freebsd-dev/sys/netinet/ip_icmp.c

1072 lines
29 KiB
C
Raw Normal View History

/*-
* SPDX-License-Identifier: BSD-3-Clause
*
1994-05-24 10:09:53 +00:00
* Copyright (c) 1982, 1986, 1988, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
1994-05-24 10:09:53 +00:00
* 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.
*
* @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/time.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/rmlock.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
1994-05-24 10:09:53 +00:00
#include <net/if.h>
#include <net/if_var.h>
#include <net/if_types.h>
1994-05-24 10:09:53 +00:00
#include <net/route.h>
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
#include <net/vnet.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
#include <netinet/in_fib.h>
#include <netinet/in_pcb.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
#include <netinet/ip.h>
#include <netinet/ip_icmp.h>
#include <netinet/ip_var.h>
#include <netinet/ip_options.h>
#include <netinet/sctp.h>
#include <netinet/tcp.h>
#include <netinet/tcp_var.h>
#include <netinet/tcpip.h>
1994-05-24 10:09:53 +00:00
#include <netinet/icmp_var.h>
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
#ifdef INET
#include <machine/in_cksum.h>
#include <security/mac/mac_framework.h>
#endif /* INET */
1994-05-24 10:09:53 +00:00
/*
* ICMP routines: error generation, receive packet processing, and
* routines to turnaround packets back to the originator, and
* host table maintenance routines.
*/
VNET_DEFINE_STATIC(int, icmplim) = 200;
#define V_icmplim VNET(icmplim)
SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
&VNET_NAME(icmplim), 0,
"Maximum number of ICMP responses per second");
VNET_DEFINE_STATIC(int, icmplim_output) = 1;
#define V_icmplim_output VNET(icmplim_output)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
&VNET_NAME(icmplim_output), 0,
"Enable logging of ICMP response rate limiting");
#ifdef INET
VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
VNET_PCPUSTAT_SYSINIT(icmpstat);
SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
#ifdef VIMAGE
VNET_PCPUSTAT_SYSUNINIT(icmpstat);
#endif /* VIMAGE */
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
#define V_icmpmaskrepl VNET(icmpmaskrepl)
SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(icmpmaskrepl), 0,
"Reply to ICMP Address Mask Request packets");
VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
#define V_icmpmaskfake VNET(icmpmaskfake)
SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(icmpmaskfake), 0,
"Fake reply to ICMP Address Mask Request packets");
VNET_DEFINE(int, drop_redirect) = 0;
Kill custom in_matroute() radix mathing function removing one rte mutex lock. Initially in_matrote() in_clsroute() in their current state was introduced by r4105 20 years ago. Instead of deleting inactive routes immediately, we kept them in route table, setting RTPRF_OURS flag and some expire time. After that, either GC came or RTPRF_OURS got removed on first-packet. It was a good solution in that days (and probably another decade after that) to keep TCP metrics. However, after moving metrics to TCP hostcache in r122922, most of in_rmx functionality became unused. It might had been used for flushing icmp-originated routes before rte mutexes/refcounting, but I'm not sure about that. So it looks like this is nearly impossible to make GC do its work nowadays: in_rtkill() ignores non-RTPRF_OURS routes. route can only become RTPRF_OURS after dropping last reference via rtfree() which calls in_clsroute(), which, it turn, ignores UP and non-RTF_DYNAMIC routes. Dynamic routes can still be installed via received redirect, but they have default lifetime (no specific rt_expire) and no one has another trie walker to call RTFREE() on them. So, the changelist: * remove custom rnh_match / rnh_close matching function. * remove all GC functions * partially revert r256695 (proto3 is no more used inside kernel, it is not possible to use rt_expire from user point of view, proto3 support is not complete) * Finish r241884 (similar to this commit) and remove remaining IPv6 parts MFC after: 1 month
2014-11-11 02:52:40 +00:00
#define V_drop_redirect VNET(drop_redirect)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
&VNET_NAME(drop_redirect), 0,
"Ignore ICMP redirects");
VNET_DEFINE_STATIC(int, log_redirect) = 0;
#define V_log_redirect VNET(log_redirect)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(log_redirect), 0,
"Log ICMP redirects to the console");
1999-09-14 16:40:28 +00:00
VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
#define V_reply_src VNET(reply_src)
SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(reply_src), IFNAMSIZ,
"ICMP reply source for non-local packets");
VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
#define V_icmp_rfi VNET(icmp_rfi)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(icmp_rfi), 0,
"ICMP reply from incoming interface for non-local packets");
/* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
