freebsd-dev/sys/netinet/ip_divert.c

805 lines
21 KiB
C
Raw Normal View History

/*-
* 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.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#if !defined(KLD_MODULE)
#include "opt_inet.h"
#include "opt_ipfw.h"
#include "opt_sctp.h"
#ifndef INET
#error "IPDIVERT requires INET."
#endif
#ifndef IPFIREWALL
#error "IPDIVERT requires IPFIREWALL"
#endif
#endif
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/module.h>
#include <sys/kernel.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/protosw.h>
Conditionally compile out V_ globals while instantiating the appropriate container structures, depending on VIMAGE_GLOBALS compile time option. Make VIMAGE_GLOBALS a new compile-time option, which by default will not be defined, resulting in instatiations of global variables selected for V_irtualization (enclosed in #ifdef VIMAGE_GLOBALS blocks) to be effectively compiled out. Instantiate new global container structures to hold V_irtualized variables: vnet_net_0, vnet_inet_0, vnet_inet6_0, vnet_ipsec_0, vnet_netgraph_0, and vnet_gif_0. Update the VSYM() macro so that depending on VIMAGE_GLOBALS the V_ macros resolve either to the original globals, or to fields inside container structures, i.e. effectively #ifdef VIMAGE_GLOBALS #define V_rt_tables rt_tables #else #define V_rt_tables vnet_net_0._rt_tables #endif Update SYSCTL_V_*() macros to operate either on globals or on fields inside container structs. Extend the internal kldsym() lookups with the ability to resolve selected fields inside the virtualization container structs. This applies only to the fields which are explicitly registered for kldsym() visibility via VNET_MOD_DECLARE() and vnet_mod_register(), currently this is done only in sys/net/if.c. Fix a few broken instances of MODULE_GLOBAL() macro use in SCTP code, and modify the MODULE_GLOBAL() macro to resolve to V_ macros, which in turn result in proper code being generated depending on VIMAGE_GLOBALS. De-virtualize local static variables in sys/contrib/pf/net/pf_subr.c which were prematurely V_irtualized by automated V_ prepending scripts during earlier merging steps. PF virtualization will be done separately, most probably after next PF import. Convert a few variable initializations at instantiation to initialization in init functions, most notably in ipfw. Also convert TUNABLE_INT() initializers for V_ variables to TUNABLE_FETCH_INT() in initializer functions. Discussed at: devsummit Strassburg Reviewed by: bz, julian Approved by: julian (mentor) Obtained from: //depot/projects/vimage-commit2/... X-MFC after: never Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-12-10 23:12:39 +00:00
#include <sys/rwlock.h>
#include <sys/signalvar.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <vm/uma.h>
#include <net/if.h>
Over the past couple of years, there have been a number of reports relating the use of divert sockets to dead locks. A number of LORs have been reported between divert and a number of other network subsystems including: IPSEC, Pfil, multicast, ipfw and others. Other dead locks could occur because of recursive entry into the IP stack. This change should take care of most if not all of these issues. A summary of the changes follow: - We disallow multicast operations on divert sockets. It really doesn't make semantic sense to allow this, since typically you would set multicast parameters on multicast end points. NOTE: As a part of this change, we actually dis-allow multicast options on any socket that IS a divert socket OR IS NOT a SOCK_RAW or SOCK_DGRAM family - We check to see if there are any socket options that have been specified on the socket, and if there was (which is very un-common and also probably doesnt make sense to support) we duplicate the mbuf carrying the options. - We then drop the INP/INFO locks over the call to ip_output(). It should be noted that since we no longer support multicast operations on divert sockets and we have duplicated any socket options, we no longer need the reference to the pcb to be coherent. - Finally, we replaced the call to ip_input() to use netisr queuing. This should remove the recursive entry into the IP stack from divert. By dropping the locks over the call to ip_output() we eliminate all the lock ordering issues above. By switching over to netisr on the inbound path, we can no longer recursively enter the ip_input() code via divert. I have tested this change by using the following command: ipfwpcap -r 8000 - | tcpdump -r - -nn -v This should exercise the input and re-injection (outbound) path, which is very similar to the work load performed by natd(8). Additionally, I have run some ospf daemons which have a heavy reliance on raw sockets and multicast. Approved by: re@ (kensmith) MFC after: 1 month LOR: 163 LOR: 181 LOR: 202 LOR: 203 Discussed with: julian, andre et al (on freebsd-net) In collaboration with: bms [1], rwatson [2] [1] bms helped out with the multicast decisions [2] rwatson submitted the original netisr patches and came up with some of the original ideas on how to combat this issue.
