +00:00
|
|
|
|
|
|
|
|
TUNABLE_INT_FETCH("net.inet.tcp.sack.enable", &V_tcp_do_sack);
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
V_sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole),
|
|
|
|
|
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
|
|
|
|
|
|
|
|
|
|
/* Skip initialization of globals for non-default instances. */
|
|
|
|
|
if (!IS_DEFAULT_VNET(curvnet))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
/* XXX virtualize those bellow? */
|
|
|
|
|
tcp_delacktime = TCPTV_DELACK;
|
|
|
|
|
tcp_keepinit = TCPTV_KEEP_INIT;
|
|
|
|
|
tcp_keepidle = TCPTV_KEEP_IDLE;
|
|
|
|
|
tcp_keepintvl = TCPTV_KEEPINTVL;
|
|
|
|
|
tcp_maxpersistidle = TCPTV_KEEP_IDLE;
|
|
|
|
|
tcp_msl = TCPTV_MSL;
|
|
|
|
|
tcp_rexmit_min = TCPTV_MIN;
|
|
|
|
|
if (tcp_rexmit_min < 1)
|
|
|
|
|
tcp_rexmit_min = 1;
|
|
|
|
|
tcp_rexmit_slop = TCPTV_CPU_VAR;
|
|
|
|
|
tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT;
|
|
|
|
|
tcp_tcbhashsize = hashsize;
|
|
|
|
|
|
2009-09-15 22:23:45 +00:00
|
|
|
TUNABLE_INT_FETCH("net.inet.tcp.soreceive_stream", &tcp_soreceive_stream);
|
|
|
|
|
if (tcp_soreceive_stream) {
|
2011-07-14 13:44:48 +00:00
|
|
|
#ifdef INET
|
2009-09-15 22:23:45 +00:00
|
|
|
tcp_usrreqs.pru_soreceive = soreceive_stream;
|
2011-07-14 13:44:48 +00:00
|
|
|
#endif
|
2011-07-07 10:37:14 +00:00
|
|
|
#ifdef INET6
|
2009-09-15 22:23:45 +00:00
|
|
|
tcp6_usrreqs.pru_soreceive = soreceive_stream;
|
2011-07-07 10:37:14 +00:00
|
|
|
#endif /* INET6 */
|
2009-09-15 22:23:45 +00:00
|
|
|
}
|
|
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
|
|
|
|
|
#else /* INET6 */
|
|
|
|
|
#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
if (max_protohdr < TCP_MINPROTOHDR)
|
|
|
|
|
max_protohdr = TCP_MINPROTOHDR;
|
|
|
|
|
if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
|
1994-05-24 10:09:53 +00:00
|
|
|
panic("tcp_init");
|
2000-01-09 19:17:30 +00:00
|
|
|
#undef TCP_MINPROTOHDR
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
|
2006-04-23 12:27:42 +00:00
|
|
|
ISN_LOCK_INIT();
|
2004-04-20 06:33:39 +00:00
|
|
|
EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL,
|
|
|
|
|
SHUTDOWN_PRI_DEFAULT);
|
2006-04-21 09:25:40 +00:00
|
|
|
EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL,
|
|
|
|
|
EVENTHANDLER_PRI_ANY);
|
2004-04-20 06:33:39 +00:00
|
|
|
}
|
|
|
|
|
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
#ifdef VIMAGE
|
|
|
|
|
void
|
|
|
|
|
tcp_destroy(void)
|
|
|
|
|
{
|
|
|
|
|
|
2010-03-07 15:58:44 +00:00
|
|
|
tcp_reass_destroy();
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
tcp_hc_destroy();
|
|
|
|
|
syncache_destroy();
|
2010-03-07 15:58:44 +00:00
|
|
|
tcp_tw_destroy();
|
2010-03-14 18:59:11 +00:00
|
|
|
in_pcbinfo_destroy(&V_tcbinfo);
|
2010-03-07 15:58:44 +00:00
|
|
|
uma_zdestroy(V_sack_hole_zone);
|
|
|
|
|
uma_zdestroy(V_tcpcb_zone);
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
2004-04-20 06:33:39 +00:00
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_fini(void *xtp)
|
2004-04-20 06:33:39 +00:00
|
|
|
{
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
2001-06-23 03:21:46 +00:00
|
|
|
* Fill in the IP and TCP headers for an outgoing packet, given the tcpcb.
|
|
|
|
|
* tcp_template used to store this data in mbufs, but we now recopy it out
|
|
|
|
|
* of the tcpcb each time to conserve mbufs.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2001-06-23 03:21:46 +00:00
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-02-19 22:18:06 +00:00
|
|
|
struct tcphdr *th = (struct tcphdr *)tcp_ptr;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2004-12-05 22:27:53 +00:00
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if ((inp->inp_vflag & INP_IPV6) != 0) {
|
2001-06-23 03:21:46 +00:00
|
|
|
struct ip6_hdr *ip6;
|
2000-01-09 19:17:30 +00:00
|
|
|
|
2001-06-23 03:21:46 +00:00
|
|
|
ip6 = (struct ip6_hdr *)ip_ptr;
|
2000-01-09 19:17:30 +00:00
|
|
|
ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
|
Another step assimilating IPv[46] PCB code - directly use
the inpcb names rather than the following IPv6 compat macros:
in6pcb,in6p_sp, in6p_ip6_nxt,in6p_flowinfo,in6p_vflag,
in6p_flags,in6p_socket,in6p_lport,in6p_fport,in6p_ppcb and
sotoin6pcb().
Apart from removing duplicate code in netipsec, this is a pure
whitespace, not a functional change.
Discussed with: rwatson
Reviewed by: rwatson (version before review requested changes)
MFC after: 4 weeks (set the timer and see then)
2008-12-15 21:50:54 +00:00
|
|
|
(inp->inp_flow & IPV6_FLOWINFO_MASK);
|
2000-01-09 19:17:30 +00:00
|
|
|
ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
|
|
|
|
|
(IPV6_VERSION & IPV6_VERSION_MASK);
|
|
|
|
|
ip6->ip6_nxt = IPPROTO_TCP;
|
2008-09-07 20:44:45 +00:00
|
|
|
ip6->ip6_plen = htons(sizeof(struct tcphdr));
|
2000-01-09 19:17:30 +00:00
|
|
|
ip6->ip6_src = inp->in6p_laddr;
|
|
|
|
|
ip6->ip6_dst = inp->in6p_faddr;
|
2011-04-30 11:21:29 +00:00
|
|
|
}
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
#if defined(INET6) && defined(INET)
|
|
|
|
|
else
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2001-06-23 03:21:46 +00:00
|
|
|
{
|
2003-02-19 22:18:06 +00:00
|
|
|
struct ip *ip;
|
|
|
|
|
|
|
|
|
|
ip = (struct ip *)ip_ptr;
|
|
|
|
|
ip->ip_v = IPVERSION;
|
|
|
|
|
ip->ip_hl = 5;
|
|
|
|
|
ip->ip_tos = inp->inp_ip_tos;
|
|
|
|
|
ip->ip_len = 0;
|
|
|
|
|
ip->ip_id = 0;
|
|
|
|
|
ip->ip_off = 0;
|
|
|
|
|
ip->ip_ttl = inp->inp_ip_ttl;
|
|
|
|
|
ip->ip_sum = 0;
|
|
|
|
|
ip->ip_p = IPPROTO_TCP;
|
|
|
|
|
ip->ip_src = inp->inp_laddr;
|
|
|
|
|
ip->ip_dst = inp->inp_faddr;
|
2001-06-23 03:21:46 +00:00
|
|
|
}
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
2003-02-19 22:18:06 +00:00
|
|
|
th->th_sport = inp->inp_lport;
|
|
|
|
|
th->th_dport = inp->inp_fport;
|
|
|
|
|
th->th_seq = 0;
|
|
|
|
|
th->th_ack = 0;
|
|
|
|
|
th->th_x2 = 0;
|
|
|
|
|
th->th_off = 5;
|
|
|
|
|
th->th_flags = 0;
|
|
|
|
|
th->th_win = 0;
|
|
|
|
|
th->th_urp = 0;
|
|
|
|
|
th->th_sum = 0; /* in_pseudo() is called later for ipv4 */
|
2001-06-23 03:21:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Create template to be used to send tcp packets on a connection.
|
|
|
|
|
* Allocates an mbuf and fills in a skeletal tcp/ip header. The only
|
|
|
|
|
* use for this function is in keepalives, which use tcp_respond.
|
|
|
|
|
*/
|
|
|
|
|
struct tcptemp *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcpip_maketemplate(struct inpcb *inp)
|
2001-06-23 03:21:46 +00:00
|
|
|
{
|
2008-06-02 14:20:26 +00:00
|
|
|
struct tcptemp *t;
|
2001-06-23 03:21:46 +00:00
|
|
|
|
2008-06-02 14:20:26 +00:00
|
|
|
t = malloc(sizeof(*t), M_TEMP, M_NOWAIT);
|
|
|
|
|
if (t == NULL)
|
|
|
|
|
return (NULL);
|
|
|
|
|
tcpip_fillheaders(inp, (void *)&t->tt_ipgen, (void *)&t->tt_t);
|
|
|
|
|
return (t);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Send a single message to the TCP at address specified by
|
2004-04-05 00:49:07 +00:00
|
|
|
* the given TCP/IP header. If m == NULL, then we make a copy
|
1994-05-24 10:09:53 +00:00
|
|
|
* of the tcpiphdr at ti and send directly to the addressed host.
|
|
|
|
|
* This is used to force keep alive messages out using the TCP
|
2001-06-23 03:21:46 +00:00
|
|
|
* template for a connection. If flags are given then we send
|
|
|
|
|
* a message back to the TCP which originated the * segment ti,
|
|
|
|
|
* and discard the mbuf containing it and any other attached mbufs.
|
1994-05-24 10:09:53 +00:00
|
|
|
*
|
|
|
|
|
* In any case the ack and sequence number of the transmitted
|
|
|
|
|
* segment are as specified by the parameters.
|
1997-12-19 03:36:15 +00:00
|
|
|
*
|
|
|
|
|
* NOTE: If m != NULL, then ti must point to *inside* the mbuf.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
void
|
2007-05-10 15:58:48 +00:00
|
|
|
tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m,
|
|
|
|
|
tcp_seq ack, tcp_seq seq, int flags)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2007-03-21 19:37:55 +00:00
|
|
|
int tlen;
|
1994-05-24 10:09:53 +00:00
|
|
|
int win = 0;
|
2000-01-09 19:17:30 +00:00
|
|
|
struct ip *ip;
|
|
|
|
|
struct tcphdr *nth;
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
struct ip6_hdr *ip6;
|
|
|
|
|
int isipv6;
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
int ipflags = 0;
|
2004-05-04 02:11:47 +00:00
|
|
|
struct inpcb *inp;
|
2000-01-09 19:17:30 +00:00
|
|
|
|
2002-08-01 03:54:43 +00:00
|
|
|
KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL"));
|
|
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
2011-04-30 11:21:29 +00:00
|
|
|
isipv6 = ((struct ip *)ipgen)->ip_v == (IPV6_VERSION >> 4);
|
2000-01-09 19:17:30 +00:00
|
|
|
ip6 = ipgen;
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
ip = ipgen;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2004-04-05 00:52:05 +00:00
|
|
|
if (tp != NULL) {
|
2003-11-08 22:59:22 +00:00
|
|
|
inp = tp->t_inpcb;
|
|
|
|
|
KASSERT(inp != NULL, ("tcp control block w/o inpcb"));
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2004-05-04 02:11:47 +00:00
|
|
|
} else
|
|
|
|
|
inp = NULL;
|
|
|
|
|
|
|
|
|
|
if (tp != NULL) {
|
2000-02-28 21:18:21 +00:00
|
|
|
if (!(flags & TH_RST)) {
|
2003-11-08 22:59:22 +00:00
|
|
|
win = sbspace(&inp->inp_socket->so_rcv);
|
2000-02-28 21:18:21 +00:00
|
|
|
if (win > (long)TCP_MAXWIN << tp->rcv_scale)
|
|
|
|
|
win = (long)TCP_MAXWIN << tp->rcv_scale;
|
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2004-04-05 00:49:07 +00:00
|
|
|
if (m == NULL) {
|
2005-11-02 13:46:32 +00:00
|
|
|
m = m_gethdr(M_DONTWAIT, MT_DATA);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (m == NULL)
|
|
|
|
|
return;
|
|
|
|
|
tlen = 0;
|
|
|
|
|
m->m_data += max_linkhdr;
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (isipv6) {
|
2004-08-16 18:32:07 +00:00
|
|
|
bcopy((caddr_t)ip6, mtod(m, caddr_t),
|
2000-01-09 19:17:30 +00:00
|
|
|
sizeof(struct ip6_hdr));
|
|
|
|
|
ip6 = mtod(m, struct ip6_hdr *);
|
|
|
|
|
nth = (struct tcphdr *)(ip6 + 1);
|
|
|
|
|
} else
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
{
|
|
|
|
|
bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
|
|
|
|
|
ip = mtod(m, struct ip *);
|
|
|
|
|
nth = (struct tcphdr *)(ip + 1);
|
|
|
|
|
}
|
|
|
|
|
bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
|
1994-05-24 10:09:53 +00:00
|
|
|
flags = TH_ACK;
|
|
|
|
|
} else {
|
Add code to allow the system to handle multiple routing tables.
This particular implementation is designed to be fully backwards compatible
and to be MFC-able to 7.x (and 6.x)
Currently the only protocol that can make use of the multiple tables is IPv4
Similar functionality exists in OpenBSD and Linux.
From my notes:
-----
One thing where FreeBSD has been falling behind, and which by chance I
have some time to work on is "policy based routing", which allows
different
packet streams to be routed by more than just the destination address.
Constraints:
------------
I want to make some form of this available in the 6.x tree
(and by extension 7.x) , but FreeBSD in general needs it so I might as
well do it in -current and back port the portions I need.
One of the ways that this can be done is to have the ability to
instantiate multiple kernel routing tables (which I will now
refer to as "Forwarding Information Bases" or "FIBs" for political
correctness reasons). Which FIB a particular packet uses to make
the next hop decision can be decided by a number of mechanisms.
The policies these mechanisms implement are the "Policies" referred
to in "Policy based routing".
One of the constraints I have if I try to back port this work to
6.x is that it must be implemented as a EXTENSION to the existing
ABIs in 6.x so that third party applications do not need to be
recompiled in timespan of the branch.
This first version will not have some of the bells and whistles that
will come with later versions. It will, for example, be limited to 16
tables in the first commit.
Implementation method, Compatible version. (part 1)
-------------------------------
For this reason I have implemented a "sufficient subset" of a
multiple routing table solution in Perforce, and back-ported it
to 6.x. (also in Perforce though not always caught up with what I
have done in -current/P4). The subset allows a number of FIBs
to be defined at compile time (8 is sufficient for my purposes in 6.x)
and implements the changes needed to allow IPV4 to use them. I have not
done the changes for ipv6 simply because I do not need it, and I do not
have enough knowledge of ipv6 (e.g. neighbor discovery) needed to do it.
Other protocol families are left untouched and should there be
users with proprietary protocol families, they should continue to work
and be oblivious to the existence of the extra FIBs.
To understand how this is done, one must know that the current FIB
code starts everything off with a single dimensional array of
pointers to FIB head structures (One per protocol family), each of
which in turn points to the trie of routes available to that family.
The basic change in the ABI compatible version of the change is to
extent that array to be a 2 dimensional array, so that
instead of protocol family X looking at rt_tables[X] for the
table it needs, it looks at rt_tables[Y][X] when for all
protocol families except ipv4 Y is always 0.
Code that is unaware of the change always just sees the first row
of the table, which of course looks just like the one dimensional
array that existed before.
The entry points rtrequest(), rtalloc(), rtalloc1(), rtalloc_ign()
are all maintained, but refer only to the first row of the array,
so that existing callers in proprietary protocols can continue to
do the "right thing".
Some new entry points are added, for the exclusive use of ipv4 code
called in_rtrequest(), in_rtalloc(), in_rtalloc1() and in_rtalloc_ign(),
which have an extra argument which refers the code to the correct row.
In addition, there are some new entry points (currently called
rtalloc_fib() and friends) that check the Address family being
looked up and call either rtalloc() (and friends) if the protocol
is not IPv4 forcing the action to row 0 or to the appropriate row
if it IS IPv4 (and that info is available). These are for calling
from code that is not specific to any particular protocol. The way
these are implemented would change in the non ABI preserving code
to be added later.
One feature of the first version of the code is that for ipv4,
the interface routes show up automatically on all the FIBs, so
that no matter what FIB you select you always have the basic
direct attached hosts available to you. (rtinit() does this
automatically).
You CAN delete an interface route from one FIB should you want
to but by default it's there. ARP information is also available
in each FIB. It's assumed that the same machine would have the
same MAC address, regardless of which FIB you are using to get
to it.
This brings us as to how the correct FIB is selected for an outgoing
IPV4 packet.
Firstly, all packets have a FIB associated with them. if nothing
has been done to change it, it will be FIB 0. The FIB is changed
in the following ways.
Packets fall into one of a number of classes.
1/ locally generated packets, coming from a socket/PCB.
Such packets select a FIB from a number associated with the
socket/PCB. This in turn is inherited from the process,
but can be changed by a socket option. The process in turn
inherits it on fork. I have written a utility call setfib
that acts a bit like nice..
setfib -3 ping target.example.com # will use fib 3 for ping.
It is an obvious extension to make it a property of a jail
but I have not done so. It can be achieved by combining the setfib and
jail commands.
2/ packets received on an interface for forwarding.
By default these packets would use table 0,
(or possibly a number settable in a sysctl(not yet)).
but prior to routing the firewall can inspect them (see below).
(possibly in the future you may be able to associate a FIB
with packets received on an interface.. An ifconfig arg, but not yet.)
3/ packets inspected by a packet classifier, which can arbitrarily
associate a fib with it on a packet by packet basis.
A fib assigned to a packet by a packet classifier
(such as ipfw) would over-ride a fib associated by
a more default source. (such as cases 1 or 2).
4/ a tcp listen socket associated with a fib will generate
accept sockets that are associated with that same fib.
5/ Packets generated in response to some other packet (e.g. reset
or icmp packets). These should use the FIB associated with the
packet being reponded to.
