2005-01-07 01:45:51 +00:00
|
|
|
/*-
|
1995-09-21 17:55:49 +00:00
|
|
|
* Copyright (c) 1982, 1986, 1991, 1993, 1995
|
2007-02-17 21:02:38 +00:00
|
|
|
* The Regents of the University of California.
|
|
|
|
|
* Copyright (c) 2007 Robert N. M. Watson
|
|
|
|
|
* All rights reserved.
|
1994-05-24 10:09:53 +00:00
|
|
|
*
|
|
|
|
|
* 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.
|
|
|
|
|
*
|
1995-09-21 17:55:49 +00:00
|
|
|
* @(#)in_pcb.c 8.4 (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("$FreeBSD$");
|
|
|
|
|
|
2007-02-17 21:02:38 +00:00
|
|
|
#include "opt_ddb.h"
|
1999-12-22 19:13:38 +00:00
|
|
|
#include "opt_ipsec.h"
|
1999-12-07 17:39:16 +00:00
|
|
|
#include "opt_inet6.h"
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#include "opt_mac.h"
|
1999-12-07 17:39:16 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/param.h>
|
|
|
|
|
#include <sys/systm.h>
|
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
|
#include <sys/mbuf.h>
|
1999-12-07 17:39:16 +00:00
|
|
|
#include <sys/domain.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/protosw.h>
|
|
|
|
|
#include <sys/socket.h>
|
|
|
|
|
#include <sys/socketvar.h>
|
2006-11-06 13:42:10 +00:00
|
|
|
#include <sys/priv.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/proc.h>
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
#include <sys/jail.h>
|
1996-01-19 08:00:58 +00:00
|
|
|
#include <sys/kernel.h>
|
|
|
|
|
#include <sys/sysctl.h>
|
1998-03-28 10:18:26 +00:00
|
|
|
|
2007-02-17 21:02:38 +00:00
|
|
|
#ifdef DDB
|
|
|
|
|
#include <ddb/ddb.h>
|
|
|
|
|
#endif
|
|
|
|
|
|
2002-03-20 05:48:55 +00:00
|
|
|
#include <vm/uma.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
|
#include <net/if.h>
|
1999-12-07 17:39:16 +00:00
|
|
|
#include <net/if_types.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <net/route.h>
|
|
|
|
|
|
|
|
|
|
#include <netinet/in.h>
|
|
|
|
|
#include <netinet/in_pcb.h>
|
|
|
|
|
#include <netinet/in_var.h>
|
|
|
|
|
#include <netinet/ip_var.h>
|
2003-02-19 22:32:43 +00:00
|
|
|
#include <netinet/tcp_var.h>
|
2005-01-02 01:50:57 +00:00
|
|
|
#include <netinet/udp.h>
|
|
|
|
|
#include <netinet/udp_var.h>
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
#include <netinet/ip6.h>
|
|
|
|
|
#include <netinet6/ip6_var.h>
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#ifdef IPSEC
|
2002-10-16 02:25:05 +00:00
|
|
|
#include <netipsec/ipsec.h>
|
|
|
|
|
#include <netipsec/key.h>
|
2007-07-03 12:13:45 +00:00
|
|
|
#endif /* IPSEC */
|
2002-10-16 02:25:05 +00:00
|
|
|
|
2006-10-22 11:52:19 +00:00
|
|
|
#include <security/mac/mac_framework.h>
|
|
|
|
|
|
1996-01-19 08:00:58 +00:00
|
|
|
/*
|
|
|
|
|
* These configure the range of local port addresses assigned to
|
|
|
|
|
* "unspecified" outgoing connections/packets/whatever.
|
|
|
|
|
*/
|
1999-11-22 02:45:11 +00:00
|
|
|
int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
|
|
|
|
|
int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
|
2008-03-04 19:16:21 +00:00
|
|
|
int ipport_firstauto = IPPORT_EPHEMERALFIRST; /* 10000 */
|
|
|
|
|
int ipport_lastauto = IPPORT_EPHEMERALLAST; /* 65535 */
|
1999-11-22 02:45:11 +00:00
|
|
|
int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
|
|
|
|
|
int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
|
1996-01-19 08:00:58 +00:00
|
|
|
|
The ancient and outdated concept of "privileged ports" in UNIX-type
OSes has probably caused more problems than it ever solved. Allow the
user to retire the old behavior by specifying their own privileged
range with,
net.inet.ip.portrange.reservedhigh default = IPPORT_RESERVED - 1
net.inet.ip.portrange.reservedlo default = 0
Now you can run that webserver without ever needing root at all. Or
just imagine, an ftpd that can really drop privileges, rather than
just set the euid, and still do PORT data transfers from 20/tcp.
Two edge cases to note,
# sysctl net.inet.ip.portrange.reservedhigh=0
Opens all ports to everyone, and,
# sysctl net.inet.ip.portrange.reservedhigh=65535
Locks all network activity to root only (which could actually have
been achieved before with ipfw(8), but is somewhat more
complicated).
For those who stick to the old religion that 0-1023 belong to root and
root alone, don't touch the knobs (or even lock them by raising
securelevel(8)), and nothing changes.
2003-02-21 05:28:27 +00:00
|
|
|
/*
|
|
|
|
|
* Reserved ports accessible only to root. There are significant
|
|
|
|
|
* security considerations that must be accounted for when changing these,
|
|
|
|
|
* but the security benefits can be great. Please be careful.
|
|
|
|
|
*/
|
|
|
|
|
int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
|
|
|
|
|
int ipport_reservedlow = 0;
|
|
|
|
|
|
2005-01-02 01:50:57 +00:00
|
|
|
/* Variables dealing with random ephemeral port allocation. */
|
|
|
|
|
int ipport_randomized = 1; /* user controlled via sysctl */
|
|
|
|
|
int ipport_randomcps = 10; /* user controlled via sysctl */
|
|
|
|
|
int ipport_randomtime = 45; /* user controlled via sysctl */
|
|
|
|
|
int ipport_stoprandom = 0; /* toggled by ipport_tick */
|
|
|
|
|
int ipport_tcpallocs;
|
|
|
|
|
int ipport_tcplastcount;
|
2004-04-22 08:32:14 +00:00
|
|
|
|
1996-08-12 14:05:54 +00:00
|
|
|
#define RANGECHK(var, min, max) \
|
|
|
|
|
if ((var) < (min)) { (var) = (min); } \
|
|
|
|
|
else if ((var) > (max)) { (var) = (max); }
|
|
|
|
|
|
|
|
|
|
static int
|
2000-07-04 11:25:35 +00:00
|
|
|
sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
|
1996-08-12 14:05:54 +00:00
|
|
|
{
|
2004-04-06 10:59:11 +00:00
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
|
|
|
|
|
if (error == 0) {
|
1996-08-12 14:05:54 +00:00
|
|
|
RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
|
|
|
|
|
RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
|
2004-04-06 10:59:11 +00:00
|
|
|
RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
|
|
|
|
|
RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
|
|
|
|
|
RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
|
|
|
|
|
RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
|
1996-08-12 14:05:54 +00:00
|
|
|
}
|
2004-04-06 10:59:11 +00:00
|
|
|
return (error);
|
1996-08-12 14:05:54 +00:00
|
|
|
}
|
1996-02-22 21:32:23 +00:00
|
|
|
|
1996-08-12 14:05:54 +00:00
|
|
|
#undef RANGECHK
|
1996-01-19 08:00:58 +00:00
|
|
|
|
1996-08-12 14:05:54 +00:00
|
|
|
SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
|
|
|
|
|
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
|
|
|
|
|
SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
|
|
|
|
|
&ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
|
The ancient and outdated concept of "privileged ports" in UNIX-type
OSes has probably caused more problems than it ever solved. Allow the
user to retire the old behavior by specifying their own privileged
range with,
net.inet.ip.portrange.reservedhigh default = IPPORT_RESERVED - 1
net.inet.ip.portrange.reservedlo default = 0
Now you can run that webserver without ever needing root at all. Or
just imagine, an ftpd that can really drop privileges, rather than
just set the euid, and still do PORT data transfers from 20/tcp.
Two edge cases to note,
# sysctl net.inet.ip.portrange.reservedhigh=0
Opens all ports to everyone, and,
# sysctl net.inet.ip.portrange.reservedhigh=65535
Locks all network activity to root only (which could actually have
been achieved before with ipfw(8), but is somewhat more
complicated).
For those who stick to the old religion that 0-1023 belong to root and
root alone, don't touch the knobs (or even lock them by raising
securelevel(8)), and nothing changes.
2003-02-21 05:28:27 +00:00
|
|
|
SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
|
|
|
|
|
CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
|
|
|
|
|
SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
|
|
|
|
|
CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
|
2005-03-23 09:26:38 +00:00
|
|
|
SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
|
|
|
|
|
&ipport_randomized, 0, "Enable random port allocation");
|
|
|
|
|
SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
|
|
|
|
|
&ipport_randomcps, 0, "Maximum number of random port "
|
|
|
|
|
"allocations before switching to a sequental one");
|
|
|
|
|
SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
|
|
|
|
|
&ipport_randomtime, 0, "Minimum time to keep sequental port "
|
|
|
|
|
"allocation before switching to a random one");
|
1995-11-14 20:34:56 +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
|
|
|
/*
|
|
|
|
|
* in_pcb.c: manage the Protocol Control Blocks.
|
|
|
|
|
*
|
2005-07-19 12:24:27 +00:00
|
|
|
* NOTE: It is assumed that most of these functions will be called with
|
|
|
|
|
* the pcbinfo lock held, and often, the inpcb lock held, as these utility
|
|
|
|
|
* functions often modify hash chains or addresses in pcbs.
|
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
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Allocate a PCB and associate it with the socket.
|
2006-07-18 22:34:27 +00:00
|
|
|
* On success return with the PCB locked.
|
Improved connection establishment performance by doing local port lookups via
a hashed port list. In the new scheme, in_pcblookup() goes away and is
replaced by a new routine, in_pcblookup_local() for doing the local port
check. Note that this implementation is space inefficient in that the PCB
struct is now too large to fit into 128 bytes. I might deal with this in the
future by using the new zone allocator, but I wanted these changes to be
extensively tested in their current form first.
Also:
1) Fixed off-by-one errors in the port lookup loops in in_pcbbind().
2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash()
to do the initialial hash insertion.
3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability.
4) Added a new routine, in_pcbremlists() to remove the PCB from the various
hash lists.
5) Added/deleted comments where appropriate.
6) Removed unnecessary splnet() locking. In general, the PCB functions should
be called at splnet()...there are unfortunately a few exceptions, however.