2007-08-06 22:06:36 +00:00
#include <net/netisr.h>
#include <net/route.h>
#include <net/vnet.h>
#include <netinet/in.h>
#include <netinet/in_pcb.h>
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
#include <netinet/ip.h>
#include <netinet/ip_divert.h>
#include <netinet/ip_var.h>
#include <netinet/ip_fw.h>
#ifdef SCTP
#include <netinet/sctp_crc32.h>
#endif
#include <security/mac/mac_framework.h>
/*
* Divert sockets
*/
/*
* Allocate enough space to hold a full IP packet
*/
#define DIVSNDQ (65536 + 100)
#define DIVRCVQ (65536 + 100)
/*
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
* Divert sockets work in conjunction with ipfw, see the divert(4)
* manpage for features.
* Internally, packets selected by ipfw in ip_input() or ip_output(),
* and never diverted before, are passed to the input queue of the
* divert socket with a given 'divert_port' number (as specified in
* the matching ipfw rule), and they are tagged with a 16 bit cookie
* (representing the rule number of the matching ipfw rule), which
* is passed to process reading from the socket.
*
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
* Packets written to the divert socket are again tagged with a cookie
* (usually the same as above) and a destination address.
* If the destination address is INADDR_ANY then the packet is
* treated as outgoing and sent to ip_output(), otherwise it is
* treated as incoming and sent to ip_input().
* In both cases, the packet is tagged with the cookie.
*
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
* On reinjection, processing in ip_input() and ip_output()
* will be exactly the same as for the original packet, except that
* ipfw processing will start at the rule number after the one
* written in the cookie (so, tagging a packet with a cookie of 0
* will cause it to be effectively considered as a standard packet).
*/
/* Internal variables. */
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
static VNET_DEFINE(struct inpcbhead, divcb);
static VNET_DEFINE(struct inpcbinfo, divcbinfo);
#define V_divcb VNET(divcb)
#define V_divcbinfo VNET(divcbinfo)
static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
static eventhandler_tag ip_divert_event_tag;
/*
* Initialize divert connection block queue.
*/
static void
div_zone_change(void *tag)
{
uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets);
}
static int
div_inpcb_init(void *mem, int size, int flags)
{
struct inpcb *inp = mem;
INP_LOCK_INIT(inp, "inp", "divinp");
return (0);
}
static void
div_inpcb_fini(void *mem, int size)
{
struct inpcb *inp = mem;
INP_LOCK_DESTROY(inp);
}
static void
div_init(void)
{
INP_INFO_LOCK_INIT(&V_divcbinfo, "div");
LIST_INIT(&V_divcb);
V_divcbinfo.ipi_listhead = &V_divcb;
Permit buiding kernels with options VIMAGE, restricted to only a single active network stack instance. Turning on options VIMAGE at compile time yields the following changes relative to default kernel build: 1) V_ accessor macros for virtualized variables resolve to structure fields via base pointers, instead of being resolved as fields in global structs or plain global variables. As an example, V_ifnet becomes: options VIMAGE: ((struct vnet_net *) vnet_net)->_ifnet default build: vnet_net_0._ifnet options VIMAGE_GLOBALS: ifnet 2) INIT_VNET_* macros will declare and set up base pointers to be used by V_ accessor macros, instead of resolving to whitespace: INIT_VNET_NET(ifp->if_vnet); becomes struct vnet_net *vnet_net = (ifp->if_vnet)->mod_data[VNET_MOD_NET]; 3) Memory for vnet modules registered via vnet_mod_register() is now allocated at run time in sys/kern/kern_vimage.c, instead of per vnet module structs being declared as globals. If required, vnet modules can now request the framework to provide them with allocated bzeroed memory by filling in the vmi_size field in their vmi_modinfo structures. 4) structs socket, ifnet, inpcbinfo, tcpcb and syncache_head are extended to hold a pointer to the parent vnet. options VIMAGE builds will fill in those fields as required. 5) curvnet is introduced as a new global variable in options VIMAGE builds, always pointing to the default and only struct vnet. 6) struct sysctl_oid has been extended with additional two fields to store major and minor virtualization module identifiers, oid_v_subs and oid_v_mod. SYSCTL_V_* family of macros will fill in those fields accordingly, and store the offset in the appropriate vnet container struct in oid_arg1. In sysctl handlers dealing with virtualized sysctls, the SYSCTL_RESOLVE_V_ARG1() macro will compute the address of the target variable and make it available in arg1 variable for further processing. Unused fields in structs vnet_inet, vnet_inet6 and vnet_ipfw have been deleted. Reviewed by: bz, rwatson Approved by: julian (mentor)
2009-04-30 13:36:26 +00:00
#ifdef VIMAGE
V_divcbinfo.ipi_vnet = curvnet;
#endif
/*
* XXX We don't use the hash list for divert IP, but it's easier
* to allocate a one entry hash list than it is to check all
* over the place for hashbase == NULL.