6/ Packets generated during encapsulation.
gif, tun and other tunnel interfaces will encapsulate using the FIB
that was in effect withthe proces that set up the tunnel.
thus setfib 1 ifconfig gif0 [tunnel instructions]
will set the fib for the tunnel to use to be fib 1.
Routing messages would be associated with their
process, and thus select one FIB or another.
messages from the kernel would be associated with the fib they
refer to and would only be received by a routing socket associated
with that fib. (not yet implemented)
In addition Netstat has been edited to be able to cope with the
fact that the array is now 2 dimensional. (It looks in system
memory using libkvm (!)). Old versions of netstat see only the first FIB.
In addition two sysctls are added to give:
a) the number of FIBs compiled in (active)
b) the default FIB of the calling process.
Early testing experience:
-------------------------
Basically our (IronPort's) appliance does this functionality already
using ipfw fwd but that method has some drawbacks.
For example,
It can't fully simulate a routing table because it can't influence the
socket's choice of local address when a connect() is done.
Testing during the generating of these changes has been
remarkably smooth so far. Multiple tables have co-existed
with no notable side effects, and packets have been routes
accordingly.
ipfw has grown 2 new keywords:
setfib N ip from anay to any
count ip from any to any fib N
In pf there seems to be a requirement to be able to give symbolic names to the
fibs but I do not have that capacity. I am not sure if it is required.
SCTP has interestingly enough built in support for this, called VRFs
in Cisco parlance. it will be interesting to see how that handles it
when it suddenly actually does something.
Where to next:
--------------------
After committing the ABI compatible version and MFCing it, I'd
like to proceed in a forward direction in -current. this will
result in some roto-tilling in the routing code.
Firstly: the current code's idea of having a separate tree per
protocol family, all of the same format, and pointed to by the
1 dimensional array is a bit silly. Especially when one considers that
there is code that makes assumptions about every protocol having the
same internal structures there. Some protocols don't WANT that
sort of structure. (for example the whole idea of a netmask is foreign
to appletalk). This needs to be made opaque to the external code.
My suggested first change is to add routing method pointers to the
'domain' structure, along with information pointing the data.
instead of having an array of pointers to uniform structures,
there would be an array pointing to the 'domain' structures
for each protocol address domain (protocol family),
and the methods this reached would be called. The methods would have
an argument that gives FIB number, but the protocol would be free
to ignore it.
When the ABI can be changed it raises the possibilty of the
addition of a fib entry into the "struct route". Currently,
the structure contains the sockaddr of the desination, and the resulting
fib entry. To make this work fully, one could add a fib number
so that given an address and a fib, one can find the third element, the
fib entry.
Interaction with the ARP layer/ LL layer would need to be
revisited as well. Qing Li has been working on this already.
This work was sponsored by Ironport Systems/Cisco
Reviewed by: several including rwatson, bz and mlair (parts each)
Obtained from: Ironport systems/Cisco
2008-05-09 23:03:00 +00:00
|
|
|
/*
|
|
|
|
|
* reuse the mbuf.
|
|
|
|
|
* XXX MRT We inherrit the FIB, which is lucky.
|
|
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
m_freem(m->m_next);
|
2004-04-05 00:49:07 +00:00
|
|
|
m->m_next = NULL;
|
2000-01-09 19:17:30 +00:00
|
|
|
m->m_data = (caddr_t)ipgen;
|
2011-07-04 17:43:04 +00:00
|
|
|
m_addr_changed(m);
|
2000-01-09 19:17:30 +00:00
|
|
|
/* m_len is set later */
|
1994-05-24 10:09:53 +00:00
|
|
|
tlen = 0;
|
|
|
|
|
#define xchg(a,b,type) { type t; t=a; a=b; b=t; }
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (isipv6) {
|
|
|
|
|
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
|
|
|
|
|
nth = (struct tcphdr *)(ip6 + 1);
|
|
|
|
|
} else
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
{
|
2009-02-13 15:14:43 +00:00
|
|
|
xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t);
|
2000-01-09 19:17:30 +00:00
|
|
|
nth = (struct tcphdr *)(ip + 1);
|
|
|
|
|
}
|
|
|
|
|
if (th != nth) {
|
|
|
|
|
/*
|
|
|
|
|
* this is usually a case when an extension header
|
|
|
|
|
* exists between the IPv6 header and the
|
|
|
|
|
* TCP header.
|
|
|
|
|
*/
|
|
|
|
|
nth->th_sport = th->th_sport;
|
|
|
|
|
nth->th_dport = th->th_dport;
|
|
|
|
|
}
|
2009-02-13 15:14:43 +00:00
|
|
|
xchg(nth->th_dport, nth->th_sport, uint16_t);
|
1994-05-24 10:09:53 +00:00
|
|
|
#undef xchg
|
|
|
|
|
}
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (isipv6) {
|
2002-02-04 17:37:06 +00:00
|
|
|
ip6->ip6_flow = 0;
|
|
|
|
|
ip6->ip6_vfc = IPV6_VERSION;
|
|
|
|
|
ip6->ip6_nxt = IPPROTO_TCP;
|
2000-01-09 19:17:30 +00:00
|
|
|
ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) +
|
|
|
|
|
tlen));
|
|
|
|
|
tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
|
2011-04-30 11:21:29 +00:00
|
|
|
}
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
#if defined(INET) && defined(INET6)
|
|
|
|
|
else
|
|
|
|
|
#endif
|
|
|
|
|
#ifdef INET
|
2004-08-16 18:32:07 +00:00
|
|
|
{
|
|
|
|
|
tlen += sizeof (struct tcpiphdr);
|
|
|
|
|
ip->ip_len = tlen;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
ip->ip_ttl = V_ip_defttl;
|
|
|
|
|
if (V_path_mtu_discovery)
|
2004-08-16 18:32:07 +00:00
|
|
|
ip->ip_off |= IP_DF;
|
|
|
|
|
}
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
m->m_len = tlen;
|
|
|
|
|
m->m_pkthdr.len = tlen;
|
2004-04-05 00:49:07 +00:00
|
|
|
m->m_pkthdr.rcvif = NULL;
|
2002-07-31 19:06:49 +00:00
|
|
|
#ifdef MAC
|
2003-11-08 22:59:22 +00:00
|
|
|
if (inp != NULL) {
|
2002-07-31 19:06:49 +00:00
|
|
|
/*
|
|
|
|
|
* Packet is associated with a socket, so allow the
|
|
|
|
|
* label of the response to reflect the socket label.
|
|
|
|
|
*/
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2007-10-24 19:04:04 +00:00
|
|
|
mac_inpcb_create_mbuf(inp, m);
|
2002-07-31 19:06:49 +00:00
|
|
|
} else {
|
|
|
|
|
/*
|
2003-08-21 18:39:16 +00:00
|
|
|
* Packet is not associated with a socket, so possibly
|
|
|
|
|
* update the label in place.
|
2002-07-31 19:06:49 +00:00
|
|
|
*/
|
2007-10-24 19:04:04 +00:00
|
|
|
mac_netinet_tcp_reply(m);
|
2002-07-31 19:06:49 +00:00
|
|
|
}
|
|
|
|
|
#endif
|
2000-01-09 19:17:30 +00:00
|
|
|
nth->th_seq = htonl(seq);
|
|
|
|
|
nth->th_ack = htonl(ack);
|
|
|
|
|
nth->th_x2 = 0;
|
|
|
|
|
nth->th_off = sizeof (struct tcphdr) >> 2;
|
|
|
|
|
nth->th_flags = flags;
|
2004-04-05 00:49:07 +00:00
|
|
|
if (tp != NULL)
|
2000-01-09 19:17:30 +00:00
|
|
|
nth->th_win = htons((u_short) (win >> tp->rcv_scale));
|
1994-05-24 10:09:53 +00:00
|
|
|
else
|
2000-01-09 19:17:30 +00:00
|
|
|
nth->th_win = htons((u_short)win);
|
|
|
|
|
nth->th_urp = 0;
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
if (isipv6) {
|
2000-04-19 15:05:00 +00:00
|
|
|
nth->th_sum = 0;
|
2000-01-09 19:17:30 +00:00
|
|
|
nth->th_sum = in6_cksum(m, IPPROTO_TCP,
|
|
|
|
|
sizeof(struct ip6_hdr),
|
|
|
|
|
tlen - sizeof(struct ip6_hdr));
|
2004-04-05 00:49:07 +00:00
|
|
|
ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb :
|
|
|
|
|
NULL, NULL);
|
2011-04-30 11:21:29 +00:00
|
|
|
}
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif /* INET6 */
|
2011-04-30 11:21:29 +00:00
|
|
|
#if defined(INET6) && defined(INET)
|
|
|
|
|
else
|
|
|
|
|
#endif
|
|
|
|
|
#ifdef INET
|
2004-08-16 18:32:07 +00:00
|
|
|
{
|
|
|
|
|
nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
|
|
|
|
|
htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p)));
|
|
|
|
|
m->m_pkthdr.csum_flags = CSUM_TCP;
|
|
|
|
|
m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
|
|
|
|
|
}
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
1995-02-09 23:13:27 +00:00
|
|
|
#ifdef TCPDEBUG
|
2003-11-08 22:59:22 +00:00
|
|
|
if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG))
|
2000-01-09 19:17:30 +00:00
|
|
|
tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0);
|
|
|
|
|
#endif
|
|
|
|
|
#ifdef INET6
|
2003-11-20 20:07:39 +00:00
|
|
|
if (isipv6)
|
|
|
|
|
(void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp);
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif /* INET6 */
|
2011-04-30 11:21:29 +00:00
|
|
|
#if defined(INET) && defined(INET6)
|
|
|
|
|
else
|
|
|
|
|
#endif
|
|
|
|
|
#ifdef INET
|
|
|
|
|
(void) ip_output(m, NULL, NULL, ipflags, NULL, inp);
|
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Create a new TCP control block, making an
|
|
|
|
|
* empty reassembly queue and hooking it to the argument
|
1998-03-24 18:06:34 +00:00
|
|
|
* protocol control block. The `inp' parameter must have
|
|
|
|
|
* come from the zone allocator set up in tcp_init().
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
struct tcpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_newtcpcb(struct inpcb *inp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-02-19 22:32:43 +00:00
|
|
|
struct tcpcb_mem *tm;
|
|
|
|
|
struct tcpcb *tp;
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
|
|
|
|
|
#endif /* INET6 */
|
1994-05-24 10:09:53 +00:00
|
|
|
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
tm = uma_zalloc(V_tcpcb_zone, M_NOWAIT | M_ZERO);
|
2003-02-19 22:32:43 +00:00
|
|
|
if (tm == NULL)
|
|
|
|
|
return (NULL);
|
|
|
|
|
tp = &tm->tcb;
|
2010-11-12 06:41:55 +00:00
|
|
|
|
|
|
|
|
/* Initialise cc_var struct for this tcpcb. */
|
|
|
|
|
tp->ccv = &tm->ccv;
|
|
|
|
|
tp->ccv->type = IPPROTO_TCP;
|
|
|
|
|
tp->ccv->ccvc.tcp = tp;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Use the current system default CC algorithm.
|
|
|
|
|
*/
|
|
|
|
|
CC_LIST_RLOCK();
|
|
|
|
|
KASSERT(!STAILQ_EMPTY(&cc_list), ("cc_list is empty!"));
|
|
|
|
|
CC_ALGO(tp) = CC_DEFAULT();
|
|
|
|
|
CC_LIST_RUNLOCK();
|
|
|
|
|
|
|
|
|
|
if (CC_ALGO(tp)->cb_init != NULL)
|
|
|
|
|
if (CC_ALGO(tp)->cb_init(tp->ccv) > 0) {
|
|
|
|
|
uma_zfree(V_tcpcb_zone, tm);
|
|
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
|
2010-12-28 12:13:30 +00:00
|
|
|
tp->osd = &tm->osd;
|
|
|
|
|
if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) {
|
|
|
|
|
uma_zfree(V_tcpcb_zone, tm);
|
|
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
|
Permit buiding kernels with options VIMAGE, restricted to only a single
active network stack instance. Turning on options VIMAGE at compile
time yields the following changes relative to default kernel build:
1) V_ accessor macros for virtualized variables resolve to structure
fields via base pointers, instead of being resolved as fields in global
structs or plain global variables. As an example, V_ifnet becomes:
options VIMAGE: ((struct vnet_net *) vnet_net)->_ifnet
default build: vnet_net_0._ifnet
options VIMAGE_GLOBALS: ifnet
2) INIT_VNET_* macros will declare and set up base pointers to be used
by V_ accessor macros, instead of resolving to whitespace:
INIT_VNET_NET(ifp->if_vnet); becomes
struct vnet_net *vnet_net = (ifp->if_vnet)->mod_data[VNET_MOD_NET];
3) Memory for vnet modules registered via vnet_mod_register() is now
allocated at run time in sys/kern/kern_vimage.c, instead of per vnet
module structs being declared as globals. If required, vnet modules
can now request the framework to provide them with allocated bzeroed
memory by filling in the vmi_size field in their vmi_modinfo structures.
4) structs socket, ifnet, inpcbinfo, tcpcb and syncache_head are
extended to hold a pointer to the parent vnet. options VIMAGE builds
will fill in those fields as required.
5) curvnet is introduced as a new global variable in options VIMAGE
builds, always pointing to the default and only struct vnet.
6) struct sysctl_oid has been extended with additional two fields to
store major and minor virtualization module identifiers, oid_v_subs and
oid_v_mod. SYSCTL_V_* family of macros will fill in those fields
accordingly, and store the offset in the appropriate vnet container
struct in oid_arg1.
In sysctl handlers dealing with virtualized sysctls, the
SYSCTL_RESOLVE_V_ARG1() macro will compute the address of the target
variable and make it available in arg1 variable for further processing.
Unused fields in structs vnet_inet, vnet_inet6 and vnet_ipfw have
been deleted.
Reviewed by: bz, rwatson
Approved by: julian (mentor)
2009-04-30 13:36:26 +00:00
|
|
|
#ifdef VIMAGE
|
|
|
|
|
tp->t_vnet = inp->inp_vnet;
|
|
|
|
|
#endif
|
2007-09-24 05:26:24 +00:00
|
|
|
tp->t_timers = &tm->tt;
|
2003-02-19 22:32:43 +00:00
|
|
|
/* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */
|
2000-01-09 19:17:30 +00:00
|
|
|
tp->t_maxseg = tp->t_maxopd =
|
|
|
|
|
#ifdef INET6
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
isipv6 ? V_tcp_v6mssdflt :
|
2000-01-09 19:17:30 +00:00
|
|
|
#endif /* INET6 */
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
V_tcp_mssdflt;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1999-08-30 21:17:07 +00:00
|
|
|
/* Set up our timeouts. */
|
2007-09-24 05:26:24 +00:00
|
|
|
callout_init(&tp->t_timers->tt_rexmt, CALLOUT_MPSAFE);
|
|
|
|
|
callout_init(&tp->t_timers->tt_persist, CALLOUT_MPSAFE);
|
|
|
|
|
callout_init(&tp->t_timers->tt_keep, CALLOUT_MPSAFE);
|
|
|
|
|
callout_init(&tp->t_timers->tt_2msl, CALLOUT_MPSAFE);
|
|
|
|
|
callout_init(&tp->t_timers->tt_delack, CALLOUT_MPSAFE);
|
1999-08-30 21:17:07 +00:00
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if (V_tcp_do_rfc1323)
|
1995-02-09 23:13:27 +00:00
|
|
|
tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if (V_tcp_do_sack)
|
2007-05-06 15:56:31 +00:00
|
|
|
tp->t_flags |= TF_SACK_PERMIT;
|
2005-07-01 22:54:18 +00:00
|
|
|
TAILQ_INIT(&tp->snd_holes);
|
1998-03-24 18:06:34 +00:00
|
|
|
tp->t_inpcb = inp; /* XXX */
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
|
* Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
|
1996-06-14 17:17:32 +00:00
|
|
|
* rtt estimate. Set rttvar so that srtt + 4 * rttvar gives
|
1994-05-24 10:09:53 +00:00
|
|
|
* reasonable initial retransmit time.
|
|
|
|
|
*/
|
|
|
|
|
tp->t_srtt = TCPTV_SRTTBASE;
|
1996-06-14 17:17:32 +00:00
|
|
|
tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4;
|
2002-07-18 19:06:12 +00:00
|
|
|
tp->t_rttmin = tcp_rexmit_min;
|
1996-06-14 17:17:32 +00:00
|
|
|
tp->t_rxtcur = TCPTV_RTOBASE;
|
1994-05-24 10:09:53 +00:00
|
|
|
tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
|
|
|
|
|
tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
|
1999-08-30 21:17:07 +00:00
|
|
|
tp->t_rcvtime = ticks;
|
2004-08-16 18:32:07 +00:00
|
|
|
/*
|
2000-01-25 01:05:18 +00:00
|
|
|
* IPv4 TTL initialization is necessary for an IPv6 socket as well,
|
|
|
|
|
* because the socket may be bound to an IPv6 wildcard address,
|
|
|
|
|
* which may match an IPv4-mapped IPv6 address.
|
|
|
|
|
*/
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
inp->inp_ip_ttl = V_ip_defttl;
|
2006-04-03 13:33:55 +00:00
|
|
|
inp->inp_ppcb = tp;
|
1998-03-24 18:06:34 +00:00
|
|
|
return (tp); /* XXX */
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
2010-11-16 08:30:39 +00:00
|
|
|
/*
|
|
|
|
|
* Switch the congestion control algorithm back to NewReno for any active
|
|
|
|
|
* control blocks using an algorithm which is about to go away.
|
|
|
|
|
* This ensures the CC framework can allow the unload to proceed without leaving
|
|
|
|
|
* any dangling pointers which would trigger a panic.
|
|
|
|
|
* Returning non-zero would inform the CC framework that something went wrong
|
|
|
|
|
* and it would be unsafe to allow the unload to proceed. However, there is no
|
|
|
|
|
* way for this to occur with this implementation so we always return zero.
|
|
|
|
|
*/
|
|
|
|
|
int
|
|
|
|
|
tcp_ccalgounload(struct cc_algo *unload_algo)
|
|
|
|
|
{
|
|
|
|
|
struct cc_algo *tmpalgo;
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
struct tcpcb *tp;
|
|
|
|
|
VNET_ITERATOR_DECL(vnet_iter);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Check all active control blocks across all network stacks and change
|
|
|
|
|
* any that are using "unload_algo" back to NewReno. If "unload_algo"
|
|
|
|
|
* requires cleanup code to be run, call it.