7) Reorganized a few structs for better cache line behavior.
8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in
the future, however.
These changes have been tested on wcarchive for more than a month. In tests
done here, connection establishment overhead is reduced by more than 50
times, thus getting rid of one of the major networking scalability problems.
Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a
large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult.
WARNING: Anything that knows about inpcb and tcpcb structs will have to be
recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
int
|
2006-07-18 22:34:27 +00:00
|
|
|
in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-01-22 01:16:25 +00:00
|
|
|
struct inpcb *inp;
|
2001-07-26 19:19:49 +00:00
|
|
|
int error;
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
|
2003-11-08 23:02:36 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(pcbinfo);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
error = 0;
|
2006-07-18 22:34:27 +00:00
|
|
|
inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (inp == NULL)
|
|
|
|
|
return (ENOBUFS);
|
2006-12-29 14:58:18 +00:00
|
|
|
bzero(inp, inp_zero_size);
|
1995-04-09 01:29:31 +00:00
|
|
|
inp->inp_pcbinfo = pcbinfo;
|
1994-05-24 10:09:53 +00:00
|
|
|
inp->inp_socket = so;
|
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
|
|
|
inp->inp_inc.inc_fibnum = so->so_fibnum;
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#ifdef MAC
|
2007-10-24 19:04:04 +00:00
|
|
|
error = mac_inpcb_init(inp, M_NOWAIT);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
if (error != 0)
|
|
|
|
|
goto out;
|
2004-06-13 02:50:07 +00:00
|
|
|
SOCK_LOCK(so);
|
2007-10-24 19:04:04 +00:00
|
|
|
mac_inpcb_create(so, inp);
|
2004-06-13 02:50:07 +00:00
|
|
|
SOCK_UNLOCK(so);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#endif
|
2007-07-01 11:41:27 +00:00
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#ifdef IPSEC
|
2001-07-26 19:19:49 +00:00
|
|
|
error = ipsec_init_policy(so, &inp->inp_sp);
|
2007-12-22 10:06:11 +00:00
|
|
|
if (error != 0) {
|
|
|
|
|
#ifdef MAC
|
|
|
|
|
mac_inpcb_destroy(inp);
|
|
|
|
|
#endif
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
goto out;
|
2008-03-17 13:04:56 +00:00
|
|
|
}
|
2007-07-03 12:13:45 +00:00
|
|
|
#endif /*IPSEC*/
|
2006-11-30 10:54:54 +00:00
|
|
|
#ifdef INET6
|
2003-02-19 22:32:43 +00:00
|
|
|
if (INP_SOCKAF(so) == AF_INET6) {
|
|
|
|
|
inp->inp_vflag |= INP_IPV6PROTO;
|
|
|
|
|
if (ip6_v6only)
|
|
|
|
|
inp->inp_flags |= IN6P_IPV6_V6ONLY;
|
|
|
|
|
}
|
2000-04-02 03:49:25 +00:00
|
|
|
#endif
|
2007-04-30 23:12:05 +00:00
|
|
|
LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
|
1998-03-24 18:06:34 +00:00
|
|
|
pcbinfo->ipi_count++;
|
1994-05-24 10:09:53 +00:00
|
|
|
so->so_pcb = (caddr_t)inp;
|
2001-06-11 12:39:29 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (ip6_auto_flowlabel)
|
|
|
|
|
inp->inp_flags |= IN6P_AUTOFLOWLABEL;
|
|
|
|
|
#endif
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inp);
|
2006-07-18 22:34:27 +00:00
|
|
|
inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
|
|
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#if defined(IPSEC) || defined(MAC)
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
out:
|
|
|
|
|
if (error != 0)
|
|
|
|
|
uma_zfree(pcbinfo->ipi_zone, inp);
|
|
|
|
|
#endif
|
|
|
|
|
return (error);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2002-10-20 21:44:31 +00:00
|
|
|
int anonport, error;
|
|
|
|
|
|
2003-11-13 05:16:56 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2003-11-08 23:02:36 +00:00
|
|
|
|
2002-10-20 21:44:31 +00:00
|
|
|
if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
anonport = inp->inp_lport == 0 && (nam == NULL ||
|
|
|
|
|
((struct sockaddr_in *)nam)->sin_port == 0);
|
|
|
|
|
error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
|
2004-03-27 21:05:46 +00:00
|
|
|
&inp->inp_lport, cred);
|
2002-10-20 21:44:31 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
if (in_pcbinshash(inp) != 0) {
|
|
|
|
|
inp->inp_laddr.s_addr = INADDR_ANY;
|
|
|
|
|
inp->inp_lport = 0;
|
|
|
|
|
return (EAGAIN);
|
|
|
|
|
}
|
|
|
|
|
if (anonport)
|
|
|
|
|
inp->inp_flags |= INP_ANONPORT;
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Set up a bind operation on a PCB, performing port allocation
|
|
|
|
|
* as required, but do not actually modify the PCB. Callers can
|
|
|
|
|
* either complete the bind by setting inp_laddr/inp_lport and
|
|
|
|
|
* calling in_pcbinshash(), or they can just use the resulting
|
|
|
|
|
* port and address to authorise the sending of a once-off packet.
|
|
|
|
|
*
|
|
|
|
|
* On error, the values of *laddrp and *lportp are not changed.
|
|
|
|
|
*/
|
|
|
|
|
int
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
|
|
|
|
|
u_short *lportp, struct ucred *cred)
|
2002-10-20 21:44:31 +00:00
|
|
|
{
|
|
|
|
|
struct socket *so = inp->inp_socket;
|
1996-10-30 06:13:10 +00:00
|
|
|
unsigned short *lastport;
|
1995-04-09 01:29:31 +00:00
|
|
|
struct sockaddr_in *sin;
|
Improved connection establishment performance by doing local port lookups via
a hashed port list. In the new scheme, in_pcblookup() goes away and is
replaced by a new routine, in_pcblookup_local() for doing the local port
check. Note that this implementation is space inefficient in that the PCB
struct is now too large to fit into 128 bytes. I might deal with this in the
future by using the new zone allocator, but I wanted these changes to be
extensively tested in their current form first.
Also:
1) Fixed off-by-one errors in the port lookup loops in in_pcbbind().
2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash()
to do the initialial hash insertion.
3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability.
4) Added a new routine, in_pcbremlists() to remove the PCB from the various
hash lists.
5) Added/deleted comments where appropriate.
6) Removed unnecessary splnet() locking. In general, the PCB functions should
be called at splnet()...there are unfortunately a few exceptions, however.
7) Reorganized a few structs for better cache line behavior.
8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in
the future, however.
These changes have been tested on wcarchive for more than a month. In tests
done here, connection establishment overhead is reduced by more than 50
times, thus getting rid of one of the major networking scalability problems.
Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a
large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult.
WARNING: Anything that knows about inpcb and tcpcb structs will have to be
recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
|
|
|
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
|
2002-10-20 21:44:31 +00:00
|
|
|
struct in_addr laddr;
|
1994-05-24 10:09:53 +00:00
|
|
|
u_short lport = 0;
|
2002-05-31 11:52:35 +00:00
|
|
|
int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
int error, prison = 0;
|
2005-01-02 01:50:57 +00:00
|
|
|
int dorandom;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
/*
|
|
|
|
|
* Because no actual state changes occur here, a write global write
|
|
|
|
|
* lock on the pcbinfo isn't required.
|
|
|
|
|
*/
|
|
|
|
|
INP_INFO_LOCK_ASSERT(pcbinfo);
|
2003-11-08 23:02:36 +00:00
|
|
|
INP_LOCK_ASSERT(inp);
|
|
|
|
|
|
1996-12-13 21:29:07 +00:00
|
|
|
if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EADDRNOTAVAIL);
|
2002-10-20 21:44:31 +00:00
|
|
|
laddr.s_addr = *laddrp;
|
|
|
|
|
if (nam != NULL && laddr.s_addr != INADDR_ANY)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EINVAL);
|
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
|
|
|
if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
|
2006-06-29 10:49:49 +00:00
|
|
|
wild = INPLOOKUP_WILDCARD;
|
1994-05-24 10:09:53 +00:00
|
|
|
if (nam) {
|
1997-08-16 19:16:27 +00:00
|
|
|
sin = (struct sockaddr_in *)nam;
|
|
|
|
|
if (nam->sa_len != sizeof (*sin))
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EINVAL);
|
|
|
|
|
#ifdef notdef
|
|
|
|
|
/*
|
|
|
|
|
* We should check the family, but old programs
|
|
|
|
|
* incorrectly fail to initialize it.
|
|
|
|
|
*/
|
|
|
|
|
if (sin->sin_family != AF_INET)
|
|
|
|
|
return (EAFNOSUPPORT);
|
|
|
|
|
#endif
|
2000-09-17 13:35:42 +00:00
|
|
|
if (sin->sin_addr.s_addr != INADDR_ANY)
|
2004-03-27 21:05:46 +00:00
|
|
|
if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
|
2000-09-17 13:35:42 +00:00
|
|
|
return(EINVAL);
|
2002-10-20 21:44:31 +00:00
|
|
|
if (sin->sin_port != *lportp) {
|
|
|
|
|
/* Don't allow the port to change. */
|
|
|
|
|
if (*lportp != 0)
|
|
|
|
|
return (EINVAL);
|
|
|
|
|
lport = sin->sin_port;
|
|
|
|
|
}
|
|
|
|
|
/* NB: lport is left as 0 if the port isn't being changed. */
|
1994-05-24 10:09:53 +00:00
|
|
|
if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
|
|
|
|
|
/*
|
|
|
|
|
* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
|
|
|
|
|
* allow complete duplication of binding if
|
|
|
|
|
* SO_REUSEPORT is set, or if SO_REUSEADDR is set
|
|
|
|
|
* and a multicast address is bound on both
|
|
|
|
|
* new and duplicated sockets.
|
|
|
|
|
*/
|
|
|
|
|
if (so->so_options & SO_REUSEADDR)
|
|
|
|
|
reuseport = SO_REUSEADDR|SO_REUSEPORT;
|
|
|
|
|
} else if (sin->sin_addr.s_addr != INADDR_ANY) {
|
|
|
|
|
sin->sin_port = 0; /* yech... */
|
2001-11-06 00:48:01 +00:00
|
|
|
bzero(&sin->sin_zero, sizeof(sin->sin_zero));
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
|
|
|
|
|
return (EADDRNOTAVAIL);
|
|
|
|
|
}
|
2002-10-20 21:44:31 +00:00
|
|
|
laddr = sin->sin_addr;
|
1994-05-24 10:09:53 +00:00
|
|
|
if (lport) {
|
|
|
|
|
struct inpcb *t;
|
2006-04-04 12:26:07 +00:00
|
|
|
struct tcptw *tw;
|
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/* GROSS */
|
The ancient and outdated concept of "privileged ports" in UNIX-type
OSes has probably caused more problems than it ever solved. Allow the
user to retire the old behavior by specifying their own privileged
range with,
net.inet.ip.portrange.reservedhigh default = IPPORT_RESERVED - 1
net.inet.ip.portrange.reservedlo default = 0
Now you can run that webserver without ever needing root at all. Or
just imagine, an ftpd that can really drop privileges, rather than
just set the euid, and still do PORT data transfers from 20/tcp.