*/
V_divcbinfo.ipi_hashbase = hashinit(1, M_PCB, &V_divcbinfo.ipi_hashmask);
V_divcbinfo.ipi_porthashbase = hashinit(1, M_PCB,
&V_divcbinfo.ipi_porthashmask);
V_divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb),
NULL, NULL, div_inpcb_init, div_inpcb_fini, UMA_ALIGN_PTR,
UMA_ZONE_NOFREE);
Conditionally compile out V_ globals while instantiating the appropriate container structures, depending on VIMAGE_GLOBALS compile time option. Make VIMAGE_GLOBALS a new compile-time option, which by default will not be defined, resulting in instatiations of global variables selected for V_irtualization (enclosed in #ifdef VIMAGE_GLOBALS blocks) to be effectively compiled out. Instantiate new global container structures to hold V_irtualized variables: vnet_net_0, vnet_inet_0, vnet_inet6_0, vnet_ipsec_0, vnet_netgraph_0, and vnet_gif_0. Update the VSYM() macro so that depending on VIMAGE_GLOBALS the V_ macros resolve either to the original globals, or to fields inside container structures, i.e. effectively #ifdef VIMAGE_GLOBALS #define V_rt_tables rt_tables #else #define V_rt_tables vnet_net_0._rt_tables #endif Update SYSCTL_V_*() macros to operate either on globals or on fields inside container structs. Extend the internal kldsym() lookups with the ability to resolve selected fields inside the virtualization container structs. This applies only to the fields which are explicitly registered for kldsym() visibility via VNET_MOD_DECLARE() and vnet_mod_register(), currently this is done only in sys/net/if.c. Fix a few broken instances of MODULE_GLOBAL() macro use in SCTP code, and modify the MODULE_GLOBAL() macro to resolve to V_ macros, which in turn result in proper code being generated depending on VIMAGE_GLOBALS. De-virtualize local static variables in sys/contrib/pf/net/pf_subr.c which were prematurely V_irtualized by automated V_ prepending scripts during earlier merging steps. PF virtualization will be done separately, most probably after next PF import. Convert a few variable initializations at instantiation to initialization in init functions, most notably in ipfw. Also convert TUNABLE_INT() initializers for V_ variables to TUNABLE_FETCH_INT() in initializer functions. Discussed at: devsummit Strassburg Reviewed by: bz, julian Approved by: julian (mentor) Obtained from: //depot/projects/vimage-commit2/... X-MFC after: never Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-12-10 23:12:39 +00:00
uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets);
}
static void
div_destroy(void)
{
INP_INFO_LOCK_DESTROY(&V_divcbinfo);
uma_zdestroy(V_divcbinfo.ipi_zone);
hashdestroy(V_divcbinfo.ipi_hashbase, M_PCB, V_divcbinfo.ipi_hashmask);
hashdestroy(V_divcbinfo.ipi_porthashbase, M_PCB,
V_divcbinfo.ipi_porthashmask);
}
/*
* IPPROTO_DIVERT is not in the real IP protocol number space; this
* function should never be called. Just in case, drop any packets.
*/
void
div_input(struct mbuf *m, int off)
{
KMOD_IPSTAT_INC(ips_noproto);
m_freem(m);
}
/*
* Divert a packet by passing it up to the divert socket at port 'port'.
*
* Setup generic address and protocol structures for div_input routine,
* then pass them along with mbuf chain.
*/
static void
divert_packet(struct mbuf *m, int incoming)
{
struct ip *ip;
struct inpcb *inp;
struct socket *sa;
u_int16_t nport;
struct sockaddr_in divsrc;
struct m_tag *mtag;
mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
if (mtag == NULL) {
printf("%s: no divert tag\n", __func__);
m_freem(m);
return;
}
/* Assure header */
if (m->m_len < sizeof(struct ip) &&
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
(m = m_pullup(m, sizeof(struct ip))) == 0)
return;
ip = mtod(m, struct ip *);
/* Delayed checksums are currently not compatible with divert. */
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
ip->ip_len = ntohs(ip->ip_len);
in_delayed_cksum(m);
m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
ip->ip_len = htons(ip->ip_len);
}
#ifdef SCTP
if (m->m_pkthdr.csum_flags & CSUM_SCTP) {
ip->ip_len = ntohs(ip->ip_len);
sctp_delayed_cksum(m);
m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
ip->ip_len = htons(ip->ip_len);
}
#endif
/*
* Record receive interface address, if any.