|
|
|
|
|
*/
|
|
|
|
|
VNET_LIST_RLOCK();
|
|
|
|
|
VNET_FOREACH(vnet_iter) {
|
|
|
|
|
CURVNET_SET(vnet_iter);
|
|
|
|
|
INP_INFO_RLOCK(&V_tcbinfo);
|
|
|
|
|
/*
|
|
|
|
|
* New connections already part way through being initialised
|
|
|
|
|
* with the CC algo we're removing will not race with this code
|
|
|
|
|
* because the INP_INFO_WLOCK is held during initialisation. We
|
|
|
|
|
* therefore don't enter the loop below until the connection
|
|
|
|
|
* list has stabilised.
|
|
|
|
|
*/
|
|
|
|
|
LIST_FOREACH(inp, &V_tcb, inp_list) {
|
|
|
|
|
INP_WLOCK(inp);
|
|
|
|
|
/* Important to skip tcptw structs. */
|
|
|
|
|
if (!(inp->inp_flags & INP_TIMEWAIT) &&
|
|
|
|
|
(tp = intotcpcb(inp)) != NULL) {
|
|
|
|
|
/*
|
|
|
|
|
* By holding INP_WLOCK here, we are assured
|
|
|
|
|
* that the connection is not currently
|
|
|
|
|
* executing inside the CC module's functions
|
|
|
|
|
* i.e. it is safe to make the switch back to
|
|
|
|
|
* NewReno.
|
|
|
|
|
*/
|
|
|
|
|
if (CC_ALGO(tp) == unload_algo) {
|
|
|
|
|
tmpalgo = CC_ALGO(tp);
|
|
|
|
|
/* NewReno does not require any init. */
|
|
|
|
|
CC_ALGO(tp) = &newreno_cc_algo;
|
|
|
|
|
if (tmpalgo->cb_destroy != NULL)
|
|
|
|
|
tmpalgo->cb_destroy(tp->ccv);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
INP_WUNLOCK(inp);
|
|
|
|
|
}
|
|
|
|
|
INP_INFO_RUNLOCK(&V_tcbinfo);
|
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
|
}
|
|
|
|
|
VNET_LIST_RUNLOCK();
|
|
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
|
* Drop a TCP connection, reporting
|
|
|
|
|
* the specified error. If connection is synchronized,
|
|
|
|
|
* then send a RST to peer.
|
|
|
|
|
*/
|
|
|
|
|
struct tcpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_drop(struct tcpcb *tp, int errno)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
|
|
|
|
struct socket *so = tp->t_inpcb->inp_socket;
|
|
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(tp->t_inpcb);
|
2005-06-01 12:06:07 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
if (TCPS_HAVERCVDSYN(tp->t_state)) {
|
|
|
|
|
tp->t_state = TCPS_CLOSED;
|
2007-12-18 22:59:07 +00:00
|
|
|
(void) tcp_output_reset(tp);
|
2009-04-11 22:07:19 +00:00
|
|
|
TCPSTAT_INC(tcps_drops);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else
|
2009-04-11 22:07:19 +00:00
|
|
|
TCPSTAT_INC(tcps_conndrops);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (errno == ETIMEDOUT && tp->t_softerror)
|
|
|
|
|
errno = tp->t_softerror;
|
|
|
|
|
so->so_error = errno;
|
|
|
|
|
return (tcp_close(tp));
|
|
|
|
|
}
|
|
|
|
|
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_discardcb(struct tcpcb *tp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
|
|
|
|
struct inpcb *inp = tp->t_inpcb;
|
|
|
|
|
struct socket *so = inp->inp_socket;
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
|
|
|
|
|
#endif /* INET6 */
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2004-12-05 22:27:53 +00:00
|
|
|
|
1999-08-30 21:17:07 +00:00
|
|
|
/*
|
2010-03-07 14:13:59 +00:00
|
|
|
* Make sure that all of our timers are stopped before we delete the
|
|
|
|
|
* PCB.
|
|
|
|
|
*
|
|
|
|
|
* XXXRW: Really, we would like to use callout_drain() here in order
|
|
|
|
|
* to avoid races experienced in tcp_timer.c where a timer is already
|
|
|
|
|
* executing at this point. However, we can't, both because we're
|
|
|
|
|
* running in a context where we can't sleep, and also because we
|
|
|
|
|
* hold locks required by the timers. What we instead need to do is
|
|
|
|
|
* test to see if callout_drain() is required, and if so, defer some
|
|
|
|
|
* portion of the remainder of tcp_discardcb() to an asynchronous
|
|
|
|
|
* context that can callout_drain() and then continue. Some care
|
|
|
|
|
* will be required to ensure that no further processing takes place
|
|
|
|
|
* on the tcpcb, even though it hasn't been freed (a flag?).
|
1999-08-30 21:17:07 +00:00
|
|
|
*/
|
2007-09-24 05:26:24 +00:00
|
|
|
callout_stop(&tp->t_timers->tt_rexmt);
|
|
|
|
|
callout_stop(&tp->t_timers->tt_persist);
|
|
|
|
|
callout_stop(&tp->t_timers->tt_keep);
|
|
|
|
|
callout_stop(&tp->t_timers->tt_2msl);
|
|
|
|
|
callout_stop(&tp->t_timers->tt_delack);
|
1999-08-30 21:17:07 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
1995-06-29 18:11:24 +00:00
|
|
|
* If we got enough samples through the srtt filter,
|
|
|
|
|
* save the rtt and rttvar in the routing entry.
|
2003-11-20 20:07:39 +00:00
|
|
|
* 'Enough' is arbitrarily defined as 4 rtt samples.
|
|
|
|
|
* 4 samples is enough for the srtt filter to converge
|
|
|
|
|
* to within enough % of the correct value; fewer samples
|
|
|
|
|
* and we could save a bogus rtt. The danger is not high
|
|
|
|
|
* as tcp quickly recovers from everything.
|
|
|
|
|
* XXX: Works very well but needs some more statistics!
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2003-11-20 20:07:39 +00:00
|
|
|
if (tp->t_rttupdated >= 4) {
|
|
|
|
|
struct hc_metrics_lite metrics;
|
|
|
|
|
u_long ssthresh;
|
2000-01-09 19:17:30 +00:00
|
|
|
|
2003-11-20 20:07:39 +00:00
|
|
|
bzero(&metrics, sizeof(metrics));
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
2003-11-20 20:07:39 +00:00
|
|
|
* Update the ssthresh always when the conditions below
|
|
|
|
|
* are satisfied. This gives us better new start value
|
|
|
|
|
* for the congestion avoidance for new connections.
|
|
|
|
|
* ssthresh is only set if packet loss occured on a session.
|
2006-08-02 16:18:05 +00:00
|
|
|
*
|
|
|
|
|
* XXXRW: 'so' may be NULL here, and/or socket buffer may be
|
|
|
|
|
* being torn down. Ideally this code would not use 'so'.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2003-11-20 20:07:39 +00:00
|
|
|
ssthresh = tp->snd_ssthresh;
|
|
|
|
|
if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) {
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
|
* convert the limit from user data bytes to
|
|
|
|
|
* packets then to packet data bytes.
|
|
|
|
|
*/
|
2003-11-20 20:07:39 +00:00
|
|
|
ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg;
|
|
|
|
|
if (ssthresh < 2)
|
|
|
|
|
ssthresh = 2;
|
|
|
|
|
ssthresh *= (u_long)(tp->t_maxseg +
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
(isipv6 ? sizeof (struct ip6_hdr) +
|
|
|
|
|
sizeof (struct tcphdr) :
|
|
|
|
|
#endif
|
|
|
|
|
sizeof (struct tcpiphdr)
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
)
|
|
|
|
|
#endif
|
|
|
|
|
);
|
2003-11-20 20:07:39 +00:00
|
|
|
} else
|
|
|
|
|
ssthresh = 0;
|
|
|
|
|
metrics.rmx_ssthresh = ssthresh;
|
|
|
|
|
|
|
|
|
|
metrics.rmx_rtt = tp->t_srtt;
|
|
|
|
|
metrics.rmx_rttvar = tp->t_rttvar;
|
|
|
|
|
metrics.rmx_cwnd = tp->snd_cwnd;
|
2004-08-16 18:32:07 +00:00
|
|
|
metrics.rmx_sendpipe = 0;
|
2003-11-20 20:07:39 +00:00
|
|
|
metrics.rmx_recvpipe = 0;
|
|
|
|
|
|
|
|
|
|
tcp_hc_update(&inp->inp_inc, &metrics);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2003-11-20 20:07:39 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/* free the reassembly queue, if any */
|
2010-09-25 04:58:46 +00:00
|
|
|
tcp_reass_flush(tp);
|
2007-12-18 22:59:07 +00:00
|
|
|
/* Disconnect offload device, if any. */
|
|
|
|
|
tcp_offload_detach(tp);
|
|
|
|
|
|
2004-06-23 21:04:37 +00:00
|
|
|
tcp_free_sackholes(tp);
|
2010-11-12 06:41:55 +00:00
|
|
|
|
|
|
|
|
/* Allow the CC algorithm to clean up after itself. */
|
|
|
|
|
if (CC_ALGO(tp)->cb_destroy != NULL)
|
|
|
|
|
CC_ALGO(tp)->cb_destroy(tp->ccv);
|
|
|
|
|
|
2010-12-28 12:13:30 +00:00
|
|
|
khelp_destroy_osd(tp->osd);
|
|
|
|
|
|
2010-11-12 06:41:55 +00:00
|
|
|
CC_ALGO(tp) = NULL;
|
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
|
|
|
inp->inp_ppcb = NULL;
|
2002-12-24 21:00:31 +00:00
|
|
|
tp->t_inpcb = NULL;
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
uma_zfree(V_tcpcb_zone, tp);
|
2003-02-19 22:32:43 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
2006-04-03 12:52:13 +00:00
|
|
|
* Attempt to close a TCP control block, marking it as dropped, and freeing
|
|
|
|
|
* the socket if we hold the only reference.
|
2003-02-19 22:32:43 +00:00
|
|
|
*/
|
|
|
|
|
struct tcpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_close(struct tcpcb *tp)
|
2003-02-19 22:32:43 +00:00
|
|
|
{
|
|
|
|
|
struct inpcb *inp = tp->t_inpcb;
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
struct socket *so;
|
2003-02-19 22:32:43 +00:00
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2004-12-05 22:27:53 +00:00
|
|
|
|
2007-12-18 22:59:07 +00:00
|
|
|
/* Notify any offload devices of listener close */
|
|
|
|
|
if (tp->t_state == TCPS_LISTEN)
|
|
|
|
|
tcp_offload_listen_close(tp);
|
2006-04-25 11:17:35 +00:00
|
|
|
in_pcbdrop(inp);
|
2009-04-11 22:07:19 +00:00
|
|
|
TCPSTAT_INC(tcps_closed);
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL"));
|
|
|
|
|
so = inp->inp_socket;
|
|
|
|
|
soisdisconnected(so);
|
2009-03-15 09:58:31 +00:00
|
|
|
if (inp->inp_flags & INP_SOCKREF) {
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
KASSERT(so->so_state & SS_PROTOREF,
|
|
|
|
|
("tcp_close: !SS_PROTOREF"));
|
2009-03-15 09:58:31 +00:00
|
|
|
inp->inp_flags &= ~INP_SOCKREF;
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
ACCEPT_LOCK();
|
|
|
|
|
SOCK_LOCK(so);
|
|
|
|
|
so->so_state &= ~SS_PROTOREF;
|
|
|
|
|
sofree(so);
|
|
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
return (tp);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_drain(void)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
VNET_ITERATOR_DECL(vnet_iter);
|
2006-04-03 12:59:27 +00:00
|
|
|
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
if (!do_tcpdrain)
|
|
|
|
|
return;
|
|
|
|
|
|
2009-07-19 14:20:53 +00:00
|
|
|
VNET_LIST_RLOCK_NOSLEEP();
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
VNET_FOREACH(vnet_iter) {
|
|
|
|
|
CURVNET_SET(vnet_iter);
|
1999-12-28 23:18:33 +00:00
|
|
|
struct inpcb *inpb;
|
|
|
|
|
struct tcpcb *tcpb;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Walk the tcpbs, if existing, and flush the reassembly queue,
|
|
|
|
|
* if there is one...
|
|
|
|
|
* XXX: The "Net/3" implementation doesn't imply that the TCP
|
|
|
|
|
* reassembly queue should be flushed, but in a situation
|
2004-08-16 18:32:07 +00:00
|
|
|
* where we're really low on mbufs, this is potentially
|
|
|
|
|
* usefull.
|
1999-12-28 23:18:33 +00:00
|
|
|
*/
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RLOCK(&V_tcbinfo);
|
|
|
|
|
LIST_FOREACH(inpb, V_tcbinfo.ipi_listhead, inp_list) {
|
2009-03-15 09:58:31 +00:00
|
|
|
if (inpb->inp_flags & INP_TIMEWAIT)
|
2003-02-19 22:32:43 +00:00
|
|
|
continue;
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inpb);
|
2004-04-05 00:49:07 +00:00
|
|
|
if ((tcpb = intotcpcb(inpb)) != NULL) {
|
2010-09-25 04:58:46 +00:00
|
|
|
tcp_reass_flush(tcpb);
|
2005-04-10 05:21:29 +00:00
|
|
|
tcp_clean_sackreport(tcpb);
|
1999-12-28 23:18:33 +00:00
|
|
|
}
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inpb);
|
1999-12-28 23:18:33 +00:00
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RUNLOCK(&V_tcbinfo);
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
CURVNET_RESTORE();
|
1999-12-28 23:18:33 +00:00
|
|
|
}
|
2009-07-19 14:20:53 +00:00
|
|
|
VNET_LIST_RUNLOCK_NOSLEEP();
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Notify a tcp user of an asynchronous error;
|
|
|
|
|
* store error as soft error, but wake up user
|
|
|
|
|
* (for now, won't do anything until can select for soft error).
|
2001-02-23 21:07:06 +00:00
|
|
|
*
|
|
|
|
|
* Do not wake up user since there currently is no mechanism for
|
|
|
|
|
* reporting soft errors (yet - a kqueue filter may be added).
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2002-06-14 08:35:21 +00:00
|
|
|
static struct inpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_notify(struct inpcb *inp, int error)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-04-03 13:33:55 +00:00
|
|
|
struct tcpcb *tp;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2006-04-03 14:07:50 +00:00
|
|
|
|
2009-03-15 09:58:31 +00:00
|
|
|
if ((inp->inp_flags & INP_TIMEWAIT) ||
|
|
|
|
|
(inp->inp_flags & INP_DROPPED))
|
2006-04-03 14:07:50 +00:00
|
|
|
return (inp);
|
|
|
|
|
|
2006-04-03 13:33:55 +00:00
|
|
|
tp = intotcpcb(inp);
|
2006-04-03 14:07:50 +00:00
|
|
|
KASSERT(tp != NULL, ("tcp_notify: tp == NULL"));
|
2004-12-05 22:27:53 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
|
* Ignore some errors if we are hooked up.
|
|
|
|
|
* If connection hasn't completed, has retransmitted several times,
|
|
|
|
|
* and receives a second error, give up now. This is better
|
|
|
|
|
* than waiting a long time to establish a connection that
|
|
|
|
|
* can never complete.
|
|
|
|
|
*/
|
|
|
|
|
if (tp->t_state == TCPS_ESTABLISHED &&
|
2003-02-15 02:37:57 +00:00
|
|
|
(error == EHOSTUNREACH || error == ENETUNREACH ||
|
|
|
|
|
error == EHOSTDOWN)) {
|
2004-12-23 01:34:26 +00:00
|
|
|
return (inp);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 &&
|
2002-06-14 08:35:21 +00:00
|
|
|
tp->t_softerror) {
|
Update TCP for infrastructural changes to the socket/pcb refcount model,
pru_abort(), pru_detach(), and in_pcbdetach():
- Universally support and enforce the invariant that so_pcb is
never NULL, converting dozens of unnecessary NULL checks into
assertions, and eliminating dozens of unnecessary error handling
cases in protocol code.
- In some cases, eliminate unnecessary pcbinfo locking, as it is no
longer required to ensure so_pcb != NULL. For example, the receive
code no longer requires the pcbinfo lock, and the send code only
requires it if building a new connection on an otherwise unconnected
socket triggered via sendto() with an address. This should
significnatly reduce tcbinfo lock contention in the receive and send
cases.
- In order to support the invariant that so_pcb != NULL, it is now
necessary for the TCP code to not discard the tcpcb any time a
connection is dropped, but instead leave the tcpcb until the socket
is shutdown. This case is handled by setting INP_DROPPED, to
substitute for using a NULL so_pcb to indicate that the connection
has been dropped. This requires the inpcb lock, but not the pcbinfo
lock.
- Unlike all other protocols in the tree, TCP may need to retain access
to the socket after the file descriptor has been closed. Set
SS_PROTOREF in tcp_detach() in order to prevent the socket from being
freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether
or not it needs to free the socket when the connection finally does
close. The typical case where this occurs is if close() is called on
a TCP socket before all sent data in the send socket buffer has been
transmitted or acknowledged. If INP_SOCKREF is found when the
connection is dropped, we release the inpcb, tcpcb, and socket instead
of flagging INP_DROPPED.
- Abort and detach protocol switch methods no longer return failures,
nor attempt to free sockets, as the socket layer does this.
- Annotate the existence of a long-standing race in the TCP timer code,
in which timers are stopped but not drained when the socket is freed,
as waiting for drain may lead to deadlocks, or have to occur in a
context where waiting is not permitted. This race has been handled
by testing to see if the tcpcb pointer in the inpcb is NULL (and vice
versa), which is not normally permitted, but may be true of a inpcb
and tcpcb have been freed. Add a counter to test how often this race
has actually occurred, and a large comment for each instance where
we compare potentially freed memory with NULL. This will have to be
fixed in the near future, but requires is to further address how to
handle the timer shutdown shutdown issue.
- Several TCP calls no longer potentially free the passed inpcb/tcpcb,
so no longer need to return a pointer to indicate whether the argument
passed in is still valid.
- Un-macroize debugging and locking setup for various protocol switch
methods for TCP, as it lead to more obscurity, and as locking becomes
more customized to the methods, offers less benefit.
- Assert copyright on tcp_usrreq.c due to significant modifications that
have been made as part of this work.