Two edge cases to note,
# sysctl net.inet.ip.portrange.reservedhigh=0
Opens all ports to everyone, and,
# sysctl net.inet.ip.portrange.reservedhigh=65535
Locks all network activity to root only (which could actually have
been achieved before with ipfw(8), but is somewhat more
complicated).
For those who stick to the old religion that 0-1023 belong to root and
root alone, don't touch the knobs (or even lock them by raising
securelevel(8)), and nothing changes.
2003-02-21 05:28:27 +00:00
|
|
|
if (ntohs(lport) <= ipport_reservedhigh &&
|
|
|
|
|
ntohs(lport) >= ipport_reservedlow &&
|
2006-11-06 13:42:10 +00:00
|
|
|
priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
|
2007-06-12 00:12:01 +00:00
|
|
|
0))
|
1995-09-21 17:55:49 +00:00
|
|
|
return (EACCES);
|
2004-03-27 21:05:46 +00:00
|
|
|
if (jailed(cred))
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
prison = 1;
|
2006-06-27 11:35:53 +00:00
|
|
|
if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
|
2007-04-10 15:58:38 +00:00
|
|
|
priv_check_cred(so->so_cred,
|
2007-06-12 00:12:01 +00:00
|
|
|
PRIV_NETINET_REUSEPORT, 0) != 0) {
|
1998-03-01 19:39:29 +00:00
|
|
|
t = in_pcblookup_local(inp->inp_pcbinfo,
|
1999-11-22 02:45:11 +00:00
|
|
|
sin->sin_addr, lport,
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
prison ? 0 : INPLOOKUP_WILDCARD);
|
2003-02-19 22:32:43 +00:00
|
|
|
/*
|
|
|
|
|
* XXX
|
|
|
|
|
* This entire block sorely needs a rewrite.
|
|
|
|
|
*/
|
2002-05-31 11:52:35 +00:00
|
|
|
if (t &&
|
2004-05-20 06:35:02 +00:00
|
|
|
((t->inp_vflag & INP_TIMEWAIT) == 0) &&
|
|
|
|
|
(so->so_type != SOCK_STREAM ||
|
|
|
|
|
ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
|
2002-05-31 11:52:35 +00:00
|
|
|
(ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
|
|
|
|
|
ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
|
|
|
|
|
(t->inp_socket->so_options &
|
|
|
|
|
SO_REUSEPORT) == 0) &&
|
|
|
|
|
(so->so_cred->cr_uid !=
|
2004-07-28 13:03:07 +00:00
|
|
|
t->inp_socket->so_cred->cr_uid))
|
2002-05-31 11:52:35 +00:00
|
|
|
return (EADDRINUSE);
|
1998-03-01 19:39:29 +00:00
|
|
|
}
|
2004-03-27 21:05:46 +00:00
|
|
|
if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
|
2001-02-28 09:38:48 +00:00
|
|
|
return (EADDRNOTAVAIL);
|
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
|
|
|
t = in_pcblookup_local(pcbinfo, sin->sin_addr,
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
lport, prison ? 0 : wild);
|
2003-02-19 22:32:43 +00:00
|
|
|
if (t && (t->inp_vflag & INP_TIMEWAIT)) {
|
2006-04-04 12:26:07 +00:00
|
|
|
/*
|
|
|
|
|
* XXXRW: If an incpb has had its timewait
|
|
|
|
|
* state recycled, we treat the address as
|
|
|
|
|
* being in use (for now). This is better
|
|
|
|
|
* than a panic, but not desirable.
|
|
|
|
|
*/
|
|
|
|
|
tw = intotw(inp);
|
|
|
|
|
if (tw == NULL ||
|
|
|
|
|
(reuseport & tw->tw_so_options) == 0)
|
2003-02-19 22:32:43 +00:00
|
|
|
return (EADDRINUSE);
|
2006-04-04 12:26:07 +00:00
|
|
|
} else if (t &&
|
2002-05-31 11:52:35 +00:00
|
|
|
(reuseport & t->inp_socket->so_options) == 0) {
|
2006-11-30 10:54:54 +00:00
|
|
|
#ifdef INET6
|
2002-05-31 11:52:35 +00:00
|
|
|
if (ntohl(sin->sin_addr.s_addr) !=
|
|
|
|
|
INADDR_ANY ||
|
|
|
|
|
ntohl(t->inp_laddr.s_addr) !=
|
|
|
|
|
INADDR_ANY ||
|
|
|
|
|
INP_SOCKAF(so) ==
|
|
|
|
|
INP_SOCKAF(t->inp_socket))
|
2006-11-30 10:54:54 +00:00
|
|
|
#endif
|
2002-05-31 11:52:35 +00:00
|
|
|
return (EADDRINUSE);
|
1999-12-07 17:39:16 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
}
|
2002-10-20 21:44:31 +00:00
|
|
|
if (*lportp != 0)
|
|
|
|
|
lport = *lportp;
|
1996-02-22 21:32:23 +00:00
|
|
|
if (lport == 0) {
|
2008-03-04 19:16:21 +00:00
|
|
|
u_short first, last, aux;
|
2004-04-22 08:34:55 +00:00
|
|
|
int count;
|
1996-02-22 21:32:23 +00:00
|
|
|
|
2002-10-20 21:44:31 +00:00
|
|
|
if (laddr.s_addr != INADDR_ANY)
|
2004-03-27 21:05:46 +00:00
|
|
|
if (prison_ip(cred, 0, &laddr.s_addr))
|
2000-09-17 13:35:42 +00:00
|
|
|
return (EINVAL);
|
1996-08-23 18:59:07 +00:00
|
|
|
|
1996-02-22 21:32:23 +00:00
|
|
|
if (inp->inp_flags & INP_HIGHPORT) {
|
|
|
|
|
first = ipport_hifirstauto; /* sysctl */
|
|
|
|
|
last = ipport_hilastauto;
|
2007-04-30 23:12:05 +00:00
|
|
|
lastport = &pcbinfo->ipi_lasthi;
|
1996-02-22 21:32:23 +00:00
|
|
|
} else if (inp->inp_flags & INP_LOWPORT) {
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check_cred(cred,
|
2007-06-12 00:12:01 +00:00
|
|
|
PRIV_NETINET_RESERVEDPORT, 0);
|
2006-11-06 13:42:10 +00:00
|
|
|
if (error)
|
1997-04-27 20:01:29 +00:00
|
|
|
return error;
|
1996-08-12 14:05:54 +00:00
|
|
|
first = ipport_lowfirstauto; /* 1023 */
|
|
|
|
|
last = ipport_lowlastauto; /* 600 */
|
2007-04-30 23:12:05 +00:00
|
|
|
lastport = &pcbinfo->ipi_lastlow;
|
1996-02-22 21:32:23 +00:00
|
|
|
} else {
|
|
|
|
|
first = ipport_firstauto; /* sysctl */
|
|
|
|
|
last = ipport_lastauto;
|
2007-04-30 23:12:05 +00:00
|
|
|
lastport = &pcbinfo->ipi_lastport;
|
1996-02-22 21:32:23 +00:00
|
|
|
}
|
2005-01-02 01:50:57 +00:00
|
|
|
/*
|
|
|
|
|
* For UDP, use random port allocation as long as the user
|
|
|
|
|
* allows it. For TCP (and as of yet unknown) connections,
|
|
|
|
|
* use random port allocation only if the user allows it AND
|
2005-04-08 08:43:21 +00:00
|
|
|
* ipport_tick() allows it.
|
2005-01-02 01:50:57 +00:00
|
|
|
*/
|
|
|
|
|
if (ipport_randomized &&
|
|
|
|
|
(!ipport_stoprandom || pcbinfo == &udbinfo))
|
|
|
|
|
dorandom = 1;
|
|
|
|
|
else
|
|
|
|
|
dorandom = 0;
|
2005-04-08 08:42:10 +00:00
|
|
|
/*
|
|
|
|
|
* It makes no sense to do random port allocation if
|
|
|
|
|
* we have the only port available.
|
|
|
|
|
*/
|
|
|
|
|
if (first == last)
|
|
|
|
|
dorandom = 0;
|
2005-01-02 01:50:57 +00:00
|
|
|
/* Make sure to not include UDP packets in the count. */
|
|
|
|
|
if (pcbinfo != &udbinfo)
|
|
|
|
|
ipport_tcpallocs++;
|
1996-02-22 21:32:23 +00:00
|
|
|
/*
|
|
|
|
|
* Simple check to ensure all ports are not used up causing
|
|
|
|
|
* a deadlock here.
|
|
|
|
|
*/
|
|
|
|
|
if (first > last) {
|
2008-03-04 19:16:21 +00:00
|
|
|
aux = first;
|
|
|
|
|
first = last;
|
|
|
|
|
last = aux;
|
1996-02-22 21:32:23 +00:00
|
|
|
}
|
2008-03-04 19:16:21 +00:00
|
|
|
|
|
|
|
|
if (dorandom)
|
|
|
|
|
*lastport = first +
|
|
|
|
|
(arc4random() % (last - first));
|
|
|
|
|
|
|
|
|
|
count = last - first;
|
|
|
|
|
|
|
|
|
|
do {
|
|
|
|
|
if (count-- < 0) /* completely used? */
|
|
|
|
|
return (EADDRNOTAVAIL);
|
|
|
|
|
++*lastport;
|
|
|
|
|
if (*lastport < first || *lastport > last)
|
|
|
|
|
*lastport = first;
|
|
|
|
|
lport = htons(*lastport);
|
|
|
|
|
} while (in_pcblookup_local(pcbinfo, laddr, lport,
|
|
|
|
|
wild));
|
1996-02-22 21:32:23 +00:00
|
|
|
}
|
2004-03-27 21:05:46 +00:00
|
|
|
if (prison_ip(cred, 0, &laddr.s_addr))
|
2001-12-13 04:01:23 +00:00
|
|
|
return (EINVAL);
|
2002-10-20 21:44:31 +00:00
|
|
|
*laddrp = laddr.s_addr;
|
|
|
|
|
*lportp = lport;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
1995-02-08 20:22:09 +00:00
|
|
|
/*
|
2002-10-21 13:55:50 +00:00
|
|
|
* Connect from a socket to a specified address.