* But only for incoming packets.
*/
bzero(&divsrc, sizeof(divsrc));
divsrc.sin_len = sizeof(divsrc);
divsrc.sin_family = AF_INET;
divsrc.sin_port = divert_cookie(mtag); /* record matching rule */
if (incoming) {
struct ifaddr *ifa;
struct ifnet *ifp;
/* Sanity check */
M_ASSERTPKTHDR(m);
/* Find IP address for receive interface */
ifp = m->m_pkthdr.rcvif;
if_addr_rlock(ifp);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
if (ifa->ifa_addr->sa_family != AF_INET)
continue;
divsrc.sin_addr =
((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
break;
}
if_addr_runlock(ifp);
}
/*
* Record the incoming interface name whenever we have one.
*/
if (m->m_pkthdr.rcvif) {
/*
* Hide the actual interface name in there in the
* sin_zero array. XXX This needs to be moved to a
* different sockaddr type for divert, e.g.
* sockaddr_div with multiple fields like
* sockaddr_dl. Presently we have only 7 bytes
* but that will do for now as most interfaces
* are 4 or less + 2 or less bytes for unit.
* There is probably a faster way of doing this,
* possibly taking it from the sockaddr_dl on the iface.
* This solves the problem of a P2P link and a LAN interface
* having the same address, which can result in the wrong
* interface being assigned to the packet when fed back
* into the divert socket. Theoretically if the daemon saves
* and re-uses the sockaddr_in as suggested in the man pages,
* this iface name will come along for the ride.
* (see div_output for the other half of this.)
*/
strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname,
sizeof(divsrc.sin_zero));
}
/* Put packet on socket queue, if any */
sa = NULL;
nport = htons((u_int16_t)divert_info(mtag));
INP_INFO_RLOCK(&V_divcbinfo);
LIST_FOREACH(inp, &V_divcb, inp_list) {
/* XXX why does only one socket match? */
if (inp->inp_lport == nport) {
INP_RLOCK(inp);
sa = inp->inp_socket;
SOCKBUF_LOCK(&sa->so_rcv);
if (sbappendaddr_locked(&sa->so_rcv,
(struct sockaddr *)&divsrc, m,
(struct mbuf *)0) == 0) {
SOCKBUF_UNLOCK(&sa->so_rcv);
sa = NULL; /* force mbuf reclaim below */
} else
sorwakeup_locked(sa);
INP_RUNLOCK(inp);
break;
}
}
INP_INFO_RUNLOCK(&V_divcbinfo);
if (sa == NULL) {
m_freem(m);
KMOD_IPSTAT_INC(ips_noproto);
KMOD_IPSTAT_DEC(ips_delivered);
}
}
/*
* Deliver packet back into the IP processing machinery.
*
* If no address specified, or address is 0.0.0.0, send to ip_output();
* otherwise, send to ip_input() and mark as having been received on
* the interface with that address.
*/
static int
div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin,
struct mbuf *control)
{
struct m_tag *mtag;
struct divert_tag *dt;
int error = 0;
Over the past couple of years, there have been a number of reports relating the use of divert sockets to dead locks. A number of LORs have been reported between divert and a number of other network subsystems including: IPSEC, Pfil, multicast, ipfw and others. Other dead locks could occur because of recursive entry into the IP stack. This change should take care of most if not all of these issues. A summary of the changes follow: - We disallow multicast operations on divert sockets. It really doesn't make semantic sense to allow this, since typically you would set multicast parameters on multicast end points. NOTE: As a part of this change, we actually dis-allow multicast options on any socket that IS a divert socket OR IS NOT a SOCK_RAW or SOCK_DGRAM family - We check to see if there are any socket options that have been specified on the socket, and if there was (which is very un-common and also probably doesnt make sense to support) we duplicate the mbuf carrying the options. - We then drop the INP/INFO locks over the call to ip_output(). It should be noted that since we no longer support multicast operations on divert sockets and we have duplicated any socket options, we no longer need the reference to the pcb to be coherent. - Finally, we replaced the call to ip_input() to use netisr queuing. This should remove the recursive entry into the IP stack from divert. By dropping the locks over the call to ip_output() we eliminate all the lock ordering issues above. By switching over to netisr on the inbound path, we can no longer recursively enter the ip_input() code via divert. I have tested this change by using the following command: ipfwpcap -r 8000 - | tcpdump -r - -nn -v This should exercise the input and re-injection (outbound) path, which is very similar to the work load performed by natd(8). Additionally, I have run some ospf daemons which have a heavy reliance on raw sockets and multicast. Approved by: re@ (kensmith) MFC after: 1 month LOR: 163 LOR: 181 LOR: 202 LOR: 203 Discussed with: julian, andre et al (on freebsd-net) In collaboration with: bms [1], rwatson [2] [1] bms helped out with the multicast decisions [2] rwatson submitted the original netisr patches and came up with some of the original ideas on how to combat this issue.