#define V_icmp_quotelen VNET(icmp_quotelen)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(icmp_quotelen), 0,
"Number of bytes from original packet to quote in ICMP reply");
VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
#define V_icmpbmcastecho VNET(icmpbmcastecho)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator (DPCPU), as suggested by Peter Wemm, and implement a new per-virtual network stack memory allocator. Modify vnet to use the allocator instead of monolithic global container structures (vinet, ...). This change solves many binary compatibility problems associated with VIMAGE, and restores ELF symbols for virtualized global variables. Each virtualized global variable exists as a "reference copy", and also once per virtual network stack. Virtualized global variables are tagged at compile-time, placing the in a special linker set, which is loaded into a contiguous region of kernel memory. Virtualized global variables in the base kernel are linked as normal, but those in modules are copied and relocated to a reserved portion of the kernel's vnet region with the help of a the kernel linker. Virtualized global variables exist in per-vnet memory set up when the network stack instance is created, and are initialized statically from the reference copy. Run-time access occurs via an accessor macro, which converts from the current vnet and requested symbol to a per-vnet address. When "options VIMAGE" is not compiled into the kernel, normal global ELF symbols will be used instead and indirection is avoided. This change restores static initialization for network stack global variables, restores support for non-global symbols and types, eliminates the need for many subsystem constructors, eliminates large per-subsystem structures that caused many binary compatibility issues both for monitoring applications (netstat) and kernel modules, removes the per-function INIT_VNET_*() macros throughout the stack, eliminates the need for vnet_symmap ksym(2) munging, and eliminates duplicate definitions of virtualized globals under VIMAGE_GLOBALS. Bump __FreeBSD_version and update UPDATING. Portions submitted by: bz Reviewed by: bz, zec Discussed with: gnn, jamie, jeff, jhb, julian, sam Suggested by: peter Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
&VNET_NAME(icmpbmcastecho), 0,
"Reply to multicast ICMP Echo Request and Timestamp packets");
VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
#define V_icmptstamprepl VNET(icmptstamprepl)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
&VNET_NAME(icmptstamprepl), 0,
"Respond to ICMP Timestamp packets");
VNET_DEFINE_STATIC(int, error_keeptags) = 0;
#define V_error_keeptags VNET(error_keeptags)
SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
&VNET_NAME(error_keeptags), 0,
"ICMP error response keeps copy of mbuf_tags of original packet");
1994-05-24 10:09:53 +00:00
#ifdef ICMPPRINTFS
int icmpprintfs = 0;
#endif
2002-03-19 21:25:46 +00:00
static void icmp_reflect(struct mbuf *);
static void icmp_send(struct mbuf *, struct mbuf *);
1994-05-24 10:09:53 +00:00
extern struct protosw inetsw[];
/*
* Kernel module interface for updating icmpstat. The argument is an index
* into icmpstat treated as an array of u_long. While this encodes the
* general layout of icmpstat into the caller, it doesn't encode its
* location, so that future changes to add, for example, per-CPU stats
* support won't cause binary compatibility problems for kernel modules.
*/
void
kmod_icmpstat_inc(int statnum)
{
counter_u64_add(VNET(icmpstat)[statnum], 1);
}
1994-05-24 10:09:53 +00:00
/*
* Generate an error packet of type error
* in response to bad packet ip.
*/
void
icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
1994-05-24 10:09:53 +00:00
{
struct ip *oip, *nip;
struct icmp *icp;
struct mbuf *m;
unsigned icmplen, icmpelen, nlen, oiphlen;
KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
__func__));
1994-05-24 10:09:53 +00:00
if (type != ICMP_REDIRECT)
ICMPSTAT_INC(icps_error);
1994-05-24 10:09:53 +00:00
/*
* Don't send error:
* if the original packet was encrypted.
* if not the first fragment of message.
* in response to a multicast or broadcast packet.
* if the old packet protocol was an ICMP error message.
1994-05-24 10:09:53 +00:00
*/
if (n->m_flags & M_DECRYPTED)
goto freeit;
if (n->m_flags & (M_BCAST|M_MCAST))
1994-05-24 10:09:53 +00:00
goto freeit;
/* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
goto freeit;
oip = mtod(n, struct ip *);
oiphlen = oip->ip_hl << 2;
if (n->m_len < oiphlen + ICMP_MINLEN)
goto freeit;
#ifdef ICMPPRINTFS
if (icmpprintfs)
printf("icmp_error(%p, %x, %d)\n", oip, type, code);
#endif
if (oip->ip_off & htons(~(IP_MF|IP_DF)))
goto freeit;
1994-05-24 10:09:53 +00:00
if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
!ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
oiphlen))->icmp_type)) {
ICMPSTAT_INC(icps_oldicmp);
1994-05-24 10:09:53 +00:00
goto freeit;
}
/*
* Calculate length to quote from original packet and
* prevent the ICMP mbuf from overflowing.
* Unfortunately this is non-trivial since ip_forward()
* sends us truncated packets.