|
|
|
|
|
* Both address and port must be specified in argument sin.
|
|
|
|
|
* If don't have a local address for this socket yet,
|
|
|
|
|
* then pick one.
|
1995-02-08 20:22:09 +00:00
|
|
|
*/
|
2002-10-21 13:55:50 +00:00
|
|
|
int
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
|
2002-10-21 13:55:50 +00:00
|
|
|
{
|
|
|
|
|
u_short lport, fport;
|
|
|
|
|
in_addr_t laddr, faddr;
|
|
|
|
|
int anonport, error;
|
|
|
|
|
|
2004-08-11 04:35:20 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2004-08-11 04:35:20 +00:00
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
lport = inp->inp_lport;
|
|
|
|
|
laddr = inp->inp_laddr.s_addr;
|
|
|
|
|
anonport = (lport == 0);
|
|
|
|
|
error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
|
2004-03-27 21:05:46 +00:00
|
|
|
NULL, cred);
|
2002-10-21 13:55:50 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
|
|
|
|
|
/* Do the initial binding of the local address if required. */
|
|
|
|
|
if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
|
|
|
|
|
inp->inp_lport = lport;
|
|
|
|
|
inp->inp_laddr.s_addr = laddr;
|
|
|
|
|
if (in_pcbinshash(inp) != 0) {
|
|
|
|
|
inp->inp_laddr.s_addr = INADDR_ANY;
|
|
|
|
|
inp->inp_lport = 0;
|
|
|
|
|
return (EAGAIN);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Commit the remaining changes. */
|
|
|
|
|
inp->inp_lport = lport;
|
|
|
|
|
inp->inp_laddr.s_addr = laddr;
|
|
|
|
|
inp->inp_faddr.s_addr = faddr;
|
|
|
|
|
inp->inp_fport = fport;
|
|
|
|
|
in_pcbrehash(inp);
|
2007-07-01 11:41:27 +00:00
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
if (anonport)
|
|
|
|
|
inp->inp_flags |= INP_ANONPORT;
|
|
|
|
|
return (0);
|
|
|
|
|
}
|
1995-02-08 20:22:09 +00:00
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
/*
|
|
|
|
|
* Set up for a connect from a socket to the specified address.
|
|
|
|
|
* On entry, *laddrp and *lportp should contain the current local
|
|
|
|
|
* address and port for the PCB; these are updated to the values
|
|
|
|
|
* that should be placed in inp_laddr and inp_lport to complete
|
|
|
|
|
* the connect.
|
|
|
|
|
*
|
|
|
|
|
* On success, *faddrp and *fportp will be set to the remote address
|
|
|
|
|
* and port. These are not updated in the error case.
|
|
|
|
|
*
|
|
|
|
|
* If the operation fails because the connection already exists,
|
|
|
|
|
* *oinpp will be set to the PCB of that connection so that the
|
|
|
|
|
* caller can decide to override it. In all other cases, *oinpp
|
|
|
|
|
* is set to NULL.
|
|
|
|
|
*/
|
1995-02-08 20:22:09 +00:00
|
|
|
int
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
|
|
|
|
|
in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
|
|
|
|
|
struct inpcb **oinpp, struct ucred *cred)
|
1995-02-08 20:22:09 +00:00
|
|
|
{
|
2002-10-21 13:55:50 +00:00
|
|
|
struct sockaddr_in *sin = (struct sockaddr_in *)nam;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct in_ifaddr *ia;
|
2002-10-21 13:55:50 +00:00
|
|
|
struct sockaddr_in sa;
|
2004-03-27 21:05:46 +00:00
|
|
|
struct ucred *socred;
|
2002-10-21 13:55:50 +00:00
|
|
|
struct inpcb *oinp;
|
|
|
|
|
struct in_addr laddr, faddr;
|
|
|
|
|
u_short lport, fport;
|
|
|
|
|
int error;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
/*
|
|
|
|
|
* Because a global state change doesn't actually occur here, a read
|
|
|
|
|
* lock is sufficient.
|
|
|
|
|
*/
|
|
|
|
|
INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
|
2004-08-11 04:35:20 +00:00
|
|
|
INP_LOCK_ASSERT(inp);
|
|
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
if (oinpp != NULL)
|
|
|
|
|
*oinpp = NULL;
|
1997-08-16 19:16:27 +00:00
|
|
|
if (nam->sa_len != sizeof (*sin))
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EINVAL);
|
|
|
|
|
if (sin->sin_family != AF_INET)
|
|
|
|
|
return (EAFNOSUPPORT);
|
|
|
|
|
if (sin->sin_port == 0)
|
|
|
|
|
return (EADDRNOTAVAIL);
|
2002-10-21 13:55:50 +00:00
|
|
|
laddr.s_addr = *laddrp;
|
|
|
|
|
lport = *lportp;
|
|
|
|
|
faddr = sin->sin_addr;
|
|
|
|
|
fport = sin->sin_port;
|
2004-03-27 21:05:46 +00:00
|
|
|
socred = inp->inp_socket->so_cred;
|
|
|
|
|
if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
|
2002-10-21 13:55:50 +00:00
|
|
|
bzero(&sa, sizeof(sa));
|
2004-03-27 21:05:46 +00:00
|
|
|
sa.sin_addr.s_addr = htonl(prison_getip(socred));
|
2002-10-21 13:55:50 +00:00
|
|
|
sa.sin_len = sizeof(sa);
|
|
|
|
|
sa.sin_family = AF_INET;
|
|
|
|
|
error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
|
2004-03-27 21:05:46 +00:00
|
|
|
&laddr.s_addr, &lport, cred);
|
2002-10-21 13:55:50 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
|
|
|
|
}
|
1996-12-13 21:29:07 +00:00
|
|
|
if (!TAILQ_EMPTY(&in_ifaddrhead)) {
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
|
* If the destination address is INADDR_ANY,
|
|
|
|
|
* use the primary local address.
|
|
|
|
|
* If the supplied address is INADDR_BROADCAST,
|
|
|
|
|
* and the primary interface supports broadcast,
|
|
|
|
|
* choose the broadcast address for that interface.
|
|
|
|
|
*/
|
2002-10-21 13:55:50 +00:00
|
|
|
if (faddr.s_addr == INADDR_ANY)
|
|
|
|
|
faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
|
|
|
|
|
else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
|
|
|
|
|
(TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
|
|
|
|
|
IFF_BROADCAST))
|
|
|
|
|
faddr = satosin(&TAILQ_FIRST(
|
|
|
|
|
&in_ifaddrhead)->ia_broadaddr)->sin_addr;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2002-10-21 13:55:50 +00:00
|
|
|
if (laddr.s_addr == INADDR_ANY) {
|
1994-05-24 10:09:53 +00:00
|
|
|
ia = (struct in_ifaddr *)0;
|
1995-05-30 08:16:23 +00:00
|
|
|
/*
|
2003-11-20 20:07:39 +00:00
|
|
|
* If route is known our src addr is taken from the i/f,
|
|
|
|
|
* else punt.
|
2006-02-16 15:45:28 +00:00
|
|
|
*
|
|
|
|
|
* Find out route to destination
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2006-02-16 15:45:28 +00:00
|
|
|
if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
|
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
|
|
|
ia = ip_rtaddr(faddr, inp->inp_inc.inc_fibnum);
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
2006-02-16 15:45:28 +00:00
|
|
|
* If we found a route, use the address corresponding to
|
|
|
|
|
* the outgoing interface.
|
|
|
|
|
*
|
|
|
|
|
* Otherwise assume faddr is reachable on a directly connected
|
|
|
|
|
* network and try to find a corresponding interface to take
|
|
|
|
|
* the source address from.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
if (ia == 0) {
|
2002-10-21 13:55:50 +00:00
|
|
|
bzero(&sa, sizeof(sa));
|
|
|
|
|
sa.sin_addr = faddr;
|
|
|
|
|
sa.sin_len = sizeof(sa);
|
|
|
|
|
sa.sin_family = AF_INET;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ia == 0)
|
2002-10-21 13:55:50 +00:00
|
|
|
ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ia == 0)
|
2004-06-16 10:02:36 +00:00
|
|
|
return (ENETUNREACH);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* If the destination address is multicast and an outgoing
|
|
|
|
|
* interface has been set as a multicast option, use the
|
|
|
|
|
* address of that interface as our source address.
|
|
|
|
|
*/
|
2002-10-21 13:55:50 +00:00
|
|
|
if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
|
1994-05-24 10:09:53 +00:00
|
|
|
inp->inp_moptions != NULL) {
|
|
|
|
|
struct ip_moptions *imo;
|
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
|
|
|
|
|
imo = inp->inp_moptions;
|
|
|
|
|
if (imo->imo_multicast_ifp != NULL) {
|
|
|
|
|
ifp = imo->imo_multicast_ifp;
|
2001-02-04 16:08:18 +00:00
|
|
|
TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ia->ia_ifp == ifp)
|
|
|
|
|
break;
|
|
|
|
|
if (ia == 0)
|
|
|
|
|
return (EADDRNOTAVAIL);
|
|
|
|
|
}
|
|
|
|
|
}
|
2002-10-21 13:55:50 +00:00
|
|
|
laddr = ia->ia_addr.sin_addr;
|
1995-02-08 20:22:09 +00:00
|
|
|
}
|
|
|
|
|
|
2002-10-21 13:55:50 +00:00
|
|
|
oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
|
|
|
|
|
0, NULL);
|
|
|
|
|
if (oinp != NULL) {
|
|
|
|
|
if (oinpp != NULL)
|
|
|
|
|
*oinpp = oinp;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (EADDRINUSE);
|
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
|
|
|
}
|
2002-10-21 13:55:50 +00:00
|
|
|
if (lport == 0) {
|
2004-03-27 21:05:46 +00:00
|
|
|
error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
|
|
|
|
|
cred);
|
2002-10-21 13:55:50 +00:00
|
|
|
if (error)
|
|
|
|
|
return (error);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2002-10-21 13:55:50 +00:00
|
|
|
*laddrp = laddr.s_addr;
|
|
|
|
|
*lportp = lport;
|
|
|
|
|
*faddrp = faddr.s_addr;
|
|
|
|
|
*fportp = fport;
|
1994-05-24 10:09:53 +00:00
|
|
|
return (0);
|
|
|
|
|
}
|
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbdisconnect(struct inpcb *inp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2005-06-01 11:39:42 +00:00
|
|
|
|
2005-06-01 11:43:39 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
|
inp->inp_faddr.s_addr = INADDR_ANY;
|
|
|
|
|
inp->inp_fport = 0;
|
1995-04-09 01:29:31 +00:00
|
|
|
in_pcbrehash(inp);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
2006-04-01 16:04:42 +00:00
|
|
|
/*
|
|
|
|
|
* In the old world order, in_pcbdetach() served two functions: to detach the
|
|
|
|
|
* pcb from the socket/potentially free the socket, and to free the pcb
|
|
|
|
|
* itself. In the new world order, the protocol code is responsible for
|
|
|
|
|
* managing the relationship with the socket, and this code simply frees the
|
|
|
|
|
* pcb.