These changes significantly modify the memory management and connection
logic of our TCP implementation, and are (as such) High Risk Changes,
and likely to contain serious bugs. Please report problems to the
current@ mailing list ASAP, ideally with simple test cases, and
optionally, packet traces.
MFC after: 3 months
2006-04-01 16:36:36 +00:00
|
|
|
tp = tcp_drop(tp, error);
|
|
|
|
|
if (tp != NULL)
|
|
|
|
|
return (inp);
|
|
|
|
|
else
|
|
|
|
|
return (NULL);
|
2002-06-14 08:35:21 +00:00
|
|
|
} else {
|
1994-05-24 10:09:53 +00:00
|
|
|
tp->t_softerror = error;
|
2004-12-23 01:34:26 +00:00
|
|
|
return (inp);
|
2002-06-14 08:35:21 +00:00
|
|
|
}
|
2001-02-23 21:07:06 +00:00
|
|
|
#if 0
|
2003-03-02 16:54:40 +00:00
|
|
|
wakeup( &so->so_timeo);
|
1994-05-24 10:09:53 +00:00
|
|
|
sorwakeup(so);
|
|
|
|
|
sowwakeup(so);
|
2001-02-23 21:07:06 +00:00
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
1998-05-15 20:11:40 +00:00
|
|
|
static int
|
2000-07-04 11:25:35 +00:00
|
|
|
tcp_pcblist(SYSCTL_HANDLER_ARGS)
|
1998-05-15 20:11:40 +00:00
|
|
|
{
|
2007-07-27 00:57:06 +00:00
|
|
|
int error, i, m, n, pcb_count;
|
1998-05-15 20:11:40 +00:00
|
|
|
struct inpcb *inp, **inp_list;
|
|
|
|
|
inp_gen_t gencnt;
|
|
|
|
|
struct xinpgen xig;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* The process of preparing the TCB list is too time-consuming and
|
|
|
|
|
* resource-intensive to repeat twice on every request.
|
|
|
|
|
*/
|
2004-04-05 00:49:07 +00:00
|
|
|
if (req->oldptr == NULL) {
|
2010-08-27 18:17:46 +00:00
|
|
|
n = V_tcbinfo.ipi_count + syncache_pcbcount();
|
|
|
|
|
n += imax(n / 8, 10);
|
|
|
|
|
req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xtcpcb);
|
2004-12-23 01:34:26 +00:00
|
|
|
return (0);
|
1998-05-15 20:11:40 +00:00
|
|
|
}
|
|
|
|
|
|
2004-04-05 00:49:07 +00:00
|
|
|
if (req->newptr != NULL)
|
2004-12-23 01:34:26 +00:00
|
|
|
return (EPERM);
|
1998-05-15 20:11:40 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* OK, now we're committed to doing something.
|
|
|
|
|
*/
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RLOCK(&V_tcbinfo);
|
|
|
|
|
gencnt = V_tcbinfo.ipi_gencnt;
|
|
|
|
|
n = V_tcbinfo.ipi_count;
|
|
|
|
|
INP_INFO_RUNLOCK(&V_tcbinfo);
|
1998-05-15 20:11:40 +00:00
|
|
|
|
2007-07-27 00:57:06 +00:00
|
|
|
m = syncache_pcbcount();
|
|
|
|
|
|
2004-02-26 00:27:04 +00:00
|
|
|
error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
|
2007-07-27 00:57:06 +00:00
|
|
|
+ (n + m) * sizeof(struct xtcpcb));
|
2004-02-26 00:27:04 +00:00
|
|
|
if (error != 0)
|
|
|
|
|
return (error);
|
2002-07-28 19:59:31 +00:00
|
|
|
|
1998-05-15 20:11:40 +00:00
|
|
|
xig.xig_len = sizeof xig;
|
2007-07-27 00:57:06 +00:00
|
|
|
xig.xig_count = n + m;
|
1998-05-15 20:11:40 +00:00
|
|
|
xig.xig_gen = gencnt;
|
|
|
|
|
xig.xig_sogen = so_gencnt;
|
|
|
|
|
error = SYSCTL_OUT(req, &xig, sizeof xig);
|
|
|
|
|
if (error)
|
2004-12-23 01:34:26 +00:00
|
|
|
return (error);
|
1998-05-15 20:11:40 +00:00
|
|
|
|
2007-07-27 00:57:06 +00:00
|
|
|
error = syncache_pcblist(req, m, &pcb_count);
|
|
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
|
2003-02-19 05:47:46 +00:00
|
|
|
inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
|
2004-04-05 00:49:07 +00:00
|
|
|
if (inp_list == NULL)
|
2004-12-23 01:34:26 +00:00
|
|
|
return (ENOMEM);
|
2004-08-16 18:32:07 +00:00
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RLOCK(&V_tcbinfo);
|
2008-08-20 01:05:56 +00:00
|
|
|
for (inp = LIST_FIRST(V_tcbinfo.ipi_listhead), i = 0;
|
2008-08-20 01:24:55 +00:00
|
|
|
inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) {
|
2010-03-17 18:28:27 +00:00
|
|
|
INP_WLOCK(inp);
|
2003-04-10 20:33:10 +00:00
|
|
|
if (inp->inp_gencnt <= gencnt) {
|
|
|
|
|
/*
|
|
|
|
|
* XXX: This use of cr_cansee(), introduced with
|
|
|
|
|
* TCP state changes, is not quite right, but for
|
|
|
|
|
* now, better than nothing.
|
|
|
|
|
*/
|
2009-03-15 09:58:31 +00:00
|
|
|
if (inp->inp_flags & INP_TIMEWAIT) {
|
2006-04-04 12:26:07 +00:00
|
|
|
if (intotw(inp) != NULL)
|
|
|
|
|
error = cr_cansee(req->td->td_ucred,
|
|
|
|
|
intotw(inp)->tw_cred);
|
|
|
|
|
else
|
|
|
|
|
error = EINVAL; /* Skip this inp. */
|
|
|
|
|
} else
|
2008-10-17 16:26:16 +00:00
|
|
|
error = cr_canseeinpcb(req->td->td_ucred, inp);
|
2010-03-17 18:28:27 +00:00
|
|
|
if (error == 0) {
|
|
|
|
|
in_pcbref(inp);
|
2003-04-10 20:33:10 +00:00
|
|
|
inp_list[i++] = inp;
|
2010-03-17 18:28:27 +00:00
|
|
|
}
|
2003-04-10 20:33:10 +00:00
|
|
|
}
|
2010-03-17 18:28:27 +00:00
|
|
|
INP_WUNLOCK(inp);
|
1998-05-15 20:11:40 +00:00
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RUNLOCK(&V_tcbinfo);
|
1998-05-15 20:11:40 +00:00
|
|
|
n = i;
|
|
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
for (i = 0; i < n; i++) {
|
|
|
|
|
inp = inp_list[i];
|
2008-05-29 14:28:26 +00:00
|
|
|
INP_RLOCK(inp);
|
1998-05-15 20:11:40 +00:00
|
|
|
if (inp->inp_gencnt <= gencnt) {
|
|
|
|
|
struct xtcpcb xt;
|
2006-04-03 13:33:55 +00:00
|
|
|
void *inp_ppcb;
|
2005-05-07 00:41:36 +00:00
|
|
|
|
|
|
|
|
bzero(&xt, sizeof(xt));
|
1998-05-15 20:11:40 +00:00
|
|
|
xt.xt_len = sizeof xt;
|
|
|
|
|
/* XXX should avoid extra copy */
|
|
|
|
|
bcopy(inp, &xt.xt_inp, sizeof *inp);
|
1999-06-16 19:05:17 +00:00
|
|
|
inp_ppcb = inp->inp_ppcb;
|
2003-02-19 22:32:43 +00:00
|
|
|
if (inp_ppcb == NULL)
|
|
|
|
|
bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);
|
2009-03-15 09:58:31 +00:00
|
|
|
else if (inp->inp_flags & INP_TIMEWAIT) {
|
1999-06-16 19:05:17 +00:00
|
|
|
bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);
|
2003-02-19 22:32:43 +00:00
|
|
|
xt.xt_tp.t_state = TCPS_TIME_WAIT;
|
2009-09-16 05:33:15 +00:00
|
|
|
} else {
|
2003-02-19 22:32:43 +00:00
|
|
|
bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp);
|
2009-09-16 05:33:15 +00:00
|
|
|
if (xt.xt_tp.t_timers)
|
|
|
|
|
tcp_timer_to_xtimer(&xt.xt_tp, xt.xt_tp.t_timers, &xt.xt_timer);
|
|
|
|
|
}
|
2004-04-05 00:49:07 +00:00
|
|
|
if (inp->inp_socket != NULL)
|
1998-05-15 20:11:40 +00:00
|
|
|
sotoxsocket(inp->inp_socket, &xt.xt_socket);
|
2003-02-19 22:32:43 +00:00
|
|
|
else {
|
|
|
|
|
bzero(&xt.xt_socket, sizeof xt.xt_socket);
|
|
|
|
|
xt.xt_socket.xso_protocol = IPPROTO_TCP;
|
|
|
|
|
}
|
2003-02-15 02:37:57 +00:00
|
|
|
xt.xt_inp.inp_gencnt = inp->inp_gencnt;
|
2008-05-29 14:28:26 +00:00
|
|
|
INP_RUNLOCK(inp);
|
1998-05-15 20:11:40 +00:00
|
|
|
error = SYSCTL_OUT(req, &xt, sizeof xt);
|
2006-07-18 22:34:27 +00:00
|
|
|
} else
|
2008-05-29 14:28:26 +00:00
|
|
|
INP_RUNLOCK(inp);
|
1998-05-15 20:11:40 +00:00
|
|
|
}
|
2010-03-17 18:28:27 +00:00
|
|
|
INP_INFO_WLOCK(&V_tcbinfo);
|
|
|
|
|
for (i = 0; i < n; i++) {
|
|
|
|
|
inp = inp_list[i];
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
INP_RLOCK(inp);
|
|
|
|
|
if (!in_pcbrele_rlocked(inp))
|
|
|
|
|
INP_RUNLOCK(inp);
|
2010-03-17 18:28:27 +00:00
|
|
|
}
|
|
|
|
|
INP_INFO_WUNLOCK(&V_tcbinfo);
|
|
|
|
|
|
1998-05-15 20:11:40 +00:00
|
|
|
if (!error) {
|
|
|
|
|
/*
|
|
|
|
|
* Give the user an updated idea of our state.
|
|
|
|
|
* If the generation differs from what we told
|
|
|
|
|
* her before, she knows that something happened
|
|
|
|
|
* while we were processing this request, and it
|
|
|
|
|
* might be necessary to retry.
|
|
|
|
|
*/
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_RLOCK(&V_tcbinfo);
|
|
|
|
|
xig.xig_gen = V_tcbinfo.ipi_gencnt;
|
1998-05-15 20:11:40 +00:00
|
|
|
xig.xig_sogen = so_gencnt;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
xig.xig_count = V_tcbinfo.ipi_count + pcb_count;
|
|
|
|
|
INP_INFO_RUNLOCK(&V_tcbinfo);
|
1998-05-15 20:11:40 +00:00
|
|
|
error = SYSCTL_OUT(req, &xig, sizeof xig);
|
|
|
|
|
}
|
|
|
|
|
free(inp_list, M_TEMP);
|
2004-12-23 01:34:26 +00:00
|
|
|
return (error);
|
1998-05-15 20:11:40 +00:00
|
|
|
}
|
|
|
|
|
|
2011-01-18 21:14:13 +00:00
|
|
|
SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist,
|
|
|
|
|
CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
|
2007-03-21 19:37:55 +00:00
|
|
|
tcp_pcblist, "S,xtcpcb", "List of active TCP connections");
|
1998-05-15 20:11:40 +00:00
|
|
|
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
1999-07-11 18:32:46 +00:00
|
|
|
static int
|
2000-07-04 11:25:35 +00:00
|
|
|
tcp_getcred(SYSCTL_HANDLER_ARGS)
|
1999-07-11 18:32:46 +00:00
|
|
|
{
|
2001-02-18 13:30:20 +00:00
|
|
|
struct xucred xuc;
|
1999-07-11 18:32:46 +00:00
|
|
|
struct sockaddr_in addrs[2];
|
|
|
|
|
struct inpcb *inp;
|
2005-07-19 12:21:26 +00:00
|
|
|
int error;
|
1999-07-11 18:32:46 +00:00
|
|
|
|
2007-06-12 00:12:01 +00:00
|
|
|
error = priv_check(req->td, PRIV_NETINET_GETCRED);
|
1999-07-11 18:32:46 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
error = SYSCTL_IN(req, addrs, sizeof(addrs));
|
|
|
|
|
if (error)
|
|
|
|
|
return (error);
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in_pcblookup(&V_tcbinfo, addrs[1].sin_addr, addrs[1].sin_port,
|
|
|
|
|
addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_RLOCKPCB, NULL);
|
2008-05-29 14:28:26 +00:00
|
|
|
if (inp != NULL) {
|
|
|
|
|
if (inp->inp_socket == NULL)
|
|
|
|
|
error = ENOENT;
|
|
|
|
|
if (error == 0)
|
2008-10-17 16:26:16 +00:00
|
|
|
error = cr_canseeinpcb(req->td->td_ucred, inp);
|
2008-05-29 14:28:26 +00:00
|
|
|
if (error == 0)
|
2008-10-04 15:06:34 +00:00
|
|
|
cru2x(inp->inp_cred, &xuc);
|
2008-05-29 14:28:26 +00:00
|
|
|
INP_RUNLOCK(inp);
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
} else
|
2002-07-11 23:13:31 +00:00
|
|
|
error = ENOENT;
|
2002-07-11 23:18:43 +00:00
|
|
|
if (error == 0)
|
|
|
|
|
error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
|
1999-07-11 18:32:46 +00:00
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
2001-06-24 12:18:27 +00:00
|
|
|
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred,
|
|
|
|
|
CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
|
|
|
|
|
tcp_getcred, "S,xucred", "Get the xucred of a TCP connection");
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
1999-07-11 18:32:46 +00:00
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
static int
|
2000-07-04 11:25:35 +00:00
|
|
|
tcp6_getcred(SYSCTL_HANDLER_ARGS)
|
2000-01-09 19:17:30 +00:00
|
|
|
{
|
2001-02-18 13:30:20 +00:00
|
|
|
struct xucred xuc;
|
2000-01-09 19:17:30 +00:00
|
|
|
struct sockaddr_in6 addrs[2];
|
|
|
|
|
struct inpcb *inp;
|
2011-04-30 11:21:29 +00:00
|
|
|
int error;
|
|
|
|
|
#ifdef INET
|
|
|
|
|
int mapped = 0;
|
|
|
|
|
#endif
|
2000-01-09 19:17:30 +00:00
|
|
|
|
2007-06-12 00:12:01 +00:00
|
|
|
error = priv_check(req->td, PRIV_NETINET_GETCRED);
|
2000-01-09 19:17:30 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
error = SYSCTL_IN(req, addrs, sizeof(addrs));
|
|
|
|
|
if (error)
|
|
|
|
|
return (error);
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
|
|
|
|
|
(error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
|
2005-07-25 12:31:43 +00:00
|
|
|
return (error);
|
|
|
|
|
}
|
2000-01-09 19:17:30 +00:00
|
|
|
if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) {
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2000-01-09 19:17:30 +00:00
|
|
|
if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr))
|
|
|
|
|
mapped = 1;
|
|
|
|
|
else
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2000-01-09 19:17:30 +00:00
|
|
|
return (EINVAL);
|
|
|
|
|
}
|
2005-07-25 12:31:43 +00:00
|
|
|
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2000-01-09 19:17:30 +00:00
|
|
|
if (mapped == 1)
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in_pcblookup(&V_tcbinfo,
|
2000-01-09 19:17:30 +00:00
|
|
|
*(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12],
|
|
|
|
|
addrs[1].sin6_port,
|
|
|
|
|
*(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12],
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL);
|
2000-01-09 19:17:30 +00:00
|
|
|
else
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in6_pcblookup(&V_tcbinfo,
|
2005-10-12 09:24:18 +00:00
|
|
|
&addrs[1].sin6_addr, addrs[1].sin6_port,
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
&addrs[0].sin6_addr, addrs[0].sin6_port,
|
|
|
|
|
INPLOOKUP_RLOCKPCB, NULL);
|
2008-05-29 14:28:26 +00:00
|
|
|
if (inp != NULL) {
|
|
|
|
|
if (inp->inp_socket == NULL)
|
|
|
|
|
error = ENOENT;
|
|
|
|
|
if (error == 0)
|
2008-10-17 16:26:16 +00:00
|
|
|
error = cr_canseeinpcb(req->td->td_ucred, inp);
|
2008-05-29 14:28:26 +00:00
|
|
|
if (error == 0)
|
2008-10-04 15:06:34 +00:00
|
|
|
cru2x(inp->inp_cred, &xuc);
|
2008-05-29 14:28:26 +00:00
|
|
|
INP_RUNLOCK(inp);
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
} else
|
2002-07-11 23:13:31 +00:00
|
|
|
error = ENOENT;
|
2002-07-11 23:18:43 +00:00
|
|
|
if (error == 0)
|
|
|
|
|
error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
|
2000-01-09 19:17:30 +00:00
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
2001-06-24 12:18:27 +00:00
|
|
|
SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred,
|
|
|
|
|
CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
|
|
|
|
|
tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection");
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET6 */
|
2000-01-09 19:17:30 +00:00
|
|
|
|
|
|
|
|
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
1994-05-24 10:09:53 +00:00
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-02-23 20:51:46 +00:00
|
|
|
struct ip *ip = vip;
|
|
|
|
|
struct tcphdr *th;
|
2001-02-26 21:19:47 +00:00
|
|
|
struct in_addr faddr;
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
struct tcpcb *tp;
|
2002-06-14 08:35:21 +00:00
|
|
|
struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify;
|
2005-04-21 14:29:34 +00:00
|
|
|
struct icmp *icp;
|
|
|
|
|
struct in_conninfo inc;
|
|
|
|
|
tcp_seq icmp_tcp_seq;
|
2005-07-19 12:21:26 +00:00
|
|
|
int mtu;
|
2001-02-26 21:19:47 +00:00
|
|
|
|
|
|
|
|
faddr = ((struct sockaddr_in *)sa)->sin_addr;
|
|
|
|
|
if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
|
|
|
|
|
return;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2005-04-21 12:37:12 +00:00
|
|
|
if (cmd == PRC_MSGSIZE)
|
|
|
|
|
notify = tcp_mtudisc;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
else if (V_icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
|
2002-06-30 20:07:21 +00:00
|
|
|
cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip)
|
2001-02-23 20:51:46 +00:00
|
|
|
notify = tcp_drop_syn_sent;
|
2003-11-20 20:07:39 +00:00
|
|
|
/*
|
|
|
|
|
* Redirects don't need to be handled up here.