*/
nlen = m_length(n, NULL);
if (oip->ip_p == IPPROTO_TCP) {
struct tcphdr *th;
int tcphlen;
if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
n->m_next == NULL)
goto stdreply;
if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
(n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
goto freeit;
oip = mtod(n, struct ip *);
th = mtodo(n, oiphlen);
tcphlen = th->th_off << 2;
if (tcphlen < sizeof(struct tcphdr))
goto freeit;
if (ntohs(oip->ip_len) < oiphlen + tcphlen)
goto freeit;
if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
goto stdreply;
if (n->m_len < oiphlen + tcphlen &&
(n = m_pullup(n, oiphlen + tcphlen)) == NULL)
goto freeit;
oip = mtod(n, struct ip *);
icmpelen = max(tcphlen, min(V_icmp_quotelen,
ntohs(oip->ip_len) - oiphlen));
} else if (oip->ip_p == IPPROTO_SCTP) {
struct sctphdr *sh;
struct sctp_chunkhdr *ch;
if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
goto stdreply;
if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
n->m_next == NULL)
goto stdreply;
if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
(n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
goto freeit;
oip = mtod(n, struct ip *);
icmpelen = max(sizeof(struct sctphdr),
min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
sh = mtodo(n, oiphlen);
if (ntohl(sh->v_tag) == 0 &&
ntohs(oip->ip_len) >= oiphlen +
sizeof(struct sctphdr) + 8 &&
(n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
n->m_next != NULL)) {
if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
(n = m_pullup(n, oiphlen +
sizeof(struct sctphdr) + 8)) == NULL)
goto freeit;
oip = mtod(n, struct ip *);
sh = mtodo(n, oiphlen);
ch = (struct sctp_chunkhdr *)(sh + 1);
if (ch->chunk_type == SCTP_INITIATION) {
icmpelen = max(sizeof(struct sctphdr) + 8,
min(V_icmp_quotelen, ntohs(oip->ip_len) -
oiphlen));
}
}
} else
stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
oiphlen));
icmplen = min(oiphlen + icmpelen, nlen);
if (icmplen < sizeof(struct ip))
goto freeit;
if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
m = m_gethdr(M_NOWAIT, MT_DATA);
else
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
if (m == NULL)
goto freeit;
#ifdef MAC
mac_netinet_icmp_reply(n, m);
#endif
icmplen = min(icmplen, M_TRAILINGSPACE(m) -
sizeof(struct ip) - ICMP_MINLEN);
m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
m->m_data += sizeof(struct ip);
m->m_len = ICMP_MINLEN + icmplen;
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
/* XXX MRT make the outgoing packet use the same FIB
* that was associated with the incoming packet
*/
M_SETFIB(m, M_GETFIB(n));
1994-05-24 10:09:53 +00:00
icp = mtod(m, struct icmp *);
ICMPSTAT_INC(icps_outhist[type]);
1994-05-24 10:09:53 +00:00
icp->icmp_type = type;
if (type == ICMP_REDIRECT)
icp->icmp_gwaddr.s_addr = dest;
else {
icp->icmp_void = 0;
1995-05-30 08:16:23 +00:00
/*
1994-05-24 10:09:53 +00:00
* The following assignments assume an overlay with the
* just zeroed icmp_void field.
1994-05-24 10:09:53 +00:00
*/
if (type == ICMP_PARAMPROB) {
icp->icmp_pptr = code;
code = 0;
} else if (type == ICMP_UNREACH &&
code == ICMP_UNREACH_NEEDFRAG && mtu) {
icp->icmp_nextmtu = htons(mtu);
1994-05-24 10:09:53 +00:00
}
}
icp->icmp_code = code;
Fixed broken ICMP error generation, unified conversion of IP header fields between host and network byte order. The details: o icmp_error() now does not add IP header length. This fixes the problem when icmp_error() is called from ip_forward(). In this case the ip_len of the original IP datagram returned with ICMP error was wrong. o icmp_error() expects all three fields, ip_len, ip_id and ip_off in host byte order, so DTRT and convert these fields back to network byte order before sending a message. This fixes the problem described in PR 16240 and PR 20877 (ip_id field was returned in host byte order). o ip_ttl decrement operation in ip_forward() was moved down to make sure that it does not corrupt the copy of original IP datagram passed later to icmp_error(). o A copy of original IP datagram in ip_forward() was made a read-write, independent copy. This fixes the problem I first reported to Garrett Wollman and Bill Fenner and later put in audit trail of PR 16240: ip_output() (not always) converts fields of original datagram to network byte order, but because copy (mcopy) and its original (m) most likely share the same mbuf cluster, ip_output()'s manipulations on original also corrupted the copy. o ip_output() now expects all three fields, ip_len, ip_off and (what is significant) ip_id in host byte order. It was a headache for years that ip_id was handled differently. The only compatibility issue here is the raw IP socket interface with IP_HDRINCL socket option set and a non-zero ip_id field, but ip.4 manual page was unclear on whether in this case ip_id field should be in host or network byte order.
2000-09-01 12:33:03 +00:00
/*
* Copy the quotation into ICMP message and
* convert quoted IP header back to network representation.
Fixed broken ICMP error generation, unified conversion of IP header fields between host and network byte order. The details: o icmp_error() now does not add IP header length. This fixes the problem when icmp_error() is called from ip_forward(). In this case the ip_len of the original IP datagram returned with ICMP error was wrong. o icmp_error() expects all three fields, ip_len, ip_id and ip_off in host byte order, so DTRT and convert these fields back to network byte order before sending a message. This fixes the problem described in PR 16240 and PR 20877 (ip_id field was returned in host byte order). o ip_ttl decrement operation in ip_forward() was moved down to make sure that it does not corrupt the copy of original IP datagram passed later to icmp_error(). o A copy of original IP datagram in ip_forward() was made a read-write, independent copy. This fixes the problem I first reported to Garrett Wollman and Bill Fenner and later put in audit trail of PR 16240: ip_output() (not always) converts fields of original datagram to network byte order, but because copy (mcopy) and its original (m) most likely share the same mbuf cluster, ip_output()'s manipulations on original also corrupted the copy. o ip_output() now expects all three fields, ip_len, ip_off and (what is significant) ip_id in host byte order. It was a headache for years that ip_id was handled differently. The only compatibility issue here is the raw IP socket interface with IP_HDRINCL socket option set and a non-zero ip_id field, but ip.4 manual page was unclear on whether in this case ip_id field should be in host or network byte order.