|
|
|
|
|
*/
|
1994-05-25 09:21:21 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbdetach(struct inpcb *inp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-04-01 16:04:42 +00:00
|
|
|
|
|
|
|
|
KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
|
|
|
|
|
inp->inp_socket->so_pcb = NULL;
|
|
|
|
|
inp->inp_socket = NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
in_pcbfree(struct inpcb *inp)
|
|
|
|
|
{
|
1998-03-24 18:06:34 +00:00
|
|
|
struct inpcbinfo *ipi = inp->inp_pcbinfo;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2006-04-01 16:04:42 +00:00
|
|
|
KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
|
2008-04-17 21:38:18 +00:00
|
|
|
|
2005-06-01 11:43:39 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(ipi);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2003-11-08 23:02:36 +00:00
|
|
|
|
2007-07-03 12:13:45 +00:00
|
|
|
#ifdef IPSEC
|
2000-07-04 16:35:15 +00:00
|
|
|
ipsec4_delete_pcbpolicy(inp);
|
2007-07-03 12:13:45 +00:00
|
|
|
#endif /*IPSEC*/
|
1998-03-24 18:06:34 +00:00
|
|
|
inp->inp_gencnt = ++ipi->ipi_gencnt;
|
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
|
|
|
in_pcbremlists(inp);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (inp->inp_options)
|
|
|
|
|
(void)m_free(inp->inp_options);
|
Import rewrite of IPv4 socket multicast layer to support source-specific
and protocol-independent host mode multicast. The code is written to
accomodate IPv6, IGMPv3 and MLDv2 with only a little additional work.
This change only pertains to FreeBSD's use as a multicast end-station and
does not concern multicast routing; for an IGMPv3/MLDv2 router
implementation, consider the XORP project.
The work is based on Wilbert de Graaf's IGMPv3 code drop for FreeBSD 4.6,
which is available at: http://www.kloosterhof.com/wilbert/igmpv3.html
Summary
* IPv4 multicast socket processing is now moved out of ip_output.c
into a new module, in_mcast.c.
* The in_mcast.c module implements the IPv4 legacy any-source API in
terms of the protocol-independent source-specific API.
* Source filters are lazy allocated as the common case does not use them.
They are part of per inpcb state and are covered by the inpcb lock.
* struct ip_mreqn is now supported to allow applications to specify
multicast joins by interface index in the legacy IPv4 any-source API.
* In UDP, an incoming multicast datagram only requires that the source
port matches the 4-tuple if the socket was already bound by source port.
An unbound socket SHOULD be able to receive multicasts sent from an
ephemeral source port.
* The UDP socket multicast filter mode defaults to exclusive, that is,
sources present in the per-socket list will be blocked from delivery.
* The RFC 3678 userland functions have been added to libc: setsourcefilter,
getsourcefilter, setipv4sourcefilter, getipv4sourcefilter.
* Definitions for IGMPv3 are merged but not yet used.
* struct sockaddr_storage is now referenced from <netinet/in.h>. It
is therefore defined there if not already declared in the same way
as for the C99 types.
* The RFC 1724 hack (specify 0.0.0.0/8 addresses to IP_MULTICAST_IF
which are then interpreted as interface indexes) is now deprecated.
* A patch for the Rhyolite.com routed in the FreeBSD base system
is available in the -net archives. This only affects individuals
running RIPv1 or RIPv2 via point-to-point and/or unnumbered interfaces.
* Make IPv6 detach path similar to IPv4's in code flow; functionally same.
* Bump __FreeBSD_version to 700048; see UPDATING.
This work was financially supported by another FreeBSD committer.
Obtained from: p4://bms_netdev
Submitted by: Wilbert de Graaf (original work)
Reviewed by: rwatson (locking), silence from fenner,
net@ (but with encouragement)
2007-06-12 16:24:56 +00:00
|
|
|
if (inp->inp_moptions != NULL)
|
|
|
|
|
inp_freemoptions(inp->inp_moptions);
|
1999-12-07 17:39:16 +00:00
|
|
|
inp->inp_vflag = 0;
|
2006-07-18 22:34:27 +00:00
|
|
|
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#ifdef MAC
|
2007-10-24 19:04:04 +00:00
|
|
|
mac_inpcb_destroy(inp);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#endif
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2002-03-20 05:48:55 +00:00
|
|
|
uma_zfree(ipi->ipi_zone, inp);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
2006-04-25 11:17:35 +00:00
|
|
|
/*
|
|
|
|
|
* TCP needs to maintain its inpcb structure after the TCP connection has
|
|
|
|
|
* been torn down. However, it must be disconnected from the inpcb hashes as
|
|
|
|
|
* it must not prevent binding of future connections to the same port/ip
|
|
|
|
|
* combination by other inpcbs.
|
|
|
|
|
*/
|
|
|
|
|
void
|
|
|
|
|
in_pcbdrop(struct inpcb *inp)
|
|
|
|
|
{
|
|
|
|
|
|
2006-04-25 23:23:13 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2006-04-25 11:17:35 +00:00
|
|
|
|
|
|
|
|
inp->inp_vflag |= INP_DROPPED;
|
|
|
|
|
if (inp->inp_lport) {
|
|
|
|
|
struct inpcbport *phd = inp->inp_phd;
|
|
|
|
|
|
|
|
|
|
LIST_REMOVE(inp, inp_hash);
|
|
|
|
|
LIST_REMOVE(inp, inp_portlist);
|
|
|
|
|
if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
|
|
|
|
|
LIST_REMOVE(phd, phd_hash);
|
|
|
|
|
free(phd, M_PCB);
|
|
|
|
|
}
|
|
|
|
|
inp->inp_lport = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2007-05-11 10:20:51 +00:00
|
|
|
/*
|
|
|
|
|
* Common routines to return the socket addresses associated with inpcbs.
|
|
|
|
|
*/
|
2002-08-21 11:57:12 +00:00
|
|
|
struct sockaddr *
|
2006-01-22 01:16:25 +00:00
|
|
|
in_sockaddr(in_port_t port, struct in_addr *addr_p)
|
2002-08-21 11:57:12 +00:00
|
|
|
{
|
|
|
|
|
struct sockaddr_in *sin;
|
|
|
|
|
|
|
|
|
|
MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
|
2003-02-19 05:47:46 +00:00
|
|
|
M_WAITOK | M_ZERO);
|
2002-08-21 11:57:12 +00:00
|
|
|
sin->sin_family = AF_INET;
|
|
|
|
|
sin->sin_len = sizeof(*sin);
|
|
|
|
|
sin->sin_addr = *addr_p;
|
|
|
|
|
sin->sin_port = port;
|
|
|
|
|
|
|
|
|
|
return (struct sockaddr *)sin;
|
|
|
|
|
}
|
|
|
|
|
|
1997-02-18 20:46:36 +00:00
|
|
|
int
|
2007-05-11 10:20:51 +00:00
|
|
|
in_getsockaddr(struct socket *so, struct sockaddr **nam)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-01-22 01:16:25 +00:00
|
|
|
struct inpcb *inp;
|
2002-08-21 11:57:12 +00:00
|
|
|
struct in_addr addr;
|
|
|
|
|
in_port_t port;
|
1997-12-25 06:57:36 +00:00
|
|
|
|
1997-05-19 01:28:39 +00:00
|
|
|
inp = sotoinpcb(so);
|
2007-05-11 10:20:51 +00:00
|
|
|
KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
|
2006-04-22 19:10:02 +00:00
|
|
|
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RLOCK(inp);
|
2002-08-21 11:57:12 +00:00
|
|
|
port = inp->inp_lport;
|
|
|
|
|
addr = inp->inp_laddr;
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RUNLOCK(inp);
|
1997-12-25 06:57:36 +00:00
|
|
|
|
2002-08-21 11:57:12 +00:00
|
|
|
*nam = in_sockaddr(port, &addr);
|
1997-02-18 20:46:36 +00:00
|
|
|
return 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
1997-02-18 20:46:36 +00:00
|
|
|
int
|
2007-05-11 10:20:51 +00:00
|
|
|
in_getpeeraddr(struct socket *so, struct sockaddr **nam)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-01-22 01:16:25 +00:00
|
|
|
struct inpcb *inp;
|
2002-08-21 11:57:12 +00:00
|
|
|
struct in_addr addr;
|
|
|
|
|
in_port_t port;
|
1997-12-25 06:57:36 +00:00
|
|
|
|
1997-05-19 01:28:39 +00:00
|
|
|
inp = sotoinpcb(so);
|
2007-05-11 10:20:51 +00:00
|
|
|
KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
|
2006-04-22 19:10:02 +00:00
|
|
|
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RLOCK(inp);
|
2002-08-21 11:57:12 +00:00
|
|
|
port = inp->inp_fport;
|
|
|
|
|
addr = inp->inp_faddr;
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RUNLOCK(inp);
|
1997-12-25 06:57:36 +00:00
|
|
|
|
2002-08-21 11:57:12 +00:00
|
|
|
*nam = in_sockaddr(port, &addr);
|
1997-02-18 20:46:36 +00:00
|
|
|
return 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
1994-05-25 09:21:21 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
|
|
|
|
|
struct inpcb *(*notify)(struct inpcb *, int))
|
2001-02-22 21:23:45 +00:00
|
|
|
{
|
2008-04-06 21:20:56 +00:00
|
|
|
struct inpcb *inp, *inp_temp;
|
2001-02-22 21:23:45 +00:00
|
|
|
|
2003-02-12 23:55:07 +00:00
|
|
|
INP_INFO_WLOCK(pcbinfo);
|
2008-04-06 21:20:56 +00:00
|
|
|
LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inp);
|
2001-02-22 21:23:45 +00:00
|
|
|
#ifdef INET6
|
2002-06-10 20:05:46 +00:00
|
|
|
if ((inp->inp_vflag & INP_IPV4) == 0) {
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2001-02-22 21:23:45 +00:00
|
|
|
continue;
|
2002-06-10 20:05:46 +00:00
|
|
|
}
|
2001-02-22 21:23:45 +00:00
|
|
|
#endif
|
|
|
|
|
if (inp->inp_faddr.s_addr != faddr.s_addr ||
|
2002-06-10 20:05:46 +00:00
|
|
|
inp->inp_socket == NULL) {
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2003-02-12 23:55:07 +00:00
|
|
|
continue;
|
2002-06-10 20:05:46 +00:00
|
|
|
}
|
2003-02-12 23:55:07 +00:00
|
|
|
if ((*notify)(inp, errno))
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2001-02-22 21:23:45 +00:00
|
|
|
}
|
2003-02-12 23:55:07 +00:00
|
|
|
INP_INFO_WUNLOCK(pcbinfo);
|
2001-02-22 21:23:45 +00:00
|
|
|
}
|
|
|
|
|
|
2001-08-04 17:10:14 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
|
2001-08-04 17:10:14 +00:00
|
|
|
{
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
struct ip_moptions *imo;
|
|
|
|
|
int i, gap;
|
|
|
|
|
|
2002-06-10 20:05:46 +00:00
|
|
|
INP_INFO_RLOCK(pcbinfo);
|
2007-04-30 23:12:05 +00:00
|
|
|
LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inp);
|
2001-08-04 17:10:14 +00:00
|
|
|
imo = inp->inp_moptions;
|
|
|
|
|
if ((inp->inp_vflag & INP_IPV4) &&
|
|
|
|
|
imo != NULL) {
|
|
|
|
|
/*
|
|
|
|
|
* Unselect the outgoing interface if it is being
|
|
|
|
|
* detached.