|
|
|
|
|
*/
|
|
|
|
|
else if (PRC_IS_REDIRECT(cmd))
|
|
|
|
|
return;
|
2005-04-21 12:37:12 +00:00
|
|
|
/*
|
|
|
|
|
* Source quench is depreciated.
|
|
|
|
|
*/
|
|
|
|
|
else if (cmd == PRC_QUENCH)
|
|
|
|
|
return;
|
2003-11-20 20:07:39 +00:00
|
|
|
/*
|
|
|
|
|
* Hostdead is ugly because it goes linearly through all PCBs.
|
|
|
|
|
* XXX: We never get this from ICMP, otherwise it makes an
|
|
|
|
|
* excellent DoS attack on machines with many connections.
|
|
|
|
|
*/
|
|
|
|
|
else if (cmd == PRC_HOSTDEAD)
|
2004-04-05 00:49:07 +00:00
|
|
|
ip = NULL;
|
2003-09-11 21:40:21 +00:00
|
|
|
else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
|
1994-05-24 10:09:53 +00:00
|
|
|
return;
|
2004-04-05 00:49:07 +00:00
|
|
|
if (ip != NULL) {
|
2005-04-21 14:29:34 +00:00
|
|
|
icp = (struct icmp *)((caddr_t)ip
|
|
|
|
|
- offsetof(struct icmp, icmp_ip));
|
2004-08-16 18:32:07 +00:00
|
|
|
th = (struct tcphdr *)((caddr_t)ip
|
2002-10-20 22:52:07 +00:00
|
|
|
+ (ip->ip_hl << 2));
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK(&V_tcbinfo);
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport,
|
|
|
|
|
ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL);
|
2002-06-10 20:05:46 +00:00
|
|
|
if (inp != NULL) {
|
2009-03-15 09:58:31 +00:00
|
|
|
if (!(inp->inp_flags & INP_TIMEWAIT) &&
|
|
|
|
|
!(inp->inp_flags & INP_DROPPED) &&
|
2006-04-03 14:07:50 +00:00
|
|
|
!(inp->inp_socket == NULL)) {
|
2005-04-21 14:29:34 +00:00
|
|
|
icmp_tcp_seq = htonl(th->th_seq);
|
2002-06-10 20:05:46 +00:00
|
|
|
tp = intotcpcb(inp);
|
2005-04-21 14:29:34 +00:00
|
|
|
if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
|
|
|
|
|
SEQ_LT(icmp_tcp_seq, tp->snd_max)) {
|
|
|
|
|
if (cmd == PRC_MSGSIZE) {
|
|
|
|
|
/*
|
|
|
|
|
* MTU discovery:
|
|
|
|
|
* If we got a needfrag set the MTU
|
|
|
|
|
* in the route to the suggested new
|
|
|
|
|
* value (if given) and then notify.
|
|
|
|
|
*/
|
|
|
|
|
bzero(&inc, sizeof(inc));
|
|
|
|
|
inc.inc_faddr = faddr;
|
Add code to allow the system to handle multiple routing tables.
This particular implementation is designed to be fully backwards compatible
and to be MFC-able to 7.x (and 6.x)
Currently the only protocol that can make use of the multiple tables is IPv4
Similar functionality exists in OpenBSD and Linux.
From my notes:
-----
One thing where FreeBSD has been falling behind, and which by chance I
have some time to work on is "policy based routing", which allows
different
packet streams to be routed by more than just the destination address.
Constraints:
------------
I want to make some form of this available in the 6.x tree
(and by extension 7.x) , but FreeBSD in general needs it so I might as
well do it in -current and back port the portions I need.
One of the ways that this can be done is to have the ability to
instantiate multiple kernel routing tables (which I will now
refer to as "Forwarding Information Bases" or "FIBs" for political
correctness reasons). Which FIB a particular packet uses to make
the next hop decision can be decided by a number of mechanisms.
The policies these mechanisms implement are the "Policies" referred
to in "Policy based routing".
One of the constraints I have if I try to back port this work to
6.x is that it must be implemented as a EXTENSION to the existing
ABIs in 6.x so that third party applications do not need to be
recompiled in timespan of the branch.
This first version will not have some of the bells and whistles that
will come with later versions. It will, for example, be limited to 16
tables in the first commit.
Implementation method, Compatible version. (part 1)
-------------------------------
For this reason I have implemented a "sufficient subset" of a
multiple routing table solution in Perforce, and back-ported it
to 6.x. (also in Perforce though not always caught up with what I
have done in -current/P4). The subset allows a number of FIBs
to be defined at compile time (8 is sufficient for my purposes in 6.x)
and implements the changes needed to allow IPV4 to use them. I have not
done the changes for ipv6 simply because I do not need it, and I do not
have enough knowledge of ipv6 (e.g. neighbor discovery) needed to do it.
Other protocol families are left untouched and should there be
users with proprietary protocol families, they should continue to work
and be oblivious to the existence of the extra FIBs.
To understand how this is done, one must know that the current FIB
code starts everything off with a single dimensional array of
pointers to FIB head structures (One per protocol family), each of
which in turn points to the trie of routes available to that family.
The basic change in the ABI compatible version of the change is to
extent that array to be a 2 dimensional array, so that
instead of protocol family X looking at rt_tables[X] for the
table it needs, it looks at rt_tables[Y][X] when for all
protocol families except ipv4 Y is always 0.
Code that is unaware of the change always just sees the first row
of the table, which of course looks just like the one dimensional
array that existed before.
The entry points rtrequest(), rtalloc(), rtalloc1(), rtalloc_ign()
are all maintained, but refer only to the first row of the array,
so that existing callers in proprietary protocols can continue to
do the "right thing".
Some new entry points are added, for the exclusive use of ipv4 code
called in_rtrequest(), in_rtalloc(), in_rtalloc1() and in_rtalloc_ign(),
which have an extra argument which refers the code to the correct row.
In addition, there are some new entry points (currently called
rtalloc_fib() and friends) that check the Address family being
looked up and call either rtalloc() (and friends) if the protocol
is not IPv4 forcing the action to row 0 or to the appropriate row
if it IS IPv4 (and that info is available). These are for calling
from code that is not specific to any particular protocol. The way
these are implemented would change in the non ABI preserving code
to be added later.
One feature of the first version of the code is that for ipv4,
the interface routes show up automatically on all the FIBs, so
that no matter what FIB you select you always have the basic
direct attached hosts available to you. (rtinit() does this
automatically).
You CAN delete an interface route from one FIB should you want
to but by default it's there. ARP information is also available
in each FIB. It's assumed that the same machine would have the
same MAC address, regardless of which FIB you are using to get
to it.
This brings us as to how the correct FIB is selected for an outgoing
IPV4 packet.
Firstly, all packets have a FIB associated with them. if nothing
has been done to change it, it will be FIB 0. The FIB is changed
in the following ways.
Packets fall into one of a number of classes.
1/ locally generated packets, coming from a socket/PCB.
Such packets select a FIB from a number associated with the
socket/PCB. This in turn is inherited from the process,
but can be changed by a socket option. The process in turn
inherits it on fork. I have written a utility call setfib
that acts a bit like nice..
setfib -3 ping target.example.com # will use fib 3 for ping.
It is an obvious extension to make it a property of a jail
but I have not done so. It can be achieved by combining the setfib and
jail commands.
2/ packets received on an interface for forwarding.
By default these packets would use table 0,
(or possibly a number settable in a sysctl(not yet)).
but prior to routing the firewall can inspect them (see below).
(possibly in the future you may be able to associate a FIB
with packets received on an interface.. An ifconfig arg, but not yet.)
3/ packets inspected by a packet classifier, which can arbitrarily
associate a fib with it on a packet by packet basis.
A fib assigned to a packet by a packet classifier
(such as ipfw) would over-ride a fib associated by
a more default source. (such as cases 1 or 2).
4/ a tcp listen socket associated with a fib will generate
accept sockets that are associated with that same fib.
5/ Packets generated in response to some other packet (e.g. reset
or icmp packets). These should use the FIB associated with the
packet being reponded to.
6/ Packets generated during encapsulation.
gif, tun and other tunnel interfaces will encapsulate using the FIB
that was in effect withthe proces that set up the tunnel.
thus setfib 1 ifconfig gif0 [tunnel instructions]
will set the fib for the tunnel to use to be fib 1.
Routing messages would be associated with their
process, and thus select one FIB or another.
messages from the kernel would be associated with the fib they
refer to and would only be received by a routing socket associated
with that fib. (not yet implemented)
In addition Netstat has been edited to be able to cope with the
fact that the array is now 2 dimensional. (It looks in system
memory using libkvm (!)). Old versions of netstat see only the first FIB.
In addition two sysctls are added to give:
a) the number of FIBs compiled in (active)
b) the default FIB of the calling process.
Early testing experience:
-------------------------
Basically our (IronPort's) appliance does this functionality already
using ipfw fwd but that method has some drawbacks.
For example,
It can't fully simulate a routing table because it can't influence the
socket's choice of local address when a connect() is done.
Testing during the generating of these changes has been
remarkably smooth so far. Multiple tables have co-existed
with no notable side effects, and packets have been routes
accordingly.
ipfw has grown 2 new keywords:
setfib N ip from anay to any
count ip from any to any fib N
In pf there seems to be a requirement to be able to give symbolic names to the
fibs but I do not have that capacity. I am not sure if it is required.
SCTP has interestingly enough built in support for this, called VRFs
in Cisco parlance. it will be interesting to see how that handles it
when it suddenly actually does something.
Where to next:
--------------------
After committing the ABI compatible version and MFCing it, I'd
like to proceed in a forward direction in -current. this will
result in some roto-tilling in the routing code.
Firstly: the current code's idea of having a separate tree per
protocol family, all of the same format, and pointed to by the
1 dimensional array is a bit silly. Especially when one considers that
there is code that makes assumptions about every protocol having the
same internal structures there. Some protocols don't WANT that
sort of structure. (for example the whole idea of a netmask is foreign
to appletalk). This needs to be made opaque to the external code.
My suggested first change is to add routing method pointers to the
'domain' structure, along with information pointing the data.
instead of having an array of pointers to uniform structures,
there would be an array pointing to the 'domain' structures
for each protocol address domain (protocol family),
and the methods this reached would be called. The methods would have
an argument that gives FIB number, but the protocol would be free
to ignore it.
When the ABI can be changed it raises the possibilty of the
addition of a fib entry into the "struct route". Currently,
the structure contains the sockaddr of the desination, and the resulting
fib entry. To make this work fully, one could add a fib number
so that given an address and a fib, one can find the third element, the
fib entry.
Interaction with the ARP layer/ LL layer would need to be
revisited as well. Qing Li has been working on this already.
This work was sponsored by Ironport Systems/Cisco
Reviewed by: several including rwatson, bz and mlair (parts each)
Obtained from: Ironport systems/Cisco
2008-05-09 23:03:00 +00:00
|
|
|
inc.inc_fibnum =
|
|
|
|
|
inp->inp_inc.inc_fibnum;
|
2005-04-21 14:29:34 +00:00
|
|
|
|
|
|
|
|
mtu = ntohs(icp->icmp_nextmtu);
|
2005-05-04 13:48:44 +00:00
|
|
|
/*
|
|
|
|
|
* If no alternative MTU was
|
|
|
|
|
* proposed, try the next smaller
|
2005-09-10 07:43:29 +00:00
|
|
|
* one. ip->ip_len has already
|
|
|
|
|
* been swapped in icmp_input().
|
2005-05-04 13:48:44 +00:00
|
|
|
*/
|
|
|
|
|
if (!mtu)
|
2005-09-10 07:43:29 +00:00
|
|
|
mtu = ip_next_mtu(ip->ip_len,
|
2005-05-04 13:48:44 +00:00
|
|
|
1);
|
2010-08-15 13:25:18 +00:00
|
|
|
if (mtu < V_tcp_minmss
|
|
|
|
|
+ sizeof(struct tcpiphdr))
|
|
|
|
|
mtu = V_tcp_minmss
|
2005-05-04 13:48:44 +00:00
|
|
|
+ sizeof(struct tcpiphdr);
|
|
|
|
|
/*
|
2011-02-21 09:01:34 +00:00
|
|
|
* Only cache the MTU if it
|
2005-05-04 13:48:44 +00:00
|
|
|
* is smaller than the interface
|
|
|
|
|
* or route MTU. tcp_mtudisc()
|
|
|
|
|
* will do right thing by itself.
|
|
|
|
|
*/
|
2006-09-06 21:51:59 +00:00
|
|
|
if (mtu <= tcp_maxmtu(&inc, NULL))
|
2005-04-21 14:29:34 +00:00
|
|
|
tcp_hc_updatemtu(&inc, mtu);
|
|
|
|
|
}
|
|
|
|
|
|
2002-06-14 08:35:21 +00:00
|
|
|
inp = (*notify)(inp, inetctlerrmap[cmd]);
|
2005-04-21 14:29:34 +00:00
|
|
|
}
|
2002-06-10 20:05:46 +00:00
|
|
|
}
|
2004-04-05 00:49:07 +00:00
|
|
|
if (inp != NULL)
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2001-11-22 04:50:44 +00:00
|
|
|
} else {
|
2008-12-17 12:52:34 +00:00
|
|
|
bzero(&inc, sizeof(inc));
|
2001-11-22 04:50:44 +00:00
|
|
|
inc.inc_fport = th->th_dport;
|
|
|
|
|
inc.inc_lport = th->th_sport;
|
|
|
|
|
inc.inc_faddr = faddr;
|
|
|
|
|
inc.inc_laddr = ip->ip_src;
|
|
|
|
|
syncache_unreach(&inc, th);
|
2001-02-26 21:19:47 +00:00
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WUNLOCK(&V_tcbinfo);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
void
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
|
2000-01-09 19:17:30 +00:00
|
|
|
{
|
|
|
|
|
struct tcphdr th;
|
2002-06-14 08:35:21 +00:00
|
|
|
struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify;
|
2000-01-09 19:17:30 +00:00
|
|
|
struct ip6_hdr *ip6;
|
|
|
|
|
struct mbuf *m;
|
2001-06-11 12:39:29 +00:00
|
|
|
struct ip6ctlparam *ip6cp = NULL;
|
|
|
|
|
const struct sockaddr_in6 *sa6_src = NULL;
|
2000-01-09 19:17:30 +00:00
|
|
|
int off;
|
2001-06-11 12:39:29 +00:00
|
|
|
struct tcp_portonly {
|
|
|
|
|
u_int16_t th_sport;
|
|
|
|
|
u_int16_t th_dport;
|
|
|
|
|
} *thp;
|
2000-01-09 19:17:30 +00:00
|
|
|
|
|
|
|
|
if (sa->sa_family != AF_INET6 ||
|
|
|
|
|
sa->sa_len != sizeof(struct sockaddr_in6))
|
|
|
|
|
return;
|
|
|
|
|
|
2005-04-21 12:37:12 +00:00
|
|
|
if (cmd == PRC_MSGSIZE)
|
2000-01-09 19:17:30 +00:00
|
|
|
notify = tcp_mtudisc;
|
|
|
|
|
else if (!PRC_IS_REDIRECT(cmd) &&
|
2003-09-11 21:40:21 +00:00
|
|
|
((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0))
|
2000-01-09 19:17:30 +00:00
|
|
|
return;
|
2005-04-21 12:37:12 +00:00
|
|
|
/* Source quench is depreciated. */
|
|
|
|
|
else if (cmd == PRC_QUENCH)
|
|
|
|
|
return;
|
2000-01-09 19:17:30 +00:00
|
|
|
|
|
|
|
|
/* if the parameter is from icmp6, decode it. */
|
|
|
|
|
if (d != NULL) {
|
2001-06-11 12:39:29 +00:00
|
|
|
ip6cp = (struct ip6ctlparam *)d;
|
2000-01-09 19:17:30 +00:00
|
|
|
m = ip6cp->ip6c_m;
|
|
|
|
|
ip6 = ip6cp->ip6c_ip6;
|
|
|
|
|
off = ip6cp->ip6c_off;
|
2001-06-11 12:39:29 +00:00
|
|
|
sa6_src = ip6cp->ip6c_src;
|
2000-01-09 19:17:30 +00:00
|
|
|
} else {
|
|
|
|
|
m = NULL;
|
|
|
|
|
ip6 = NULL;
|
2000-10-23 07:11:01 +00:00
|
|
|
off = 0; /* fool gcc */
|
2001-06-11 12:39:29 +00:00
|
|
|
sa6_src = &sa6_any;
|
2000-01-09 19:17:30 +00:00
|
|
|
}
|
|
|
|
|
|
2004-04-05 00:49:07 +00:00
|
|
|
if (ip6 != NULL) {
|
2001-11-22 04:50:44 +00:00
|
|
|
struct in_conninfo inc;
|
2000-01-09 19:17:30 +00:00
|
|
|
/*
|
|
|
|
|
* XXX: We assume that when IPV6 is non NULL,
|
|
|
|
|
* M and OFF are valid.