2000-09-01 12:33:03 +00:00
*/
m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
nip = &icp->icmp_ip;
1994-05-24 10:09:53 +00:00
/*
* Set up ICMP message mbuf and copy old IP header (without options
* in front of ICMP message.
* If the original mbuf was meant to bypass the firewall, the error
* reply should bypass as well.
*/
m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
("insufficient space for ip header"));
1994-05-24 10:09:53 +00:00
m->m_data -= sizeof(struct ip);
m->m_len += sizeof(struct ip);
m->m_pkthdr.len = m->m_len;
m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
nip = mtod(m, struct ip *);
bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
nip->ip_len = htons(m->m_len);
nip->ip_v = IPVERSION;
nip->ip_hl = 5;
1994-05-24 10:09:53 +00:00
nip->ip_p = IPPROTO_ICMP;
nip->ip_tos = 0;
nip->ip_off = 0;
if (V_error_keeptags)
m_tag_copy_chain(m, n, M_NOWAIT);
1994-05-24 10:09:53 +00:00
icmp_reflect(m);
freeit:
m_freem(n);
}
/*
* Process a received ICMP message.
*/
int
icmp_input(struct mbuf **mp, int *offp, int proto)
1994-05-24 10:09:53 +00:00
{
struct epoch_tracker et;
struct icmp *icp;
struct in_ifaddr *ia;
struct mbuf *m = *mp;
struct ip *ip = mtod(m, struct ip *);
struct sockaddr_in icmpsrc, icmpdst, icmpgw;
int hlen = *offp;
int icmplen = ntohs(ip->ip_len) - *offp;
int i, code;
2002-03-19 21:25:46 +00:00
void (*ctlfunc)(int, struct sockaddr *, void *);
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
int fibnum;
1994-05-24 10:09:53 +00:00
*mp = NULL;
1994-05-24 10:09:53 +00:00
/*
* Locate icmp structure in mbuf, and check
* that not corrupted and of at least minimum length.
*/
#ifdef ICMPPRINTFS
if (icmpprintfs) {
char srcbuf[INET_ADDRSTRLEN];
char dstbuf[INET_ADDRSTRLEN];
printf("icmp_input from %s to %s, len %d\n",
inet_ntoa_r(ip->ip_src, srcbuf),
inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
}
1994-05-24 10:09:53 +00:00
#endif
NET_EPOCH_ENTER(et);
1994-05-24 10:09:53 +00:00
if (icmplen < ICMP_MINLEN) {
ICMPSTAT_INC(icps_tooshort);
1994-05-24 10:09:53 +00:00
goto freeit;
}
i = hlen + min(icmplen, ICMP_ADVLENMIN);
if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
ICMPSTAT_INC(icps_tooshort);
NET_EPOCH_EXIT(et);
return (IPPROTO_DONE);
1994-05-24 10:09:53 +00:00
}
ip = mtod(m, struct ip *);
m->m_len -= hlen;
m->m_data += hlen;
icp = mtod(m, struct icmp *);
if (in_cksum(m, icmplen)) {
ICMPSTAT_INC(icps_checksum);
1994-05-24 10:09:53 +00:00
goto freeit;
}
m->m_len += hlen;
m->m_data -= hlen;
#ifdef ICMPPRINTFS
if (icmpprintfs)
printf("icmp_input, type %d code %d\n", icp->icmp_type,
icp->icmp_code);
#endif
/*
* Message type specific processing.