|
|
|
|
|
*/
|
|
|
|
|
if (imo->imo_multicast_ifp == ifp)
|
|
|
|
|
imo->imo_multicast_ifp = NULL;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Drop multicast group membership if we joined
|
|
|
|
|
* through the interface being detached.
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0, gap = 0; i < imo->imo_num_memberships;
|
|
|
|
|
i++) {
|
|
|
|
|
if (imo->imo_membership[i]->inm_ifp == ifp) {
|
|
|
|
|
in_delmulti(imo->imo_membership[i]);
|
|
|
|
|
gap++;
|
|
|
|
|
} else if (gap != 0)
|
|
|
|
|
imo->imo_membership[i - gap] =
|
|
|
|
|
imo->imo_membership[i];
|
|
|
|
|
}
|
|
|
|
|
imo->imo_num_memberships -= gap;
|
|
|
|
|
}
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2001-08-04 17:10:14 +00:00
|
|
|
}
|
2002-06-12 03:08:08 +00:00
|
|
|
INP_INFO_RUNLOCK(pcbinfo);
|
2001-08-04 17:10:14 +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
|
|
|
/*
|
|
|
|
|
* Lookup a PCB based on the local address and port.
|
|
|
|
|
*/
|
2006-02-04 07:59:17 +00:00
|
|
|
#define INP_LOOKUP_MAPPED_PCB_COST 3
|
1994-05-24 10:09:53 +00:00
|
|
|
struct inpcb *
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
|
|
|
|
|
u_int lport_arg, int wild_okay)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2006-01-22 01:16:25 +00:00
|
|
|
struct inpcb *inp;
|
2006-02-04 07:59:17 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
|
|
|
|
|
#else
|
|
|
|
|
int matchwild = 3;
|
|
|
|
|
#endif
|
|
|
|
|
int wildcard;
|
Improved connection establishment performance by doing local port lookups via
a hashed port list. In the new scheme, in_pcblookup() goes away and is
replaced by a new routine, in_pcblookup_local() for doing the local port
check. Note that this implementation is space inefficient in that the PCB
struct is now too large to fit into 128 bytes. I might deal with this in the
future by using the new zone allocator, but I wanted these changes to be
extensively tested in their current form first.
Also:
1) Fixed off-by-one errors in the port lookup loops in in_pcbbind().
2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash()
to do the initialial hash insertion.
3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability.
4) Added a new routine, in_pcbremlists() to remove the PCB from the various
hash lists.
5) Added/deleted comments where appropriate.
6) Removed unnecessary splnet() locking. In general, the PCB functions should
be called at splnet()...there are unfortunately a few exceptions, however.
7) Reorganized a few structs for better cache line behavior.
8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in
the future, however.
These changes have been tested on wcarchive for more than a month. In tests
done here, connection establishment overhead is reduced by more than 50
times, thus getting rid of one of the major networking scalability problems.
Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a
large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult.
WARNING: Anything that knows about inpcb and tcpcb structs will have to be
recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
|
|
|
u_short lport = lport_arg;
|
1995-04-10 08:52:45 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_INFO_LOCK_ASSERT(pcbinfo);
|
2003-11-13 05:16:56 +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
|
|
|
if (!wild_okay) {
|
|
|
|
|
struct inpcbhead *head;
|
|
|
|
|
/*
|
|
|
|
|
* Look for an unconnected (wildcard foreign addr) PCB that
|
|
|
|
|
* matches the local address and port we're looking for.
|
|
|
|
|
*/
|
2007-04-30 23:12:05 +00:00
|
|
|
head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
|
|
|
|
|
0, pcbinfo->ipi_hashmask)];
|
2001-02-04 13:13:25 +00:00
|
|
|
LIST_FOREACH(inp, head, inp_hash) {
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
1999-12-21 11:14:12 +00:00
|
|
|
if ((inp->inp_vflag & INP_IPV4) == 0)
|
1999-12-07 17:39:16 +00:00
|
|
|
continue;
|
|
|
|
|
#endif
|
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
|
|
|
if (inp->inp_faddr.s_addr == INADDR_ANY &&
|
|
|
|
|
inp->inp_laddr.s_addr == laddr.s_addr &&
|
|
|
|
|
inp->inp_lport == lport) {
|
|
|
|
|
/*
|
|
|
|
|
* Found.
|
|
|
|
|
*/
|
|
|
|
|
return (inp);
|
|
|
|
|
}
|
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
|
|
|
/*
|
|
|
|
|
* Not found.
|
|
|
|
|
*/
|
|
|
|
|
return (NULL);
|
|
|
|
|
} else {
|
|
|
|
|
struct inpcbporthead *porthash;
|
|
|
|
|
struct inpcbport *phd;
|
|
|
|
|
struct inpcb *match = NULL;
|
|
|
|
|
/*
|
|
|
|
|
* Best fit PCB lookup.
|
|
|
|
|
*
|
|
|
|
|
* First see if this local port is in use by looking on the
|
|
|
|
|
* port hash list.
|
|
|
|
|
*/
|
2007-04-30 23:12:05 +00:00
|
|
|
porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
|
|
|
|
|
pcbinfo->ipi_porthashmask)];
|
2001-02-04 13:13:25 +00:00
|
|
|
LIST_FOREACH(phd, porthash, phd_hash) {
|
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
|
|
|
if (phd->phd_port == lport)
|
1994-05-24 10:09:53 +00:00
|
|
|
break;
|
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
|
|
|
}
|
|
|
|
|
if (phd != NULL) {
|
|
|
|
|
/*
|
|
|
|
|
* Port is in use by one or more PCBs. Look for best
|
|
|
|
|
* fit.
|
|
|
|
|
*/
|
2001-02-04 16:08:18 +00:00
|
|
|
LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
|
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
|
|
|
wildcard = 0;
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
1999-12-21 11:14:12 +00:00
|
|
|
if ((inp->inp_vflag & INP_IPV4) == 0)
|
1999-12-07 17:39:16 +00:00
|
|
|
continue;
|
2006-02-04 07:59:17 +00:00
|
|
|
/*
|
|
|
|
|
* We never select the PCB that has
|
|
|
|
|
* INP_IPV6 flag and is bound to :: if
|
|
|
|
|
* we have another PCB which is bound
|
|
|
|
|
* to 0.0.0.0. If a PCB has the
|
|
|
|
|
* INP_IPV6 flag, then we set its cost
|
|
|
|
|
* higher than IPv4 only PCBs.
|
|
|
|
|
*
|
|
|
|
|
* Note that the case only happens
|
|
|
|
|
* when a socket is bound to ::, under
|
|
|
|
|
* the condition that the use of the
|
|
|
|
|
* mapped address is allowed.
|
|
|
|
|
*/
|
|
|
|
|
if ((inp->inp_vflag & INP_IPV6) != 0)
|
|
|
|
|
wildcard += INP_LOOKUP_MAPPED_PCB_COST;
|
1999-12-07 17:39:16 +00:00
|
|
|
#endif
|
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
|
|
|
if (inp->inp_faddr.s_addr != INADDR_ANY)
|
|
|
|
|
wildcard++;
|
|
|
|
|
if (inp->inp_laddr.s_addr != INADDR_ANY) {
|
|
|
|
|
if (laddr.s_addr == INADDR_ANY)
|
|
|
|
|
wildcard++;
|
|
|
|
|
else if (inp->inp_laddr.s_addr != laddr.s_addr)
|
|
|
|
|
continue;
|
|
|
|
|
} else {
|
|
|
|
|
if (laddr.s_addr != INADDR_ANY)
|
|
|
|
|
wildcard++;
|
|
|
|
|
}
|
|
|
|
|
if (wildcard < matchwild) {
|
|
|
|
|
match = inp;
|
|
|
|
|
matchwild = wildcard;
|
|
|
|
|
if (matchwild == 0) {
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
1995-03-02 19:29:42 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +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
|
|
|
return (match);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
}
|
2006-02-04 07:59:17 +00:00
|
|
|
#undef INP_LOOKUP_MAPPED_PCB_COST
|
1995-04-09 01:29:31 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Lookup PCB in hash list.
|
|
|
|
|
*/
|
|
|
|
|
struct inpcb *
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
|
|
|
|
|
u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
|
|
|
|
|
struct ifnet *ifp)
|
1995-04-09 01:29:31 +00:00
|
|
|
{
|
|
|
|
|
struct inpcbhead *head;
|
2006-01-22 01:16:25 +00:00
|
|
|
struct inpcb *inp;
|
1995-04-09 01:29:31 +00:00
|
|
|
u_short fport = fport_arg, lport = lport_arg;
|
|
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_INFO_LOCK_ASSERT(pcbinfo);
|
2006-04-22 19:15:20 +00:00
|
|
|
|
1995-04-09 01:29:31 +00:00
|
|
|
/*
|
|
|
|
|
* First look for an exact match.