|
|
|
|
|
*/
|
|
|
|
|
|
2000-10-23 07:11:01 +00:00
|
|
|
/* check if we can safely examine src and dst ports */
|
2001-06-11 12:39:29 +00:00
|
|
|
if (m->m_pkthdr.len < off + sizeof(*thp))
|
2000-10-23 07:11:01 +00:00
|
|
|
return;
|
|
|
|
|
|
2001-06-11 12:39:29 +00:00
|
|
|
bzero(&th, sizeof(th));
|
|
|
|
|
m_copydata(m, off, sizeof(*thp), (caddr_t)&th);
|
|
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
in6_pcbnotify(&V_tcbinfo, sa, th.th_dport,
|
2001-06-11 12:39:29 +00:00
|
|
|
(struct sockaddr *)ip6cp->ip6c_src,
|
2004-02-13 14:50:01 +00:00
|
|
|
th.th_sport, cmd, NULL, notify);
|
2001-11-22 04:50:44 +00:00
|
|
|
|
2008-12-17 12:52:34 +00:00
|
|
|
bzero(&inc, sizeof(inc));
|
2001-11-22 04:50:44 +00:00
|
|
|
inc.inc_fport = th.th_dport;
|
|
|
|
|
inc.inc_lport = th.th_sport;
|
|
|
|
|
inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr;
|
|
|
|
|
inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr;
|
2008-12-17 12:52:34 +00:00
|
|
|
inc.inc_flags |= INC_ISIPV6;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK(&V_tcbinfo);
|
2001-11-22 04:50:44 +00:00
|
|
|
syncache_unreach(&inc, &th);
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WUNLOCK(&V_tcbinfo);
|
2000-01-09 19:17:30 +00:00
|
|
|
} else
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src,
|
2004-02-13 14:50:01 +00:00
|
|
|
0, cmd, NULL, notify);
|
2000-01-09 19:17:30 +00:00
|
|
|
}
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
|
2001-04-17 18:08:01 +00:00
|
|
|
|
2001-08-22 00:58:16 +00:00
|
|
|
/*
|
|
|
|
|
* Following is where TCP initial sequence number generation occurs.
|
|
|
|
|
*
|
|
|
|
|
* There are two places where we must use initial sequence numbers:
|
|
|
|
|
* 1. In SYN-ACK packets.
|
|
|
|
|
* 2. In SYN packets.
|
|
|
|
|
*
|
2002-04-10 22:12:01 +00:00
|
|
|
* All ISNs for SYN-ACK packets are generated by the syncache. See
|
|
|
|
|
* tcp_syncache.c for details.
|
2001-08-22 00:58:16 +00:00
|
|
|
*
|
|
|
|
|
* The ISNs in SYN packets must be monotonic; TIME_WAIT recycling
|
|
|
|
|
* depends on this property. In addition, these ISNs should be
|
|
|
|
|
* unguessable so as to prevent connection hijacking. To satisfy
|
|
|
|
|
* the requirements of this situation, the algorithm outlined in
|
2004-08-16 18:32:07 +00:00
|
|
|
* RFC 1948 is used, with only small modifications.
|
2001-08-22 00:58:16 +00:00
|
|
|
*
|
|
|
|
|
* Implementation details:
|
|
|
|
|
*
|
|
|
|
|
* Time is based off the system timer, and is corrected so that it
|
|
|
|
|
* increases by one megabyte per second. This allows for proper
|
|
|
|
|
* recycling on high speed LANs while still leaving over an hour
|
|
|
|
|
* before rollover.
|
|
|
|
|
*
|
2004-04-20 06:33:39 +00:00
|
|
|
* As reading the *exact* system time is too expensive to be done
|
|
|
|
|
* whenever setting up a TCP connection, we increment the time
|
|
|
|
|
* offset in two ways. First, a small random positive increment
|
|
|
|
|
* is added to isn_offset for each connection that is set up.
|
|
|
|
|
* Second, the function tcp_isn_tick fires once per clock tick
|
|
|
|
|
* and increments isn_offset as necessary so that sequence numbers
|
|
|
|
|
* are incremented at approximately ISN_BYTES_PER_SECOND. The
|
|
|
|
|
* random positive increments serve only to ensure that the same
|
|
|
|
|
* exact sequence number is never sent out twice (as could otherwise
|
|
|
|
|
* happen when a port is recycled in less than the system tick
|
|
|
|
|
* interval.)
|
|
|
|
|
*
|
2001-08-22 00:58:16 +00:00
|
|
|
* net.inet.tcp.isn_reseed_interval controls the number of seconds
|
|
|
|
|
* between seeding of isn_secret. This is normally set to zero,
|
|
|
|
|
* as reseeding should not be necessary.
|
|
|
|
|
*
|
2004-11-23 15:59:43 +00:00
|
|
|
* Locking of the global variables isn_secret, isn_last_reseed, isn_offset,
|
|
|
|
|
* isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In
|
|
|
|
|
* general, this means holding an exclusive (write) lock.
|
2001-08-22 00:58:16 +00:00
|
|
|
*/
|
2001-04-17 18:08:01 +00:00
|
|
|
|
2001-08-22 00:58:16 +00:00
|
|
|
#define ISN_BYTES_PER_SECOND 1048576
|
2004-04-20 06:33:39 +00:00
|
|
|
#define ISN_STATIC_INCREMENT 4096
|
|
|
|
|
#define ISN_RANDOM_INCREMENT (4096 - 1)
|
2001-04-17 18:08:01 +00:00
|
|
|
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(u_char, isn_secret[32]);
|
2011-05-09 07:37:47 +00:00
|
|
|
static VNET_DEFINE(int, isn_last);
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(int, isn_last_reseed);
|
|
|
|
|
static VNET_DEFINE(u_int32_t, isn_offset);
|
|
|
|
|
static VNET_DEFINE(u_int32_t, isn_offset_old);
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
|
2009-07-16 21:13:04 +00:00
|
|
|
#define V_isn_secret VNET(isn_secret)
|
2011-05-09 07:37:47 +00:00
|
|
|
#define V_isn_last VNET(isn_last)
|
2009-07-16 21:13:04 +00:00
|
|
|
#define V_isn_last_reseed VNET(isn_last_reseed)
|
|
|
|
|
#define V_isn_offset VNET(isn_offset)
|
|
|
|
|
#define V_isn_offset_old VNET(isn_offset_old)
|
2001-04-17 18:08:01 +00:00
|
|
|
|
|
|
|
|
tcp_seq
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_new_isn(struct tcpcb *tp)
|
2001-04-17 18:08:01 +00:00
|
|
|
{
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5_CTX isn_ctx;
|
2001-08-22 00:58:16 +00:00
|
|
|
u_int32_t md5_buffer[4];
|
|
|
|
|
tcp_seq new_isn;
|
2011-05-09 07:37:47 +00:00
|
|
|
u_int32_t projected_offset;
|
2001-08-22 00:58:16 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(tp->t_inpcb);
|
2004-11-23 15:59:43 +00:00
|
|
|
|
2006-04-23 12:27:42 +00:00
|
|
|
ISN_LOCK();
|
2001-08-22 00:58:16 +00:00
|
|
|
/* Seed if this is the first use, reseed if requested. */
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if ((V_isn_last_reseed == 0) || ((V_tcp_isn_reseed_interval > 0) &&
|
|
|
|
|
(((u_int)V_isn_last_reseed + (u_int)V_tcp_isn_reseed_interval*hz)
|
2001-08-22 00:58:16 +00:00
|
|
|
< (u_int)ticks))) {
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
read_random(&V_isn_secret, sizeof(V_isn_secret));
|
|
|
|
|
V_isn_last_reseed = ticks;
|
2001-08-22 00:58:16 +00:00
|
|
|
}
|
2004-08-16 18:32:07 +00:00
|
|
|
|
2001-08-22 00:58:16 +00:00
|
|
|
/* Compute the md5 hash and return the ISN. */
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Init(&isn_ctx);
|
|
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short));
|
|
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short));
|
2001-08-22 00:58:16 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) {
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr,
|
2001-08-22 00:58:16 +00:00
|
|
|
sizeof(struct in6_addr));
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr,
|
2001-08-22 00:58:16 +00:00
|
|
|
sizeof(struct in6_addr));
|
|
|
|
|
} else
|
|
|
|
|
#endif
|
|
|
|
|
{
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr,
|
2001-08-22 00:58:16 +00:00
|
|
|
sizeof(struct in_addr));
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr,
|
2001-08-22 00:58:16 +00:00
|
|
|
sizeof(struct in_addr));
|
|
|
|
|
}
|
2008-12-13 21:59:18 +00:00
|
|
|
MD5Update(&isn_ctx, (u_char *) &V_isn_secret, sizeof(V_isn_secret));
|
|
|
|
|
MD5Final((u_char *) &md5_buffer, &isn_ctx);
|
2001-08-22 00:58:16 +00:00
|
|
|
new_isn = (tcp_seq) md5_buffer[0];
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
V_isn_offset += ISN_STATIC_INCREMENT +
|
2004-04-20 06:33:39 +00:00
|
|
|
(arc4random() & ISN_RANDOM_INCREMENT);
|
2011-05-09 07:37:47 +00:00
|
|
|
if (ticks != V_isn_last) {
|
|
|
|
|
projected_offset = V_isn_offset_old +
|
|
|
|
|
ISN_BYTES_PER_SECOND / hz * (ticks - V_isn_last);
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
if (SEQ_GT(projected_offset, V_isn_offset))
|
|
|
|
|
V_isn_offset = projected_offset;
|
|
|
|
|
V_isn_offset_old = V_isn_offset;
|
2011-05-09 07:37:47 +00:00
|
|
|
V_isn_last = ticks;
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
}
|
2011-05-09 07:37:47 +00:00
|
|
|
new_isn += V_isn_offset;
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
ISN_UNLOCK();
|
2011-05-09 07:37:47 +00:00
|
|
|
return (new_isn);
|
2004-04-20 06:33:39 +00:00
|
|
|
}
|
|
|
|
|
|
We currently does not react to ICMP administratively prohibited
messages send by routers when they deny our traffic, this causes
a timeout when trying to connect to TCP ports/services on a remote
host, which is blocked by routers or firewalls.
rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually
requi re that we treat such a message for a TCP session, that we
treat it like if we had recieved a RST.
quote begin.
A Destination Unreachable message that is received MUST be
reported to the transport layer. The transport layer SHOULD
use the information appropriately; for example, see Sections
4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol
that has its own mechanism for notifying the sender that a
port is unreachable (e.g., TCP, which sends RST segments)
MUST nevertheless accept an ICMP Port Unreachable for the
same purpose.
quote end.
I've written a small extension that implement this, it also create
a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if
this new behaviour is activated.
When it's activated (set to 1) we'll treat a ICMP administratively
prohibited message (icmp type 3 code 9, 10 and 13) for a TCP
sessions, as if we recived a TCP RST, but only if the TCP session
is in SYN_SENT state.
The reason for only reacting when in SYN_SENT state, is that this
will solve the problem, and at the same time minimize the risk of
this being abused.
I suggest that we enable this new behaviour by default, but it
would be a change of current behaviour, so if people prefer to
leave it disabled by default, at least for now, this would be ok
for me, the attached diff actually have the sysctl set to 0 by
default.
PR: 23086
Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
|
|
|
/*
|
2001-02-23 20:51:46 +00:00
|
|
|
* When a specific ICMP unreachable message is received and the
|
|
|
|
|
* connection state is SYN-SENT, drop the connection. This behavior
|
|
|
|
|
* is controlled by the icmp_may_rst sysctl.
|
We currently does not react to ICMP administratively prohibited
messages send by routers when they deny our traffic, this causes
a timeout when trying to connect to TCP ports/services on a remote
host, which is blocked by routers or firewalls.
rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually
requi re that we treat such a message for a TCP session, that we
treat it like if we had recieved a RST.
quote begin.
A Destination Unreachable message that is received MUST be
reported to the transport layer. The transport layer SHOULD
use the information appropriately; for example, see Sections
4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol
that has its own mechanism for notifying the sender that a
port is unreachable (e.g., TCP, which sends RST segments)
MUST nevertheless accept an ICMP Port Unreachable for the
same purpose.
quote end.
I've written a small extension that implement this, it also create
a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if
this new behaviour is activated.
When it's activated (set to 1) we'll treat a ICMP administratively
prohibited message (icmp type 3 code 9, 10 and 13) for a TCP
sessions, as if we recived a TCP RST, but only if the TCP session
is in SYN_SENT state.
The reason for only reacting when in SYN_SENT state, is that this
will solve the problem, and at the same time minimize the risk of
this being abused.
I suggest that we enable this new behaviour by default, but it
would be a change of current behaviour, so if people prefer to
leave it disabled by default, at least for now, this would be ok
for me, the attached diff actually have the sysctl set to 0 by
default.
PR: 23086
Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
|
|
|
*/
|
2002-06-14 08:35:21 +00:00
|
|
|
struct inpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_drop_syn_sent(struct inpcb *inp, int errno)
|
We currently does not react to ICMP administratively prohibited
messages send by routers when they deny our traffic, this causes
a timeout when trying to connect to TCP ports/services on a remote
host, which is blocked by routers or firewalls.
rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually
requi re that we treat such a message for a TCP session, that we
treat it like if we had recieved a RST.
quote begin.
A Destination Unreachable message that is received MUST be
reported to the transport layer. The transport layer SHOULD
use the information appropriately; for example, see Sections
4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol
that has its own mechanism for notifying the sender that a
port is unreachable (e.g., TCP, which sends RST segments)
MUST nevertheless accept an ICMP Port Unreachable for the
same purpose.
quote end.
I've written a small extension that implement this, it also create
a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if
this new behaviour is activated.
When it's activated (set to 1) we'll treat a ICMP administratively
prohibited message (icmp type 3 code 9, 10 and 13) for a TCP
sessions, as if we recived a TCP RST, but only if the TCP session
is in SYN_SENT state.
The reason for only reacting when in SYN_SENT state, is that this
will solve the problem, and at the same time minimize the risk of
this being abused.
I suggest that we enable this new behaviour by default, but it
would be a change of current behaviour, so if people prefer to
leave it disabled by default, at least for now, this would be ok
for me, the attached diff actually have the sysctl set to 0 by
default.
PR: 23086
Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
|
|
|
{
|
2006-04-03 13:33:55 +00:00
|
|
|
struct tcpcb *tp;
|
We currently does not react to ICMP administratively prohibited
messages send by routers when they deny our traffic, this causes
a timeout when trying to connect to TCP ports/services on a remote
host, which is blocked by routers or firewalls.
rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually
requi re that we treat such a message for a TCP session, that we
treat it like if we had recieved a RST.
quote begin.
A Destination Unreachable message that is received MUST be
reported to the transport layer. The transport layer SHOULD
use the information appropriately; for example, see Sections
4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol
that has its own mechanism for notifying the sender that a
port is unreachable (e.g., TCP, which sends RST segments)
MUST nevertheless accept an ICMP Port Unreachable for the
same purpose.
quote end.
I've written a small extension that implement this, it also create
a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if
this new behaviour is activated.
When it's activated (set to 1) we'll treat a ICMP administratively
prohibited message (icmp type 3 code 9, 10 and 13) for a TCP
sessions, as if we recived a TCP RST, but only if the TCP session
is in SYN_SENT state.
The reason for only reacting when in SYN_SENT state, is that this
will solve the problem, and at the same time minimize the risk of
this being abused.
I suggest that we enable this new behaviour by default, but it
would be a change of current behaviour, so if people prefer to
leave it disabled by default, at least for now, this would be ok
for me, the attached diff actually have the sysctl set to 0 by
default.
PR: 23086
Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2006-04-03 14:07:50 +00:00
|
|
|
|
2009-03-15 09:58:31 +00:00
|
|
|
if ((inp->inp_flags & INP_TIMEWAIT) ||
|
|
|
|
|
(inp->inp_flags & INP_DROPPED))
|
2006-04-03 14:07:50 +00:00
|
|
|
return (inp);
|
|
|
|
|
|
2006-04-03 13:33:55 +00:00
|
|
|
tp = intotcpcb(inp);
|
2006-04-03 14:07:50 +00:00
|
|
|
if (tp->t_state != TCPS_SYN_SENT)
|
|
|
|
|
return (inp);
|
2005-06-01 12:06:07 +00:00
|
|
|
|
2006-04-03 14:07:50 +00:00
|
|
|
tp = tcp_drop(tp, errno);
|
|
|
|
|
if (tp != NULL)
|
|
|
|
|
return (inp);
|
|
|
|
|
else
|
|
|
|
|
return (NULL);
|
2001-02-18 09:34:55 +00:00
|
|
|
}
|
|
|
|
|
|
1995-09-18 15:51:40 +00:00
|
|
|
/*
|
|
|
|
|
* When `need fragmentation' ICMP is received, update our idea of the MSS
|
|
|
|
|
* based on the new value in the route. Also nudge TCP to send something,
|
|
|
|
|
* since we know the packet we just sent was dropped.
|
1995-09-20 21:00:59 +00:00
|
|
|
* This duplicates some code in the tcp_mss() function in tcp_input.c.
|
1995-09-18 15:51:40 +00:00
|
|
|
*/
|
2002-06-14 08:35:21 +00:00
|
|
|
struct inpcb *
|
2006-04-03 12:59:27 +00:00
|
|
|
tcp_mtudisc(struct inpcb *inp, int errno)
|
1995-09-18 15:51:40 +00:00
|
|
|
{
|
2006-04-03 13:33:55 +00:00
|
|
|
struct tcpcb *tp;
|
2008-09-07 18:50:25 +00:00
|
|
|
struct socket *so;
|
1995-09-18 15:51:40 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2009-03-15 09:58:31 +00:00
|
|
|
if ((inp->inp_flags & INP_TIMEWAIT) ||
|
|
|
|
|
(inp->inp_flags & INP_DROPPED))
|
2006-04-03 14:07:50 +00:00
|
|
|
return (inp);
|
|
|
|
|
|
2006-04-03 13:33:55 +00:00
|
|
|
tp = intotcpcb(inp);
|
2006-04-03 14:07:50 +00:00
|
|
|
KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL"));
|
|
|
|
|
|
2008-11-06 13:25:59 +00:00
|
|
|
tcp_mss_update(tp, -1, NULL, NULL);
|
2008-09-07 18:50:25 +00:00
|
|
|
|
|
|
|
|
so = inp->inp_socket;
|
|
|
|
|
SOCKBUF_LOCK(&so->so_snd);
|
|
|
|
|
/* If the mss is larger than the socket buffer, decrease the mss. */
|
|
|
|
|
if (so->so_snd.sb_hiwat < tp->t_maxseg)
|
|
|
|
|
tp->t_maxseg = so->so_snd.sb_hiwat;
|
|
|
|
|
SOCKBUF_UNLOCK(&so->so_snd);
|
1995-09-20 21:00:59 +00:00
|
|
|
|
2009-04-11 22:07:19 +00:00
|
|
|
TCPSTAT_INC(tcps_mturesent);
|
2006-04-03 14:07:50 +00:00
|
|
|
tp->t_rtttime = 0;
|
|
|
|
|
tp->snd_nxt = tp->snd_una;
|
2006-08-26 17:53:19 +00:00
|
|
|
tcp_free_sackholes(tp);
|
|
|
|
|
tp->snd_recover = tp->snd_max;
|
2007-05-06 15:56:31 +00:00
|
|
|
if (tp->t_flags & TF_SACK_PERMIT)
|
2010-11-12 06:41:55 +00:00
|
|
|
EXIT_FASTRECOVERY(tp->t_flags);
|
2007-12-18 22:59:07 +00:00
|
|
|
tcp_output_send(tp);
|
2004-12-23 01:34:26 +00:00
|
|
|
return (inp);
|
1995-09-18 15:51:40 +00:00
|
|
|
}
|
|
|
|
|
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
1995-02-09 23:13:27 +00:00
|
|
|
/*
|
|
|
|
|
* Look-up the routing entry to the peer of this inpcb. If no route
|
2008-11-06 12:59:00 +00:00
|
|
|
* is found and it cannot be allocated, then return 0. This routine
|
|
|
|
|
* is called by TCP routines that access the rmx structure and by
|
|
|
|
|
* tcp_mss_update to get the peer/interface MTU.
|
1995-02-09 23:13:27 +00:00
|
|
|
*/
|
2004-08-16 18:32:07 +00:00
|
|
|
u_long
|
2006-09-06 21:51:59 +00:00
|
|
|
tcp_maxmtu(struct in_conninfo *inc, int *flags)
|
1995-02-09 23:13:27 +00:00
|
|
|
{
|
2003-11-20 20:07:39 +00:00
|
|
|
struct route sro;
|
|
|
|
|
struct sockaddr_in *dst;
|
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
u_long maxmtu = 0;
|
|
|
|
|
|
|
|
|
|
KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer"));
|
|
|
|
|
|
2003-11-26 20:31:13 +00:00
|
|
|
bzero(&sro, sizeof(sro));
|
2003-11-20 20:07:39 +00:00
|
|
|
if (inc->inc_faddr.s_addr != INADDR_ANY) {
|
|
|
|
|
dst = (struct sockaddr_in *)&sro.ro_dst;
|
|
|
|
|
dst->sin_family = AF_INET;
|
|
|
|
|
dst->sin_len = sizeof(*dst);
|
|
|
|
|
dst->sin_addr = inc->inc_faddr;
|
This main goals of this project are:
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,
The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.
Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:
- Kip Macy revised the locking code completely, thus completing
the last piece of the puzzle, Kip has also been conducting
active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
me maintaining that branch before the svn conversion
2008-12-15 06:10:57 +00:00
|
|
|
in_rtalloc_ign(&sro, 0, inc->inc_fibnum);
|
2003-11-20 20:07:39 +00:00
|
|
|
}
|
|
|
|
|
if (sro.ro_rt != NULL) {
|
|
|
|
|
ifp = sro.ro_rt->rt_ifp;
|
|
|
|
|
if (sro.ro_rt->rt_rmx.rmx_mtu == 0)
|
|
|
|
|
maxmtu = ifp->if_mtu;
|
|
|
|
|
else
|
|
|
|
|
maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu);
|
2006-09-06 21:51:59 +00:00
|
|
|
|
|
|
|
|
/* Report additional interface capabilities. */
|
|
|
|
|
if (flags != NULL) {
|
|
|
|
|
if (ifp->if_capenable & IFCAP_TSO4 &&
|
|
|
|
|
ifp->if_hwassist & CSUM_TSO)
|
|
|
|
|
*flags |= CSUM_TSO;
|
|
|
|
|
}
|
2003-11-20 20:07:39 +00:00
|
|
|
RTFREE(sro.ro_rt);
|
1995-02-09 23:13:27 +00:00
|
|
|
}
|
2003-11-20 20:07:39 +00:00
|
|
|
return (maxmtu);
|
1995-02-09 23:13:27 +00:00
|
|
|
}
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
1995-02-09 23:13:27 +00:00
|
|
|
|
2000-01-09 19:17:30 +00:00
|
|
|
#ifdef INET6
|
2003-11-20 20:07:39 +00:00
|
|
|
u_long
|
2006-09-06 21:51:59 +00:00
|
|
|
tcp_maxmtu6(struct in_conninfo *inc, int *flags)
|
2000-01-09 19:17:30 +00:00
|
|
|
{
|
2003-11-20 20:07:39 +00:00
|
|
|
struct route_in6 sro6;
|
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
u_long maxmtu = 0;
|
|
|
|
|
|
|
|
|
|
KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer"));
|
|
|
|
|
|
2003-11-26 20:31:13 +00:00
|
|
|
bzero(&sro6, sizeof(sro6));
|
2003-11-20 20:07:39 +00:00
|
|
|
if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) {
|
|
|
|
|
sro6.ro_dst.sin6_family = AF_INET6;
|
|
|
|
|
sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
|
|
|
|
|
sro6.ro_dst.sin6_addr = inc->inc6_faddr;
|
2012-02-03 13:08:44 +00:00
|
|
|
in6_rtalloc_ign(&sro6, 0, inc->inc_fibnum);
|
2000-01-09 19:17:30 +00:00
|
|
|
}
|
2003-11-20 20:07:39 +00:00
|
|
|
if (sro6.ro_rt != NULL) {
|
|
|
|
|
ifp = sro6.ro_rt->rt_ifp;
|
|
|
|
|
if (sro6.ro_rt->rt_rmx.rmx_mtu == 0)
|
|
|
|
|
maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp);
|
|
|
|
|
else
|
|
|
|
|
maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu,
|
|
|
|
|
IN6_LINKMTU(sro6.ro_rt->rt_ifp));
|
2006-09-06 21:51:59 +00:00
|
|
|
|
|
|
|
|
/* Report additional interface capabilities. */
|
|
|
|
|
if (flags != NULL) {
|
|
|
|
|
if (ifp->if_capenable & IFCAP_TSO6 &&
|
|
|
|
|
ifp->if_hwassist & CSUM_TSO)
|
|
|
|
|
*flags |= CSUM_TSO;
|
|
|
|
|
}
|
2003-11-20 20:07:39 +00:00
|
|
|
RTFREE(sro6.ro_rt);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (maxmtu);
|
2000-01-09 19:17:30 +00:00
|
|
|
}
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#ifdef IPSEC
|
2000-01-09 19:17:30 +00:00
|
|
|
/* compute ESP/AH header size for TCP, including outer IP header. */
|
|
|
|
|
size_t
|
2006-04-03 12:59:27 +00:00
|
|
|
ipsec_hdrsiz_tcp(struct tcpcb *tp)
|
2000-01-09 19:17:30 +00:00
|
|
|
{
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
struct mbuf *m;
|
|
|
|
|
size_t hdrsiz;
|
|
|
|
|
struct ip *ip;
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
struct ip6_hdr *ip6;
|
2003-02-19 22:32:43 +00:00
|
|
|
#endif
|
2000-01-09 19:17:30 +00:00
|
|
|
struct tcphdr *th;
|
|
|
|
|
|
2001-06-23 03:21:46 +00:00
|
|
|
if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL))
|
2004-12-23 01:34:26 +00:00
|
|
|
return (0);
|
2003-02-19 05:47:46 +00:00
|
|
|
MGETHDR(m, M_DONTWAIT, MT_DATA);
|
2000-01-09 19:17:30 +00:00
|
|
|
if (!m)
|
2004-12-23 01:34:26 +00:00
|
|
|
return (0);
|
2000-01-09 19:17:30 +00:00
|
|
|
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
if ((inp->inp_vflag & INP_IPV6) != 0) {
|
|
|
|
|
ip6 = mtod(m, struct ip6_hdr *);
|
|
|
|
|
th = (struct tcphdr *)(ip6 + 1);
|
|
|
|
|
m->m_pkthdr.len = m->m_len =
|
|
|
|
|
sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
|
2003-02-19 22:18:06 +00:00
|
|
|
tcpip_fillheaders(inp, ip6, th);
|
2009-02-08 09:27:07 +00:00
|
|
|
hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
|
2000-01-09 19:17:30 +00:00
|
|
|
} else
|
|
|
|
|
#endif /* INET6 */
|
2004-08-16 18:32:07 +00:00
|
|
|
{
|
|
|
|
|
ip = mtod(m, struct ip *);
|
|
|
|
|
th = (struct tcphdr *)(ip + 1);
|
|
|
|
|
m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr);
|
|
|
|
|
tcpip_fillheaders(inp, ip, th);
|
2009-02-08 09:27:07 +00:00
|
|
|
hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
|
2004-08-16 18:32:07 +00:00
|
|
|
}
|
2000-01-09 19:17:30 +00:00
|
|
|
|
|
|
|
|
m_free(m);
|
2004-12-23 01:34:26 +00:00
|
|
|
return (hdrsiz);
|
2000-01-09 19:17:30 +00:00
|
|
|
}
|
2007-07-03 12:13:45 +00:00
|
|
|
#endif /* IPSEC */
|
2000-01-09 19:17:30 +00:00
|
|
|
|
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
|
2004-02-13 18:21:45 +00:00
|
|
|
/*
|
|
|
|
|
* Callback function invoked by m_apply() to digest TCP segment data
|
|
|
|
|
* contained within an mbuf chain.
|
|
|
|
|
*/
|
|
|
|
|
static int
|
|
|
|
|
tcp_signature_apply(void *fstate, void *data, u_int len)
|
|
|
|
|
{
|
|
|
|
|
|
2004-02-14 21:49:48 +00:00
|
|
|
MD5Update(fstate, (u_char *)data, len);
|
2004-02-13 18:21:45 +00:00
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
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
|
|
|
/*
|
2008-09-13 17:26:46 +00:00
|
|
|
* Compute TCP-MD5 hash of a TCP segment. (RFC2385)
|
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
|
|
|
*
|
2004-02-12 20:12:48 +00:00
|
|
|
* Parameters:
|
|
|
|
|
* m pointer to head of mbuf chain
|
2008-09-13 17:26:46 +00:00
|
|
|
* _unused
|
2004-02-12 20:12:48 +00:00
|
|
|
* len length of TCP segment data, excluding options
|
|
|
|
|
* optlen length of TCP segment options
|
|
|
|
|
* buf pointer to storage for computed MD5 digest
|
|
|
|
|
* direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND)
|
|
|
|
|
*
|
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
|
|
|
* We do this over ip, tcphdr, segment data, and the key in the SADB.
|
|
|
|
|
* When called from tcp_input(), we can be sure that th_sum has been
|
|
|
|
|
* zeroed out and verified already.
|
|
|
|
|
*
|
|
|
|
|
* Return 0 if successful, otherwise return -1.
|
|
|
|
|
*
|
|
|
|
|
* XXX The key is retrieved from the system's PF_KEY SADB, by keying a
|
|
|
|
|
* search with the destination IP address, and a 'magic SPI' to be
|
|
|
|
|
* determined by the application. This is hardcoded elsewhere to 1179
|
|
|
|
|
* right now. Another branch of this code exists which uses the SPD to
|
|
|
|
|
* specify per-application flows but it is unstable.
|
|
|
|
|
*/
|
|
|
|
|
int
|
2008-09-13 17:26:46 +00:00
|
|
|
tcp_signature_compute(struct mbuf *m, int _unused, int len, int optlen,
|
2004-02-12 20:12:48 +00:00
|
|
|
u_char *buf, u_int direction)
|
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
|
|
|
{
|
|
|
|
|
union sockaddr_union dst;
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
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
|
|
|
struct ippseudo ippseudo;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
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
|
|
|
MD5_CTX ctx;
|
|
|
|
|
int doff;
|
|
|
|
|
struct ip *ip;
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
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
|
|
|
struct ipovly *ipovly;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
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
|
|
|
struct secasvar *sav;
|
|
|
|
|
struct tcphdr *th;
|
2008-09-13 17:26:46 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
struct ip6_hdr *ip6;
|
|
|
|
|
struct in6_addr in6;
|
|
|
|
|
char ip6buf[INET6_ADDRSTRLEN];
|
|
|
|
|
uint32_t plen;
|
|
|
|
|
uint16_t nhdr;
|
|
|
|
|
#endif
|
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
|
|
|
u_short savecsum;
|
|
|
|
|
|
2004-02-12 20:12:48 +00:00
|
|
|
KASSERT(m != NULL, ("NULL mbuf chain"));
|
|
|
|
|
KASSERT(buf != NULL, ("NULL signature pointer"));
|
|
|
|
|
|
|
|
|
|
/* Extract the destination from the IP header in the mbuf. */
|
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
|
|
|
bzero(&dst, sizeof(union sockaddr_union));
|
2008-09-13 17:26:46 +00:00
|
|
|
ip = mtod(m, struct ip *);
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
ip6 = NULL; /* Make the compiler happy. */
|
|
|
|
|
#endif
|
|
|
|
|
switch (ip->ip_v) {
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2008-09-13 17:26:46 +00:00
|
|
|
case IPVERSION:
|
|
|
|
|
dst.sa.sa_len = sizeof(struct sockaddr_in);
|
|
|
|
|
dst.sa.sa_family = AF_INET;
|
|
|
|
|
dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ?
|
|
|
|
|
ip->ip_src : ip->ip_dst;
|
|
|
|
|
break;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2008-09-13 17:26:46 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
case (IPV6_VERSION >> 4):
|
|
|
|
|
ip6 = mtod(m, struct ip6_hdr *);
|
|
|
|
|
dst.sa.sa_len = sizeof(struct sockaddr_in6);
|
|
|
|
|
dst.sa.sa_family = AF_INET6;
|
|
|
|
|
dst.sin6.sin6_addr = (direction == IPSEC_DIR_INBOUND) ?
|
|
|
|
|
ip6->ip6_src : ip6->ip6_dst;
|
|
|
|
|
break;
|
|
|
|
|
#endif
|
|
|
|
|
default:
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
/* NOTREACHED */
|
|
|
|
|
break;
|
|
|
|
|
}
|
2004-02-12 20:12:48 +00:00
|
|
|
|
|
|
|
|
/* Look up an SADB entry which matches the address of the peer. */
|
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
|
|
|
sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI));
|
|
|
|
|
if (sav == NULL) {
|
2008-09-13 17:26:46 +00:00
|
|
|
ipseclog((LOG_ERR, "%s: SADB lookup failed for %s\n", __func__,
|
|
|
|
|
(ip->ip_v == IPVERSION) ? inet_ntoa(dst.sin.sin_addr) :
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
(ip->ip_v == (IPV6_VERSION >> 4)) ?
|
|
|
|
|
ip6_sprintf(ip6buf, &dst.sin6.sin6_addr) :
|
|
|
|
|
#endif
|
|
|
|
|
"(unsupported)"));
|
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
|
|
|
return (EINVAL);
|
|
|
|
|
}
|
|
|
|
|
|
2004-02-12 20:12:48 +00:00
|
|
|
MD5Init(&ctx);
|
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
|
|
|
/*
|
2008-09-13 17:26:46 +00:00
|
|
|
* Step 1: Update MD5 hash with IP(v6) pseudo-header.
|
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
|
|
|
*
|
|
|
|
|
* XXX The ippseudo header MUST be digested in network byte order,
|
|
|
|
|
* or else we'll fail the regression test. Assume all fields we've
|
|
|
|
|
* been doing arithmetic on have been in host byte order.
|
|
|
|
|
* XXX One cannot depend on ipovly->ih_len here. When called from
|
|
|
|
|
* tcp_output(), the underlying ip_len member has not yet been set.
|
|
|
|
|
*/
|
2008-09-13 17:26:46 +00:00
|
|
|
switch (ip->ip_v) {
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2008-09-13 17:26:46 +00:00
|
|
|
case IPVERSION:
|
|
|
|
|
ipovly = (struct ipovly *)ip;
|
|
|
|
|
ippseudo.ippseudo_src = ipovly->ih_src;
|
|
|
|
|
ippseudo.ippseudo_dst = ipovly->ih_dst;
|
|
|
|
|
ippseudo.ippseudo_pad = 0;
|
|
|
|
|
ippseudo.ippseudo_p = IPPROTO_TCP;
|
|
|
|
|
ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) +
|
|
|
|
|
optlen);
|
|
|
|
|
MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo));
|
|
|
|
|
|
|
|
|
|
th = (struct tcphdr *)((u_char *)ip + sizeof(struct ip));
|
|
|
|
|
doff = sizeof(struct ip) + sizeof(struct tcphdr) + optlen;
|
|
|
|
|
break;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2008-09-13 17:26:46 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
/*
|
|
|
|
|
* RFC 2385, 2.0 Proposal
|
|
|
|
|
* For IPv6, the pseudo-header is as described in RFC 2460, namely the
|
|
|
|
|
* 128-bit source IPv6 address, 128-bit destination IPv6 address, zero-
|
|
|
|
|
* extended next header value (to form 32 bits), and 32-bit segment
|
|
|
|
|
* length.
|
|
|
|
|
* Note: Upper-Layer Packet Length comes before Next Header.
|
|
|
|
|
*/
|
|
|
|
|
case (IPV6_VERSION >> 4):
|
|
|
|
|
in6 = ip6->ip6_src;
|
|
|
|
|
in6_clearscope(&in6);
|
|
|
|
|
MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr));
|
|
|
|
|
in6 = ip6->ip6_dst;
|
|
|
|
|
in6_clearscope(&in6);
|
|
|
|
|
MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr));
|
|
|
|
|
plen = htonl(len + sizeof(struct tcphdr) + optlen);
|
|
|
|
|
MD5Update(&ctx, (char *)&plen, sizeof(uint32_t));
|
|
|
|
|
nhdr = 0;
|
|
|
|
|
MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t));
|
|
|
|
|
MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t));
|
|
|
|
|
MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t));
|
|
|
|
|
nhdr = IPPROTO_TCP;
|
|
|
|
|
MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t));
|
|
|
|
|
|
|
|
|
|
th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr));
|
|
|
|
|
doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen;
|
|
|
|
|
break;
|
|
|
|
|
#endif
|
|
|
|
|
default:
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
/* NOTREACHED */
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
2004-02-12 20:12:48 +00:00
|
|
|
|
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
|
|
|
/*
|
|
|
|
|
* Step 2: Update MD5 hash with TCP header, excluding options.
|
|
|
|
|
* The TCP checksum must be set to zero.
|
|
|
|
|
*/
|
|
|
|
|
savecsum = th->th_sum;
|
|
|
|
|
th->th_sum = 0;
|
|
|
|
|
MD5Update(&ctx, (char *)th, sizeof(struct tcphdr));
|
|
|
|
|
th->th_sum = savecsum;
|
2004-02-12 20:12:48 +00:00
|
|
|
|
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
|
|
|
/*
|
|
|
|
|
* Step 3: Update MD5 hash with TCP segment data.