*/
1994-05-24 10:09:53 +00:00
if (icp->icmp_type > ICMP_MAXTYPE)
goto raw;
/* Initialize */
bzero(&icmpsrc, sizeof(icmpsrc));
icmpsrc.sin_len = sizeof(struct sockaddr_in);
icmpsrc.sin_family = AF_INET;
bzero(&icmpdst, sizeof(icmpdst));
icmpdst.sin_len = sizeof(struct sockaddr_in);
icmpdst.sin_family = AF_INET;
bzero(&icmpgw, sizeof(icmpgw));
icmpgw.sin_len = sizeof(struct sockaddr_in);
icmpgw.sin_family = AF_INET;
ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
1994-05-24 10:09:53 +00:00
code = icp->icmp_code;
switch (icp->icmp_type) {
case ICMP_UNREACH:
switch (code) {
case ICMP_UNREACH_NET:
case ICMP_UNREACH_HOST:
case ICMP_UNREACH_SRCFAIL:
case ICMP_UNREACH_NET_UNKNOWN:
case ICMP_UNREACH_HOST_UNKNOWN:
case ICMP_UNREACH_ISOLATED:
case ICMP_UNREACH_TOSNET:
case ICMP_UNREACH_TOSHOST:
case ICMP_UNREACH_HOST_PRECEDENCE:
case ICMP_UNREACH_PRECEDENCE_CUTOFF:
code = PRC_UNREACH_NET;
1994-05-24 10:09:53 +00:00
break;
case ICMP_UNREACH_NEEDFRAG:
code = PRC_MSGSIZE;
break;
1995-05-30 08:16:23 +00:00
/*
* RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
* Treat subcodes 2,3 as immediate RST
*/
case ICMP_UNREACH_PROTOCOL:
code = PRC_UNREACH_PROTOCOL;
break;
case ICMP_UNREACH_PORT:
code = PRC_UNREACH_PORT;
break;
We currently does not react to ICMP administratively prohibited messages send by routers when they deny our traffic, this causes a timeout when trying to connect to TCP ports/services on a remote host, which is blocked by routers or firewalls. rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually requi re that we treat such a message for a TCP session, that we treat it like if we had recieved a RST. quote begin. A Destination Unreachable message that is received MUST be reported to the transport layer. The transport layer SHOULD use the information appropriately; for example, see Sections 4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol that has its own mechanism for notifying the sender that a port is unreachable (e.g., TCP, which sends RST segments) MUST nevertheless accept an ICMP Port Unreachable for the same purpose. quote end. I've written a small extension that implement this, it also create a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if this new behaviour is activated. When it's activated (set to 1) we'll treat a ICMP administratively prohibited message (icmp type 3 code 9, 10 and 13) for a TCP sessions, as if we recived a TCP RST, but only if the TCP session is in SYN_SENT state. The reason for only reacting when in SYN_SENT state, is that this will solve the problem, and at the same time minimize the risk of this being abused. I suggest that we enable this new behaviour by default, but it would be a change of current behaviour, so if people prefer to leave it disabled by default, at least for now, this would be ok for me, the attached diff actually have the sysctl set to 0 by default. PR: 23086 Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
case ICMP_UNREACH_NET_PROHIB:
1994-05-24 10:09:53 +00:00
case ICMP_UNREACH_HOST_PROHIB:
case ICMP_UNREACH_FILTER_PROHIB:
code = PRC_UNREACH_ADMIN_PROHIB;
break;
We currently does not react to ICMP administratively prohibited messages send by routers when they deny our traffic, this causes a timeout when trying to connect to TCP ports/services on a remote host, which is blocked by routers or firewalls. rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually requi re that we treat such a message for a TCP session, that we treat it like if we had recieved a RST. quote begin. A Destination Unreachable message that is received MUST be reported to the transport layer. The transport layer SHOULD use the information appropriately; for example, see Sections 4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol that has its own mechanism for notifying the sender that a port is unreachable (e.g., TCP, which sends RST segments) MUST nevertheless accept an ICMP Port Unreachable for the same purpose. quote end. I've written a small extension that implement this, it also create a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if this new behaviour is activated. When it's activated (set to 1) we'll treat a ICMP administratively prohibited message (icmp type 3 code 9, 10 and 13) for a TCP sessions, as if we recived a TCP RST, but only if the TCP session is in SYN_SENT state. The reason for only reacting when in SYN_SENT state, is that this will solve the problem, and at the same time minimize the risk of this being abused. I suggest that we enable this new behaviour by default, but it would be a change of current behaviour, so if people prefer to leave it disabled by default, at least for now, this would be ok for me, the attached diff actually have the sysctl set to 0 by default. PR: 23086 Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
1994-05-24 10:09:53 +00:00
default:
goto badcode;
}
goto deliver;
case ICMP_TIMXCEED:
if (code > 1)
goto badcode;
code += PRC_TIMXCEED_INTRANS;
goto deliver;
case ICMP_PARAMPROB:
if (code > 1)
goto badcode;
code = PRC_PARAMPROB;
deliver:
/*
* Problem with datagram; advise higher level routines.
*/
if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
ICMPSTAT_INC(icps_badlen);
1994-05-24 10:09:53 +00:00
goto freeit;
}
/* Discard ICMP's in response to multicast packets */
if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
goto badcode;
1994-05-24 10:09:53 +00:00
#ifdef ICMPPRINTFS
if (icmpprintfs)
printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
#endif
icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
/*
* XXX if the packet contains [IPv4 AH TCP], we can't make a
* notification to TCP layer.
*/
i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
ip_stripoptions(m);
if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
/* This should actually not happen */
ICMPSTAT_INC(icps_tooshort);
NET_EPOCH_EXIT(et);
return (IPPROTO_DONE);
}
ip = mtod(m, struct ip *);
icp = (struct icmp *)(ip + 1);
/*
* The upper layer handler can rely on:
* - The outer IP header has no options.
* - The outer IP header, the ICMP header, the inner IP header,
* and the first n bytes of the inner payload are contiguous.
* n is at least 8, but might be larger based on
* ICMP_ADVLENPREF. See its definition in ip_icmp.h.