|
|
|
|
|
*/
|
2007-04-30 23:12:05 +00:00
|
|
|
head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
|
|
|
|
|
pcbinfo->ipi_hashmask)];
|
2001-02-04 13:13:25 +00:00
|
|
|
LIST_FOREACH(inp, head, inp_hash) {
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
1999-12-21 11:14:12 +00:00
|
|
|
if ((inp->inp_vflag & INP_IPV4) == 0)
|
1999-12-07 17:39:16 +00:00
|
|
|
continue;
|
|
|
|
|
#endif
|
1996-10-07 19:06:12 +00:00
|
|
|
if (inp->inp_faddr.s_addr == faddr.s_addr &&
|
1997-04-03 05:14:45 +00:00
|
|
|
inp->inp_laddr.s_addr == laddr.s_addr &&
|
|
|
|
|
inp->inp_fport == fport &&
|
2006-11-30 10:54:54 +00:00
|
|
|
inp->inp_lport == lport)
|
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
|
|
|
return (inp);
|
1996-10-07 19:06:12 +00:00
|
|
|
}
|
2006-11-30 10:54:54 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Then look for a wildcard match, if requested.
|
|
|
|
|
*/
|
1996-10-07 19:06:12 +00:00
|
|
|
if (wildcard) {
|
|
|
|
|
struct inpcb *local_wild = NULL;
|
2006-11-30 10:54:54 +00:00
|
|
|
#ifdef INET6
|
1999-12-07 17:39:16 +00:00
|
|
|
struct inpcb *local_wild_mapped = NULL;
|
2006-11-30 10:54:54 +00:00
|
|
|
#endif
|
1996-10-07 19:06:12 +00:00
|
|
|
|
2007-04-30 23:12:05 +00:00
|
|
|
head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
|
|
|
|
|
0, pcbinfo->ipi_hashmask)];
|
2001-02-04 13:13:25 +00:00
|
|
|
LIST_FOREACH(inp, head, inp_hash) {
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
1999-12-21 11:14:12 +00:00
|
|
|
if ((inp->inp_vflag & INP_IPV4) == 0)
|
1999-12-07 17:39:16 +00:00
|
|
|
continue;
|
|
|
|
|
#endif
|
1996-10-07 19:06:12 +00:00
|
|
|
if (inp->inp_faddr.s_addr == INADDR_ANY &&
|
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_lport == lport) {
|
1999-12-07 17:39:16 +00:00
|
|
|
if (ifp && ifp->if_type == IFT_FAITH &&
|
|
|
|
|
(inp->inp_flags & INP_FAITH) == 0)
|
|
|
|
|
continue;
|
1996-10-07 19:06:12 +00:00
|
|
|
if (inp->inp_laddr.s_addr == laddr.s_addr)
|
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
|
|
|
return (inp);
|
1999-12-07 17:39:16 +00:00
|
|
|
else if (inp->inp_laddr.s_addr == INADDR_ANY) {
|
2006-11-30 10:54:54 +00:00
|
|
|
#ifdef INET6
|
1999-12-07 17:39:16 +00:00
|
|
|
if (INP_CHECK_SOCKAF(inp->inp_socket,
|
|
|
|
|
AF_INET6))
|
|
|
|
|
local_wild_mapped = inp;
|
|
|
|
|
else
|
2006-11-30 10:54:54 +00:00
|
|
|
#endif
|
|
|
|
|
local_wild = inp;
|
1999-12-07 17:39:16 +00:00
|
|
|
}
|
1996-10-07 19:06:12 +00:00
|
|
|
}
|
|
|
|
|
}
|
2006-11-30 10:54:54 +00:00
|
|
|
#ifdef INET6
|
1999-12-07 17:39:16 +00:00
|
|
|
if (local_wild == NULL)
|
|
|
|
|
return (local_wild_mapped);
|
2006-11-30 10:54:54 +00:00
|
|
|
#endif
|
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
|
|
|
return (local_wild);
|
1996-10-07 19:06:12 +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
|
|
|
return (NULL);
|
1995-04-09 01:29:31 +00:00
|
|
|
}
|
|
|
|
|
|
1995-04-10 08:52:45 +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
|
|
|
* Insert PCB onto various hash lists.
|
1995-04-10 08:52:45 +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
|
|
|
int
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbinshash(struct inpcb *inp)
|
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
|
|
|
struct inpcbhead *pcbhash;
|
|
|
|
|
struct inpcbporthead *pcbporthash;
|
|
|
|
|
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
|
|
|
|
|
struct inpcbport *phd;
|
1999-12-07 17:39:16 +00:00
|
|
|
u_int32_t hashkey_faddr;
|
1995-04-09 01:29:31 +00:00
|
|
|
|
2003-11-08 23:02:36 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2006-04-22 19:15:20 +00:00
|
|
|
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (inp->inp_vflag & INP_IPV6)
|
|
|
|
|
hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
|
|
|
|
|
else
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
hashkey_faddr = inp->inp_faddr.s_addr;
|
|
|
|
|
|
2007-04-30 23:12:05 +00:00
|
|
|
pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
|
|
|
|
|
inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
|
1995-04-09 01:29:31 +00:00
|
|
|
|
2007-04-30 23:12:05 +00:00
|
|
|
pcbporthash = &pcbinfo->ipi_porthashbase[
|
|
|
|
|
INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
|
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
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Go through port list and look for a head for this lport.
|
|
|
|
|
*/
|
2001-02-04 13:13:25 +00:00
|
|
|
LIST_FOREACH(phd, pcbporthash, phd_hash) {
|
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
|
|
|
if (phd->phd_port == inp->inp_lport)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
/*
|
|
|
|
|
* If none exists, malloc one and tack it on.
|
|
|
|
|
*/
|
|
|
|
|
if (phd == NULL) {
|
|
|
|
|
MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
|
|
|
|
|
if (phd == NULL) {
|
|
|
|
|
return (ENOBUFS); /* XXX */
|
|
|
|
|
}
|
|
|
|
|
phd->phd_port = inp->inp_lport;
|
|
|
|
|
LIST_INIT(&phd->phd_pcblist);
|
|
|
|
|
LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
|
|
|
|
|
}
|
|
|
|
|
inp->inp_phd = phd;
|
|
|
|
|
LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
|
|
|
|
|
LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
|
|
|
|
|
return (0);
|
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
|
|
|
/*
|
|
|
|
|
* Move PCB to the proper hash bucket when { faddr, fport } have been
|
|
|
|
|
* changed. NOTE: This does not handle the case of the lport changing (the
|
|
|
|
|
* hashed port list would have to be updated as well), so the lport must
|
|
|
|
|
* not change after in_pcbinshash() has been called.
|
|
|
|
|
*/
|
1995-04-09 01:29:31 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbrehash(struct inpcb *inp)
|
1995-04-09 01:29:31 +00:00
|
|
|
{
|
2003-11-08 23:02:36 +00:00
|
|
|
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
|
1995-04-09 01:29:31 +00:00
|
|
|
struct inpcbhead *head;
|
1999-12-07 17:39:16 +00:00
|
|
|
u_int32_t hashkey_faddr;
|
|
|
|
|
|
2003-11-08 23:02:36 +00:00
|
|
|
INP_INFO_WLOCK_ASSERT(pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2006-04-22 19:15:20 +00:00
|
|
|
|
1999-12-07 17:39:16 +00:00
|
|
|
#ifdef INET6
|
|
|
|
|
if (inp->inp_vflag & INP_IPV6)
|
|
|
|
|
hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
|
|
|
|
|
else
|
|
|
|
|
#endif /* INET6 */
|
|
|
|
|
hashkey_faddr = inp->inp_faddr.s_addr;
|
1995-04-09 01:29:31 +00:00
|
|
|
|
2007-04-30 23:12:05 +00:00
|
|
|
head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
|
|
|
|
|
inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
|
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
|
|
|
LIST_REMOVE(inp, inp_hash);
|
1995-04-09 01:29:31 +00:00
|
|
|
LIST_INSERT_HEAD(head, inp, inp_hash);
|
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
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Remove PCB from various lists.
|
|
|
|
|
*/
|
1999-11-05 14:41:39 +00:00
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbremlists(struct inpcb *inp)
|
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
|
|
|
{
|
2003-11-08 23:02:36 +00:00
|
|
|
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
|
|
|
|
|
|
|
|
|
|
INP_INFO_WLOCK_ASSERT(pcbinfo);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2003-11-08 23:02:36 +00:00
|
|
|
|
|
|
|
|
inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
|
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
|
|
|
if (inp->inp_lport) {
|
|
|
|
|
struct inpcbport *phd = inp->inp_phd;
|
|
|
|
|
|
|
|
|
|
LIST_REMOVE(inp, inp_hash);
|
|
|
|
|
LIST_REMOVE(inp, inp_portlist);
|
2001-02-04 13:13:25 +00:00
|
|
|
if (LIST_FIRST(&phd->phd_pcblist) == 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
|
|
|
LIST_REMOVE(phd, phd_hash);
|
|
|
|
|
free(phd, M_PCB);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
LIST_REMOVE(inp, inp_list);
|
2003-11-08 23:02:36 +00:00
|
|
|
pcbinfo->ipi_count--;
|
1995-04-09 01:29:31 +00:00
|
|
|
}
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
/*
|
|
|
|
|
* A set label operation has occurred at the socket layer, propagate the
|
|
|
|
|
* label change into the in_pcb for the socket.
|
|
|
|
|
*/
|
|
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
in_pcbsosetlabel(struct socket *so)
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
{
|
|
|
|
|
#ifdef MAC
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
|
2006-04-01 16:04:42 +00:00
|
|
|
inp = sotoinpcb(so);
|
|
|
|
|
KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
|
2006-04-22 19:15:20 +00:00
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inp);
|
2004-06-13 02:50:07 +00:00
|
|
|
SOCK_LOCK(so);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
mac_inpcb_sosetlabel(so, inp);
|
2004-06-13 02:50:07 +00:00
|
|
|
SOCK_UNLOCK(so);
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
Introduce a MAC label reference in 'struct inpcb', which caches
the MAC label referenced from 'struct socket' in the IPv4 and
IPv6-based protocols. This permits MAC labels to be checked during
network delivery operations without dereferencing inp->inp_socket
to get to so->so_label, which will eventually avoid our having to
grab the socket lock during delivery at the network layer.
This change introduces 'struct inpcb' as a labeled object to the
MAC Framework, along with the normal circus of entry points:
initialization, creation from socket, destruction, as well as a
delivery access control check.