|
|
|
|
|
* Use m_apply() to avoid an early m_pullup().
|
|
|
|
|
*/
|
|
|
|
|
if (len > 0)
|
2004-02-13 18:21:45 +00:00
|
|
|
m_apply(m, doff, len, tcp_signature_apply, &ctx);
|
2004-02-12 20:12:48 +00:00
|
|
|
|
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
|
|
|
/*
|
|
|
|
|
* Step 4: Update MD5 hash with shared secret.
|
|
|
|
|
*/
|
2007-11-28 13:23:50 +00:00
|
|
|
MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth));
|
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
|
|
|
MD5Final(buf, &ctx);
|
2004-02-12 20:12:48 +00:00
|
|
|
|
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
|
|
|
key_sa_recordxfer(sav, m);
|
|
|
|
|
KEY_FREESAV(&sav);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
2011-04-25 17:13:40 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Verify the TCP-MD5 hash of a TCP segment. (RFC2385)
|
|
|
|
|
*
|
|
|
|
|
* Parameters:
|
|
|
|
|
* m pointer to head of mbuf chain
|
|
|
|
|
* len length of TCP segment data, excluding options
|
|
|
|
|
* optlen length of TCP segment options
|
|
|
|
|
* buf pointer to storage for computed MD5 digest
|
|
|
|
|
* direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND)
|
|
|
|
|
*
|
|
|
|
|
* Return 1 if successful, otherwise return 0.
|
|
|
|
|
*/
|
|
|
|
|
int
|
|
|
|
|
tcp_signature_verify(struct mbuf *m, int off0, int tlen, int optlen,
|
|
|
|
|
struct tcpopt *to, struct tcphdr *th, u_int tcpbflag)
|
|
|
|
|
{
|
|
|
|
|
char tmpdigest[TCP_SIGLEN];
|
|
|
|
|
|
|
|
|
|
if (tcp_sig_checksigs == 0)
|
|
|
|
|
return (1);
|
|
|
|
|
if ((tcpbflag & TF_SIGNATURE) == 0) {
|
|
|
|
|
if ((to->to_flags & TOF_SIGNATURE) != 0) {
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If this socket is not expecting signature but
|
|
|
|
|
* the segment contains signature just fail.
|
|
|
|
|
*/
|
|
|
|
|
TCPSTAT_INC(tcps_sig_err_sigopt);
|
|
|
|
|
TCPSTAT_INC(tcps_sig_rcvbadsig);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Signature is not expected, and not present in segment. */
|
|
|
|
|
return (1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If this socket is expecting signature but the segment does not
|
|
|
|
|
* contain any just fail.
|
|
|
|
|
*/
|
|
|
|
|
if ((to->to_flags & TOF_SIGNATURE) == 0) {
|
|
|
|
|
TCPSTAT_INC(tcps_sig_err_nosigopt);
|
|
|
|
|
TCPSTAT_INC(tcps_sig_rcvbadsig);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
if (tcp_signature_compute(m, off0, tlen, optlen, &tmpdigest[0],
|
|
|
|
|
IPSEC_DIR_INBOUND) == -1) {
|
|
|
|
|
TCPSTAT_INC(tcps_sig_err_buildsig);
|
|
|
|
|
TCPSTAT_INC(tcps_sig_rcvbadsig);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (bcmp(to->to_signature, &tmpdigest[0], TCP_SIGLEN) != 0) {
|
|
|
|
|
TCPSTAT_INC(tcps_sig_rcvbadsig);
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
TCPSTAT_INC(tcps_sig_rcvgoodsig);
|
|
|
|
|
return (1);
|
|
|
|
|
}
|
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
|
|
|
#endif /* TCP_SIGNATURE */
|
2005-02-14 07:37:51 +00:00
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
sysctl_drop(SYSCTL_HANDLER_ARGS)
|
|
|
|
|
{
|
|
|
|
|
/* addrs[0] is a foreign socket, addrs[1] is a local one. */
|
|
|
|
|
struct sockaddr_storage addrs[2];
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
struct tcpcb *tp;
|
2005-10-02 08:43:57 +00:00
|
|
|
struct tcptw *tw;
|
2005-02-14 07:37:51 +00:00
|
|
|
struct sockaddr_in *fin, *lin;
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
struct sockaddr_in6 *fin6, *lin6;
|
|
|
|
|
#endif
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
inp = NULL;
|
|
|
|
|
fin = lin = NULL;
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
fin6 = lin6 = NULL;
|
|
|
|
|
#endif
|
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
|
|
if (req->oldptr != NULL || req->oldlen != 0)
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
if (req->newptr == NULL)
|
|
|
|
|
return (EPERM);
|
|
|
|
|
if (req->newlen < sizeof(addrs))
|
|
|
|
|
return (ENOMEM);
|
|
|
|
|
error = SYSCTL_IN(req, &addrs, sizeof(addrs));
|
|
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
|
|
|
|
|
switch (addrs[0].ss_family) {
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
case AF_INET6:
|
|
|
|
|
fin6 = (struct sockaddr_in6 *)&addrs[0];
|
|
|
|
|
lin6 = (struct sockaddr_in6 *)&addrs[1];
|
|
|
|
|
if (fin6->sin6_len != sizeof(struct sockaddr_in6) ||
|
|
|
|
|
lin6->sin6_len != sizeof(struct sockaddr_in6))
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) {
|
|
|
|
|
if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr))
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]);
|
|
|
|
|
in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]);
|
|
|
|
|
fin = (struct sockaddr_in *)&addrs[0];
|
|
|
|
|
lin = (struct sockaddr_in *)&addrs[1];
|
|
|
|
|
break;
|
|
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
error = sa6_embedscope(fin6, V_ip6_use_defzone);
|
2005-02-14 07:37:51 +00:00
|
|
|
if (error)
|
2005-07-25 12:31:43 +00:00
|
|
|
return (error);
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
error = sa6_embedscope(lin6, V_ip6_use_defzone);
|
2005-08-01 12:08:49 +00:00
|
|
|
if (error)
|
2005-07-25 12:31:43 +00:00
|
|
|
return (error);
|
2005-02-14 07:37:51 +00:00
|
|
|
break;
|
|
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2005-02-14 07:37:51 +00:00
|
|
|
case AF_INET:
|
|
|
|
|
fin = (struct sockaddr_in *)&addrs[0];
|
|
|
|
|
lin = (struct sockaddr_in *)&addrs[1];
|
|
|
|
|
if (fin->sin_len != sizeof(struct sockaddr_in) ||
|
|
|
|
|
lin->sin_len != sizeof(struct sockaddr_in))
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
break;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2005-02-14 07:37:51 +00:00
|
|
|
default:
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WLOCK(&V_tcbinfo);
|
2005-02-14 07:37:51 +00:00
|
|
|
switch (addrs[0].ss_family) {
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
case AF_INET6:
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in6_pcblookup(&V_tcbinfo, &fin6->sin6_addr,
|
|
|
|
|
fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port,
|
|
|
|
|
INPLOOKUP_WLOCKPCB, NULL);
|
2005-02-14 07:37:51 +00:00
|
|
|
break;
|
|
|
|
|
#endif
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
2005-02-14 07:37:51 +00:00
|
|
|
case AF_INET:
|
Decompose the current single inpcbinfo lock into two locks:
- The existing ipi_lock continues to protect the global inpcb list and
inpcb counter. This lock is now relegated to a small number of
allocation and free operations, and occasional operations that walk
all connections (including, awkwardly, certain UDP multicast receive
operations -- something to revisit).
- A new ipi_hash_lock protects the two inpcbinfo hash tables for
looking up connections and bound sockets, manipulated using new
INP_HASH_*() macros. This lock, combined with inpcb locks, protects
the 4-tuple address space.
Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb
connection locks, so may be acquired while manipulating a connection on
which a lock is already held, avoiding the need to acquire the inpcbinfo
lock preemptively when a binding change might later be required. As a
result, however, lookup operations necessarily go through a reference
acquire while holding the lookup lock, later acquiring an inpcb lock --
if required.
A new function in_pcblookup() looks up connections, and accepts flags
indicating how to return the inpcb. Due to lock order changes, callers
no longer need acquire locks before performing a lookup: the lookup
routine will acquire the ipi_hash_lock as needed. In the future, it will
also be able to use alternative lookup and locking strategies
transparently to callers, such as pcbgroup lookup. New lookup flags are,
supplementing the existing INPLOOKUP_WILDCARD flag:
INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb
INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb
Callers must pass exactly one of these flags (for the time being).
Some notes:
- All protocols are updated to work within the new regime; especially,
TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely
eliminated, and global hash lock hold times are dramatically reduced
compared to previous locking.
- The TCP syncache still relies on the pcbinfo lock, something that we
may want to revisit.
- Support for reverting to the FreeBSD 7.x locking strategy in TCP input
is no longer available -- hash lookup locks are now held only very
briefly during inpcb lookup, rather than for potentially extended
periods. However, the pcbinfo ipi_lock will still be acquired if a
connection state might change such that a connection is added or
removed.
- Raw IP sockets continue to use the pcbinfo ipi_lock for protection,
due to maintaining their own hash tables.
- The interface in6_pcblookup_hash_locked() is maintained, which allows
callers to acquire hash locks and perform one or more lookups atomically
with 4-tuple allocation: this is required only for TCPv6, as there is no
in6_pcbconnect_setup(), which there should be.
- UDPv6 locking remains significantly more conservative than UDPv4
locking, which relates to source address selection. This needs
attention, as it likely significantly reduces parallelism in this code
for multithreaded socket use (such as in BIND).
- In the UDPv4 and UDPv6 multicast cases, we need to revisit locking
somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which
is no longer sufficient. A second check once the inpcb lock is held
should do the trick, keeping the general case from requiring the inpcb
lock for every inpcb visited.
- This work reminds us that we need to revisit locking of the v4/v6 flags,
which may be accessed lock-free both before and after this change.
- Right now, a single lock name is used for the pcbhash lock -- this is
undesirable, and probably another argument is required to take care of
this (or a char array name field in the pcbinfo?).
This is not an MFC candidate for 8.x due to its impact on lookup and
locking semantics. It's possible some of these issues could be worked
around with compatibility wrappers, if necessary.
Reviewed by: bz
Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
|
|
|
inp = in_pcblookup(&V_tcbinfo, fin->sin_addr, fin->sin_port,
|
|
|
|
|
lin->sin_addr, lin->sin_port, INPLOOKUP_WLOCKPCB, NULL);
|
2005-02-14 07:37:51 +00:00
|
|
|
break;
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif
|
2005-02-14 07:37:51 +00:00
|
|
|
}
|
|
|
|
|
if (inp != NULL) {
|
2009-03-15 09:58:31 +00:00
|
|
|
if (inp->inp_flags & INP_TIMEWAIT) {
|
2006-04-04 12:26:07 +00:00
|
|
|
/*
|
|
|
|
|
* XXXRW: There currently exists a state where an
|
|
|
|
|
* inpcb is present, but its timewait state has been
|
|
|
|
|
* discarded. For now, don't allow dropping of this
|
|
|
|
|
* type of inpcb.
|
|
|
|
|
*/
|
2006-04-03 11:57:12 +00:00
|
|
|
tw = intotw(inp);
|
2006-04-04 12:26:07 +00:00
|
|
|
if (tw != NULL)
|
|
|
|
|
tcp_twclose(tw, 0);
|
2007-11-24 18:43:59 +00:00
|
|
|
else
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2009-03-15 09:58:31 +00:00
|
|
|
} else if (!(inp->inp_flags & INP_DROPPED) &&
|
2006-04-03 11:57:12 +00:00
|
|
|
!(inp->inp_socket->so_options & SO_ACCEPTCONN)) {
|
|
|
|
|
tp = intotcpcb(inp);
|
2007-11-24 18:43:59 +00:00
|
|
|
tp = tcp_drop(tp, ECONNABORTED);
|
|
|
|
|
if (tp != NULL)
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2007-11-24 18:43:59 +00:00
|
|
|
} else
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2005-02-14 07:37:51 +00:00
|
|
|
} else
|
|
|
|
|
error = ESRCH;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
INP_INFO_WUNLOCK(&V_tcbinfo);
|
2005-02-14 07:37:51 +00:00
|
|
|
return (error);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop,
|
|
|
|
|
CTLTYPE_STRUCT|CTLFLAG_WR|CTLFLAG_SKIP, NULL,
|
|
|
|
|
0, sysctl_drop, "", "Drop TCP connection");
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Generate a standardized TCP log line for use throughout the
|
|
|
|
|
* tcp subsystem. Memory allocation is done with M_NOWAIT to
|
|
|
|
|
* allow use in the interrupt context.
|
|
|
|
|
*
|
|
|
|
|
* NB: The caller MUST free(s, M_TCPLOG) the returned string.
|
|
|
|
|
* NB: The function may return NULL if memory allocation failed.
|
|
|
|
|
*
|
|
|
|
|
* Due to header inclusion and ordering limitations the struct ip
|
|
|
|
|
* and ip6_hdr pointers have to be passed as void pointers.
|
|
|
|
|
*/
|
2010-08-18 17:39:47 +00:00
|
|
|
char *
|
|
|
|
|
tcp_log_vain(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
|
|
|
|
|
const void *ip6hdr)
|
|
|
|
|
{
|
|
|
|
|
|
|
|
|
|
/* Is logging enabled? */
|
|
|
|
|
if (tcp_log_in_vain == 0)
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
return (tcp_log_addr(inc, th, ip4hdr, ip6hdr));
|
|
|
|
|
}
|
|
|
|
|
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
char *
|
|
|
|
|
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
|
2007-07-05 05:55:57 +00:00
|
|
|
const void *ip6hdr)
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
{
|
2010-08-18 17:39:47 +00:00
|
|
|
|
|
|
|
|
/* Is logging enabled? */
|
|
|
|
|
if (tcp_log_debug == 0)
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
return (tcp_log_addr(inc, th, ip4hdr, ip6hdr));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static char *
|
|
|
|
|
tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
|
|
|
|
|
const void *ip6hdr)
|
|
|
|
|
{
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
char *s, *sp;
|
|
|
|
|
size_t size;
|
|
|
|
|
struct ip *ip;
|
|
|
|
|
#ifdef INET6
|
2007-07-05 06:04:46 +00:00
|
|
|
const struct ip6_hdr *ip6;
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
|
2007-07-05 06:04:46 +00:00
|
|
|
ip6 = (const struct ip6_hdr *)ip6hdr;
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
#endif /* INET6 */
|
|
|
|
|
ip = (struct ip *)ip4hdr;
|
|
|
|
|
|
|
|
|
|
/*
|
2007-05-23 19:07:53 +00:00
|
|
|
* The log line looks like this:
|
|
|
|
|
* "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags 0x2<SYN>"
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
*/
|
2007-05-23 19:07:53 +00:00
|
|
|
size = sizeof("TCP: []:12345 to []:12345 tcpflags 0x2<>") +
|
|
|
|
|
sizeof(PRINT_TH_FLAGS) + 1 +
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
#ifdef INET6
|
2007-05-23 19:07:53 +00:00
|
|
|
2 * INET6_ADDRSTRLEN;
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
#else
|
2007-05-23 19:07:53 +00:00
|
|
|
2 * INET_ADDRSTRLEN;
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
#endif /* INET6 */
|
|
|
|
|
|
2007-05-27 17:02:54 +00:00
|
|
|
s = malloc(size, M_TCPLOG, M_ZERO|M_NOWAIT);
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
if (s == NULL)
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
strcat(s, "TCP: [");
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
|
2008-12-17 12:52:34 +00:00
|
|
|
if (inc && ((inc->inc_flags & INC_ISIPV6) == 0)) {
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
inet_ntoa_r(inc->inc_faddr, sp);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i to [", ntohs(inc->inc_fport));
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
inet_ntoa_r(inc->inc_laddr, sp);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i", ntohs(inc->inc_lport));
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
} else if (inc) {
|
|
|
|
|
ip6_sprintf(sp, &inc->inc6_faddr);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i to [", ntohs(inc->inc_fport));
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
ip6_sprintf(sp, &inc->inc6_laddr);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i", ntohs(inc->inc_lport));
|
|
|
|
|
} else if (ip6 && th) {
|
|
|
|
|
ip6_sprintf(sp, &ip6->ip6_src);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i to [", ntohs(th->th_sport));
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
ip6_sprintf(sp, &ip6->ip6_dst);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i", ntohs(th->th_dport));
|
|
|
|
|
#endif /* INET6 */
|
2011-04-30 11:21:29 +00:00
|
|
|
#ifdef INET
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
} else if (ip && th) {
|
|
|
|
|
inet_ntoa_r(ip->ip_src, sp);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i to [", ntohs(th->th_sport));
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
inet_ntoa_r(ip->ip_dst, sp);
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
sprintf(sp, "]:%i", ntohs(th->th_dport));
|
2011-04-30 11:21:29 +00:00
|
|
|
#endif /* INET */
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
} else {
|
|
|
|
|
free(s, M_TCPLOG);
|
|
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
sp = s + strlen(s);
|
|
|
|
|
if (th)
|
|
|
|
|
sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS);
|
2007-05-23 19:07:53 +00:00
|
|
|
if (*(s + size - 1) != '\0')
|
Add tcp_log_addrs() function to generate and standardized TCP log line
for use thoughout the tcp subsystem.
It is IPv4 and IPv6 aware creates a line in the following format:
"TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags <RST>"
A "\n" is not included at the end. The caller is supposed to add
further information after the standard tcp log header.
The function returns a NUL terminated string which the caller has
to free(s, M_TCPLOG) after use. All memory allocation is done
with M_NOWAIT and the return value may be NULL in memory shortage
situations.
Either struct in_conninfo || (struct tcphdr && (struct ip || struct
ip6_hdr) have to be supplied.
Due to ip[6].h header inclusion limitations and ordering issues the
struct ip and struct ip6_hdr parameters have to be casted and passed
as void * pointers.
tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr,
void *ip6hdr)
Usage example:
struct ip *ip;
char *tcplog;
if (tcplog = tcp_log_addrs(NULL, th, (void *)ip, NULL)) {
log(LOG_DEBUG, "%s; %s: Connection attempt to closed port\n",
tcplog, __func__);
free(s, M_TCPLOG);
}
2007-05-18 19:58:37 +00:00
|
|
|
panic("%s: string too long", __func__);
|
|
|
|
|
return (s);
|
|
|
|
|
}
|