*/
ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
if (ctlfunc)
1994-05-24 10:09:53 +00:00
(*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
(void *)&icp->icmp_ip);
1994-05-24 10:09:53 +00:00
break;
badcode:
ICMPSTAT_INC(icps_badcode);
1994-05-24 10:09:53 +00:00
break;
case ICMP_ECHO:
if (!V_icmpbmcastecho
&& (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
ICMPSTAT_INC(icps_bmcastecho);
break;
}
if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
goto freeit;
icp->icmp_type = ICMP_ECHOREPLY;
goto reflect;
1994-05-24 10:09:53 +00:00
case ICMP_TSTAMP:
if (V_icmptstamprepl == 0)
break;
if (!V_icmpbmcastecho
&& (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
ICMPSTAT_INC(icps_bmcasttstamp);
break;
}
1994-05-24 10:09:53 +00:00
if (icmplen < ICMP_TSLEN) {
ICMPSTAT_INC(icps_badlen);
1994-05-24 10:09:53 +00:00
break;
}
if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
goto freeit;
1994-05-24 10:09:53 +00:00
icp->icmp_type = ICMP_TSTAMPREPLY;
icp->icmp_rtime = iptime();
icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
goto reflect;
1995-05-30 08:16:23 +00:00
1994-05-24 10:09:53 +00:00
case ICMP_MASKREQ:
if (V_icmpmaskrepl == 0)
1994-05-24 10:09:53 +00:00
break;
/*
* We are not able to respond with all ones broadcast
* unless we receive it over a point-to-point interface.
*/
if (icmplen < ICMP_MASKLEN)
break;
switch (ip->ip_dst.s_addr) {
case INADDR_BROADCAST:
case INADDR_ANY:
icmpdst.sin_addr = ip->ip_src;
break;
default:
icmpdst.sin_addr = ip->ip_dst;
}
ia = (struct in_ifaddr *)ifaof_ifpforaddr(
(struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
if (ia == NULL)
1994-05-24 10:09:53 +00:00
break;
if (ia->ia_ifp == NULL)
break;
1994-05-24 10:09:53 +00:00
icp->icmp_type = ICMP_MASKREPLY;
if (V_icmpmaskfake == 0)
icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
else
icp->icmp_mask = V_icmpmaskfake;
1994-05-24 10:09:53 +00:00
if (ip->ip_src.s_addr == 0) {
if (ia->ia_ifp->if_flags & IFF_BROADCAST)
ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
}
reflect:
ICMPSTAT_INC(icps_reflect);
ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
1994-05-24 10:09:53 +00:00
icmp_reflect(m);
NET_EPOCH_EXIT(et);
return (IPPROTO_DONE);
1994-05-24 10:09:53 +00:00
case ICMP_REDIRECT:
if (V_log_redirect) {
u_long src, dst, gw;
src = ntohl(ip->ip_src.s_addr);
dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
gw = ntohl(icp->icmp_gwaddr.s_addr);
printf("icmp redirect from %d.%d.%d.%d: "
"%d.%d.%d.%d => %d.%d.%d.%d\n",
(int)(src >> 24), (int)((src >> 16) & 0xff),
(int)((src >> 8) & 0xff), (int)(src & 0xff),
(int)(dst >> 24), (int)((dst >> 16) & 0xff),
(int)((dst >> 8) & 0xff), (int)(dst & 0xff),
(int)(gw >> 24), (int)((gw >> 16) & 0xff),
(int)((gw >> 8) & 0xff), (int)(gw & 0xff));
}
/*
* RFC1812 says we must ignore ICMP redirects if we
* are acting as router.
*/
if (V_drop_redirect || V_ipforwarding)
break;
1994-05-24 10:09:53 +00:00
if (code > 3)
goto badcode;
if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
ICMPSTAT_INC(icps_badlen);
1994-05-24 10:09:53 +00:00
break;
}
/*
* Short circuit routing redirects to force
* immediate change in the kernel's routing
* tables. The message is also handed to anyone
* listening on a raw socket (e.g. the routing
* daemon for use in updating its tables).
*/
icmpgw.sin_addr = ip->ip_src;
icmpdst.sin_addr = icp->icmp_gwaddr;
#ifdef ICMPPRINTFS
if (icmpprintfs) {
char dstbuf[INET_ADDRSTRLEN];
char gwbuf[INET_ADDRSTRLEN];
printf("redirect dst %s to %s\n",
inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
}
1994-05-24 10:09:53 +00:00
#endif
icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
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
for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
in_rtredirect((struct sockaddr *)&icmpsrc,
(struct sockaddr *)&icmpdst,
(struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
(struct sockaddr *)&icmpgw, fibnum);
}
1994-05-24 10:09:53 +00:00
pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
break;
/*
* No kernel processing for the following;
* just fall through to send to raw listener.