For most policies, the inpcb label will simply be a cache of the
socket label, so a new protocol switch method is introduced,
pr_sosetlabel() to notify protocols that the socket layer label
has been updated so that the cache can be updated while holding
appropriate locks. Most protocols implement this using
pru_sosetlabel_null(), but IPv4/IPv6 protocols using inpcbs use
the the worker function in_pcbsosetlabel(), which calls into the
MAC Framework to perform a cache update.
Biba, LOMAC, and MLS implement these entry points, as do the stub
policy, and test policy.
Reviewed by: sam, bms
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
2003-11-18 00:39:07 +00:00
|
|
|
#endif
|
|
|
|
|
}
|
2005-01-02 01:50:57 +00:00
|
|
|
|
|
|
|
|
/*
|
2006-06-02 08:18:27 +00:00
|
|
|
* ipport_tick runs once per second, determining if random port allocation
|
|
|
|
|
* should be continued. If more than ipport_randomcps ports have been
|
|
|
|
|
* allocated in the last second, then we return to sequential port
|
|
|
|
|
* allocation. We return to random allocation only once we drop below
|
|
|
|
|
* ipport_randomcps for at least ipport_randomtime seconds.
|
2005-01-02 01:50:57 +00:00
|
|
|
*/
|
|
|
|
|
void
|
2006-01-22 01:16:25 +00:00
|
|
|
ipport_tick(void *xtp)
|
2005-01-02 01:50:57 +00:00
|
|
|
{
|
2006-06-02 08:18:27 +00:00
|
|
|
|
|
|
|
|
if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
|
2005-01-02 01:50:57 +00:00
|
|
|
if (ipport_stoprandom > 0)
|
|
|
|
|
ipport_stoprandom--;
|
2006-06-02 08:18:27 +00:00
|
|
|
} else
|
|
|
|
|
ipport_stoprandom = ipport_randomtime;
|
2005-01-02 01:50:57 +00:00
|
|
|
ipport_tcplastcount = ipport_tcpallocs;
|
|
|
|
|
callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
|
|
|
|
|
}
|
2007-02-17 21:02:38 +00:00
|
|
|
|
2008-03-23 22:34:16 +00:00
|
|
|
void
|
|
|
|
|
inp_wlock(struct inpcb *inp)
|
|
|
|
|
{
|
|
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK(inp);
|
2008-03-23 22:34:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
inp_wunlock(struct inpcb *inp)
|
|
|
|
|
{
|
|
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WUNLOCK(inp);
|
2008-03-23 22:34:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
inp_rlock(struct inpcb *inp)
|
|
|
|
|
{
|
|
|
|
|
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RLOCK(inp);
|
2008-03-23 22:34:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
inp_runlock(struct inpcb *inp)
|
|
|
|
|
{
|
|
|
|
|
|
2008-04-19 14:34:38 +00:00
|
|
|
INP_RUNLOCK(inp);
|
2008-03-23 22:34:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#ifdef INVARIANTS
|
|
|
|
|
void
|
2008-03-24 20:24:04 +00:00
|
|
|
inp_lock_assert(struct inpcb *inp)
|
2008-03-23 22:34:16 +00:00
|
|
|
{
|
|
|
|
|
|
2008-04-17 21:38:18 +00:00
|
|
|
INP_WLOCK_ASSERT(inp);
|
2008-03-23 22:34:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
2008-03-24 20:24:04 +00:00
|
|
|
inp_unlock_assert(struct inpcb *inp)
|
2008-03-23 22:34:16 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
|
|
INP_UNLOCK_ASSERT(inp);
|
|
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
2007-02-17 21:02:38 +00:00
|
|
|
#ifdef DDB
|
|
|
|
|
static void
|
|
|
|
|
db_print_indent(int indent)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < indent; i++)
|
|
|
|
|
db_printf(" ");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
|
|
|
|
|
{
|
|
|
|
|
char faddr_str[48], laddr_str[48];
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("%s at %p\n", name, inc);
|
|
|
|
|
|
|
|
|
|
indent += 2;
|
|
|
|
|
|
2007-02-18 08:57:23 +00:00
|
|
|
#ifdef INET6
|
2007-02-17 21:02:38 +00:00
|
|
|
if (inc->inc_flags == 1) {
|
|
|
|
|
/* IPv6. */
|
|
|
|
|
ip6_sprintf(laddr_str, &inc->inc6_laddr);
|
|
|
|
|
ip6_sprintf(faddr_str, &inc->inc6_faddr);
|
|
|
|
|
} else {
|
2007-02-18 08:57:23 +00:00
|
|
|
#endif
|
2007-02-17 21:02:38 +00:00
|
|
|
/* IPv4. */
|
|
|
|
|
inet_ntoa_r(inc->inc_laddr, laddr_str);
|
|
|
|
|
inet_ntoa_r(inc->inc_faddr, faddr_str);
|
2007-02-18 08:57:23 +00:00
|
|
|
#ifdef INET6
|
2007-02-17 21:02:38 +00:00
|
|
|
}
|
2007-02-18 08:57:23 +00:00
|
|
|
#endif
|
2007-02-17 21:02:38 +00:00
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
|
|
|
|
|
ntohs(inc->inc_lport));
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
|
|
|
|
|
ntohs(inc->inc_fport));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
db_print_inpflags(int inp_flags)
|
|
|
|
|
{
|
|
|
|
|
int comma;
|
|
|
|
|
|
|
|
|
|
comma = 0;
|
|
|
|
|
if (inp_flags & INP_RECVOPTS) {
|
|
|
|
|
db_printf("%sINP_RECVOPTS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_RECVRETOPTS) {
|
|
|
|
|
db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_RECVDSTADDR) {
|
|
|
|
|
db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_HDRINCL) {
|
|
|
|
|
db_printf("%sINP_HDRINCL", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_HIGHPORT) {
|
|
|
|
|
db_printf("%sINP_HIGHPORT", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_LOWPORT) {
|
|
|
|
|
db_printf("%sINP_LOWPORT", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_ANONPORT) {
|
|
|
|
|
db_printf("%sINP_ANONPORT", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_RECVIF) {
|
|
|
|
|
db_printf("%sINP_RECVIF", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_MTUDISC) {
|
|
|
|
|
db_printf("%sINP_MTUDISC", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_FAITH) {
|
|
|
|
|
db_printf("%sINP_FAITH", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_RECVTTL) {
|
|
|
|
|
db_printf("%sINP_RECVTTL", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & INP_DONTFRAG) {
|
|
|
|
|
db_printf("%sINP_DONTFRAG", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_IPV6_V6ONLY) {
|
|
|
|
|
db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_PKTINFO) {
|
|
|
|
|
db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_HOPLIMIT) {
|
|
|
|
|
db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_HOPOPTS) {
|
|
|
|
|
db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_DSTOPTS) {
|
|
|
|
|
db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_RTHDR) {
|
|
|
|
|
db_printf("%sIN6P_RTHDR", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_RTHDRDSTOPTS) {
|
|
|
|
|
db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_TCLASS) {
|
|
|
|
|
db_printf("%sIN6P_TCLASS", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_AUTOFLOWLABEL) {
|
|
|
|
|
db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_RFC2292) {
|
|
|
|
|
db_printf("%sIN6P_RFC2292", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_flags & IN6P_MTU) {
|
|
|
|
|
db_printf("IN6P_MTU%s", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
|
db_print_inpvflag(u_char inp_vflag)
|
|
|
|
|
{
|
|
|
|
|
int comma;
|
|
|
|
|
|
|
|
|
|
comma = 0;
|
|
|
|
|
if (inp_vflag & INP_IPV4) {
|
|
|
|
|
db_printf("%sINP_IPV4", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_IPV6) {
|
|
|
|
|
db_printf("%sINP_IPV6", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_IPV6PROTO) {
|
|
|
|
|
db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_TIMEWAIT) {
|
|
|
|
|
db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_ONESBCAST) {
|
|
|
|
|
db_printf("%sINP_ONESBCAST", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_DROPPED) {
|
|
|
|
|
db_printf("%sINP_DROPPED", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
if (inp_vflag & INP_SOCKREF) {
|
|
|
|
|
db_printf("%sINP_SOCKREF", comma ? ", " : "");
|
|
|
|
|
comma = 1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void
|
|
|
|
|
db_print_inpcb(struct inpcb *inp, const char *name, int indent)
|
|
|
|
|
{
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("%s at %p\n", name, inp);
|
|
|
|
|
|
|
|
|
|
indent += 2;
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_flow: 0x%x\n", inp->inp_flow);
|
|
|
|
|
|
|
|
|
|
db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
|
|
|
|
|
inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_label: %p inp_flags: 0x%x (",
|
|
|
|
|
inp->inp_label, inp->inp_flags);
|
|
|
|
|
db_print_inpflags(inp->inp_flags);
|
|
|
|
|
db_printf(")\n");
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
|
|
|
|
|
inp->inp_vflag);
|
|
|
|
|
db_print_inpvflag(inp->inp_vflag);
|
|
|
|
|
db_printf(")\n");
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
|
|
|
|
|
inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
#ifdef INET6
|
|
|
|
|
if (inp->inp_vflag & INP_IPV6) {
|
|
|
|
|
db_printf("in6p_options: %p in6p_outputopts: %p "
|
|
|
|
|
"in6p_moptions: %p\n", inp->in6p_options,
|
|
|
|
|
inp->in6p_outputopts, inp->in6p_moptions);
|
|
|
|
|
db_printf("in6p_icmp6filt: %p in6p_cksum %d "
|
|
|
|
|
"in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
|
|
|
|
|
inp->in6p_hops);
|
|
|
|
|
} else
|
|
|
|
|
#endif
|
|
|
|
|
{
|
|
|
|
|
db_printf("inp_ip_tos: %d inp_ip_options: %p "
|
|
|
|
|
"inp_ip_moptions: %p\n", inp->inp_ip_tos,
|
|
|
|
|
inp->inp_options, inp->inp_moptions);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
db_print_indent(indent);
|
|
|
|
|
db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
|
|
|
|
|
(uintmax_t)inp->inp_gencnt);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
DB_SHOW_COMMAND(inpcb, db_show_inpcb)
|
|
|
|
|
{
|
|
|
|
|
struct inpcb *inp;
|
|
|
|
|
|
|
|
|
|
if (!have_addr) {
|
|
|
|
|
db_printf("usage: show inpcb <addr>\n");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
inp = (struct inpcb *)addr;
|
|
|
|
|
|
|
|
|
|
db_print_inpcb(inp, "inpcb", 0);
|
|
|
|
|
}
|
|
|
|
|
#endif
|