*/
case ICMP_ECHOREPLY:
case ICMP_ROUTERADVERT:
case ICMP_ROUTERSOLICIT:
case ICMP_TSTAMPREPLY:
case ICMP_IREQREPLY:
case ICMP_MASKREPLY:
case ICMP_SOURCEQUENCH:
1994-05-24 10:09:53 +00:00
default:
break;
}
raw:
NET_EPOCH_EXIT(et);
*mp = m;
rip_input(mp, offp, proto);
return (IPPROTO_DONE);
1994-05-24 10:09:53 +00:00
freeit:
NET_EPOCH_EXIT(et);
1994-05-24 10:09:53 +00:00
m_freem(m);
return (IPPROTO_DONE);
1994-05-24 10:09:53 +00:00
}
/*
* Reflect the ip packet back to the source
*/
static void
icmp_reflect(struct mbuf *m)
1994-05-24 10:09:53 +00:00
{
struct rm_priotracker in_ifa_tracker;
struct epoch_tracker et;
struct ip *ip = mtod(m, struct ip *);
struct ifaddr *ifa;
struct ifnet *ifp;
struct in_ifaddr *ia;
1994-05-24 10:09:53 +00:00
struct in_addr t;
struct nhop4_extended nh_ext;
struct mbuf *opts = NULL;
int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
1994-05-24 10:09:53 +00:00
if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
1994-05-24 10:09:53 +00:00
m_freem(m); /* Bad return address */
ICMPSTAT_INC(icps_badaddr);
1994-05-24 10:09:53 +00:00
goto done; /* Ip_output() will check for broadcast */
}
1994-05-24 10:09:53 +00:00
t = ip->ip_dst;
ip->ip_dst = ip->ip_src;
1994-05-24 10:09:53 +00:00
/*
* Source selection for ICMP replies:
*
* If the incoming packet was addressed directly to one of our
* own addresses, use dst as the src for the reply.
1994-05-24 10:09:53 +00:00
*/
IN_IFADDR_RLOCK(&in_ifa_tracker);
LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
t = IA_SIN(ia)->sin_addr;
IN_IFADDR_RUNLOCK(&in_ifa_tracker);
goto match;
}
}
IN_IFADDR_RUNLOCK(&in_ifa_tracker);
/*
* If the incoming packet was addressed to one of our broadcast
* addresses, use the first non-broadcast address which corresponds
* to the incoming interface.
*/
ifp = m->m_pkthdr.rcvif;
if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
NET_EPOCH_ENTER(et);
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
if (ifa->ifa_addr->sa_family != AF_INET)
continue;
2001-12-14 19:32:47 +00:00
ia = ifatoia(ifa);
if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
t.s_addr) {
t = IA_SIN(ia)->sin_addr;
NET_EPOCH_EXIT(et);
2001-12-14 19:32:47 +00:00
goto match;
}
2001-12-14 19:32:47 +00:00
}
NET_EPOCH_EXIT(et);
2001-12-14 19:32:47 +00:00
}
/*
* If the packet was transiting through us, use the address of
* the interface the packet came through in. If that interface
* doesn't have a suitable IP address, the normal selection
* criteria apply.
*/
if (V_icmp_rfi && ifp != NULL) {
NET_EPOCH_ENTER(et);
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
if (ifa->ifa_addr->sa_family != AF_INET)
continue;
ia = ifatoia(ifa);
t = IA_SIN(ia)->sin_addr;
NET_EPOCH_EXIT(et);
goto match;
}
NET_EPOCH_EXIT(et);
}
/*
* If the incoming packet was not addressed directly to us, use
* designated interface for icmp replies specified by sysctl
* net.inet.icmp.reply_src (default not set). Otherwise continue
* with normal source selection.
*/
if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
NET_EPOCH_ENTER(et);
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
if (ifa->ifa_addr->sa_family != AF_INET)
continue;
ia = ifatoia(ifa);
t = IA_SIN(ia)->sin_addr;
NET_EPOCH_EXIT(et);
goto match;
}
NET_EPOCH_EXIT(et);
}
/*
* If the packet was transiting through us, use the address of
* the interface that is the closest to the packet source.
* When we don't have a route back to the packet source, stop here
* and drop the packet.
*/
if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) {
m_freem(m);
ICMPSTAT_INC(icps_noroute);
goto done;
}
t = nh_ext.nh_src;
match:
#ifdef MAC
mac_netinet_icmp_replyinplace(m);
#endif
1994-05-24 10:09:53 +00:00
ip->ip_src = t;
ip->ip_ttl = V_ip_defttl;
1994-05-24 10:09:53 +00:00
if (optlen > 0) {
u_char *cp;
1994-05-24 10:09:53 +00:00
int opt, cnt;
u_int len;
/*
* Retrieve any source routing from the incoming packet;
* add on any record-route or timestamp options.
*/
cp = (u_char *) (ip + 1);
if ((opts = ip_srcroute(m)) == NULL &&
(opts = m_gethdr(M_NOWAIT, MT_DATA))) {
1994-05-24 10:09:53 +00:00
opts->m_len = sizeof(struct in_addr);
mtod(opts, struct in_addr *)->s_addr = 0;
}
if (opts) {
#ifdef ICMPPRINTFS
if (icmpprintfs)
printf("icmp_reflect optlen %d rt %d => ",
optlen, opts->m_len);
#endif
for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
opt = cp[IPOPT_OPTVAL];
if (opt == IPOPT_EOL)
break;
if (opt == IPOPT_NOP)
len = 1;
else {
if (cnt < IPOPT_OLEN + sizeof(*cp))
break;
1994-05-24 10:09:53 +00:00
len = cp[IPOPT_OLEN];
if (len < IPOPT_OLEN + sizeof(*cp) ||
len > cnt)
1994-05-24 10:09:53 +00:00
break;
}
/*
* Should check for overflow, but it "can't happen"
*/
1995-05-30 08:16:23 +00:00
if (opt == IPOPT_RR || opt == IPOPT_TS ||
1994-05-24 10:09:53 +00:00
opt == IPOPT_SECURITY) {
bcopy((caddr_t)cp,
mtod(opts, caddr_t) + opts->m_len, len);
opts->m_len += len;
}
}
/* Terminate & pad, if necessary */
cnt = opts->m_len % 4;
if (cnt) {
1994-05-24 10:09:53 +00:00
for (; cnt < 4; cnt++) {
*(mtod(opts, caddr_t) + opts->m_len) =
IPOPT_EOL;
opts->m_len++;
}
}
#ifdef ICMPPRINTFS
if (icmpprintfs)
printf("%d\n", opts->m_len);
#endif
}
ip_stripoptions(m);
1994-05-24 10:09:53 +00:00
}
m_tag_delete_nonpersistent(m);
1994-05-24 10:09:53 +00:00
m->m_flags &= ~(M_BCAST|M_MCAST);
icmp_send(m, opts);
1994-05-24 10:09:53 +00:00
done:
if (opts)
(void)m_free(opts);
}
/*
* Send an icmp packet back to the ip level,
* after supplying a checksum.
*/
static void
icmp_send(struct mbuf *m, struct mbuf *opts)
1994-05-24 10:09:53 +00:00
{
struct ip *ip = mtod(m, struct ip *);
int hlen;
struct icmp *icp;
1994-05-24 10:09:53 +00:00
hlen = ip->ip_hl << 2;
1994-05-24 10:09:53 +00:00
m->m_data += hlen;
m->m_len -= hlen;
icp = mtod(m, struct icmp *);
icp->icmp_cksum = 0;
icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
1994-05-24 10:09:53 +00:00
m->m_data -= hlen;
m->m_len += hlen;
m->m_pkthdr.rcvif = (struct ifnet *)0;
1994-05-24 10:09:53 +00:00
#ifdef ICMPPRINTFS
if (icmpprintfs) {
char dstbuf[INET_ADDRSTRLEN];
char srcbuf[INET_ADDRSTRLEN];
printf("icmp_send dst %s src %s\n",
inet_ntoa_r(ip->ip_dst, dstbuf),
inet_ntoa_r(ip->ip_src, srcbuf));
}
1994-05-24 10:09:53 +00:00
#endif
(void) ip_output(m, opts, NULL, 0, NULL, NULL);
1994-05-24 10:09:53 +00:00
}
/*
* Return milliseconds since 00:00 UTC in network format.
*/
uint32_t
iptime(void)
1994-05-24 10:09:53 +00:00
{
struct timeval atv;
u_long t;
getmicrotime(&atv);
1994-05-24 10:09:53 +00:00
t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
return (htonl(t));
}
/*
* Return the next larger or smaller MTU plateau (table from RFC 1191)
* given current value MTU. If DIR is less than zero, a larger plateau
* is returned; otherwise, a smaller value is returned.
*/
int
ip_next_mtu(int mtu, int dir)
{
static int mtutab[] = {
65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
296, 68, 0
};
int i, size;
size = (sizeof mtutab) / (sizeof mtutab[0]);
if (dir >= 0) {
2006-01-23 20:10:49 +00:00
for (i = 0; i < size; i++)
if (mtu > mtutab[i])
return mtutab[i];
} else {
for (i = size - 1; i >= 0; i--)
if (mtu < mtutab[i])
return mtutab[i];
if (mtu == mtutab[0])
return mtutab[0];
}
return 0;
}
#endif /* INET */
/*
* badport_bandlim() - check for ICMP bandwidth limit
*
* Return 0 if it is ok to send an ICMP error response, -1 if we have
* hit our bandwidth limit and it is not ok.
*
* If icmplim is <= 0, the feature is disabled and 0 is returned.
*
* For now we separate the TCP and UDP subsystems w/ different 'which'
* values. We may eventually remove this separation (and simplify the
* code further).
*
* Note that the printing of the error message is delayed so we can
* properly print the icmp error rate that the system was trying to do
* (i.e. 22000/100 pps, etc...). This can cause long delays in printing
* the 'final' error, but it doesn't make sense to solve the printing
* delay with more complex code.
*/
struct icmp_rate {
const char *descr;
struct counter_rate cr;
};
VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
{ "icmp unreach response" },
{ "icmp ping response" },
{ "icmp tstamp response" },
{ "closed port RST response" },
{ "open port RST response" },
{ "icmp6 unreach response" },
{ "sctp ootb response" }
};
#define V_icmp_rates VNET(icmp_rates)
static void
icmp_bandlimit_init(void)
{
for (int i = 0; i < BANDLIM_MAX; i++) {
V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
V_icmp_rates[i].cr.cr_ticks = ticks;
}
}
VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
icmp_bandlimit_init, NULL);
static void
icmp_bandlimit_uninit(void)
{
for (int i = 0; i < BANDLIM_MAX; i++)
counter_u64_free(V_icmp_rates[i].cr.cr_rate);
}
VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
icmp_bandlimit_uninit, NULL);
int
badport_bandlim(int which)
{
int64_t pps;
if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
return (0);
KASSERT(which >= 0 && which < BANDLIM_MAX,
("%s: which %d", __func__, which));
pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim);
if (pps == -1)
return (-1);
if (pps > 0 && V_icmplim_output)
log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim);
return (0);
}