2007-08-06 22:06:36 +00:00
struct mbuf *options;
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
/*
* An mbuf may hasn't come from userland, but we pretend
* that it has.
*/
m->m_pkthdr.rcvif = NULL;
m->m_nextpkt = NULL;
M_SETFIB(m, so->so_fibnum);
if (control)
m_freem(control); /* XXX */
if ((mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL)) == NULL) {
mtag = m_tag_get(PACKET_TAG_DIVERT, sizeof(struct divert_tag),
M_NOWAIT | M_ZERO);
if (mtag == NULL) {
error = ENOBUFS;
goto cantsend;
}
dt = (struct divert_tag *)(mtag+1);
m_tag_prepend(m, mtag);
} else
dt = (struct divert_tag *)(mtag+1);
/* Loopback avoidance and state recovery */
if (sin) {
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
int i;
dt->cookie = sin->sin_port;
/*
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
* Find receive interface with the given name, stuffed
* (if it exists) in the sin_zero[] field.
* The name is user supplied data so don't trust its size
* or that it is zero terminated.
*/
for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++)
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
;
if ( i > 0 && i < sizeof(sin->sin_zero))
m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
}
/* Reinject packet into the system as incoming or outgoing */
if (!sin || sin->sin_addr.s_addr == 0) {
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
struct ip *const ip = mtod(m, struct ip *);
struct inpcb *inp;
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
dt->info |= IP_FW_DIVERT_OUTPUT_FLAG;
INP_INFO_WLOCK(&V_divcbinfo);
inp = sotoinpcb(so);
INP_RLOCK(inp);
/*
* Don't allow both user specified and setsockopt options,
* and don't allow packet length sizes that will crash
*/
if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
error = EINVAL;
INP_RUNLOCK(inp);
INP_INFO_WUNLOCK(&V_divcbinfo);
m_freem(m);
} else {
/* Convert fields to host order for ip_output() */
ip->ip_len = ntohs(ip->ip_len);
ip->ip_off = ntohs(ip->ip_off);
/* Send packet to output processing */
KMOD_IPSTAT_INC(ips_rawout); /* XXX */
#ifdef MAC
mac_inpcb_create_mbuf(inp, m);
#endif
Over the past couple of years, there have been a number of reports relating the use of divert sockets to dead locks. A number of LORs have been reported between divert and a number of other network subsystems including: IPSEC, Pfil, multicast, ipfw and others. Other dead locks could occur because of recursive entry into the IP stack. This change should take care of most if not all of these issues. A summary of the changes follow: - We disallow multicast operations on divert sockets. It really doesn't make semantic sense to allow this, since typically you would set multicast parameters on multicast end points. NOTE: As a part of this change, we actually dis-allow multicast options on any socket that IS a divert socket OR IS NOT a SOCK_RAW or SOCK_DGRAM family - We check to see if there are any socket options that have been specified on the socket, and if there was (which is very un-common and also probably doesnt make sense to support) we duplicate the mbuf carrying the options. - We then drop the INP/INFO locks over the call to ip_output(). It should be noted that since we no longer support multicast operations on divert sockets and we have duplicated any socket options, we no longer need the reference to the pcb to be coherent. - Finally, we replaced the call to ip_input() to use netisr queuing. This should remove the recursive entry into the IP stack from divert. By dropping the locks over the call to ip_output() we eliminate all the lock ordering issues above. By switching over to netisr on the inbound path, we can no longer recursively enter the ip_input() code via divert. I have tested this change by using the following command: ipfwpcap -r 8000 - | tcpdump -r - -nn -v This should exercise the input and re-injection (outbound) path, which is very similar to the work load performed by natd(8). Additionally, I have run some ospf daemons which have a heavy reliance on raw sockets and multicast. Approved by: re@ (kensmith) MFC after: 1 month LOR: 163 LOR: 181 LOR: 202 LOR: 203 Discussed with: julian, andre et al (on freebsd-net) In collaboration with: bms [1], rwatson [2] [1] bms helped out with the multicast decisions [2] rwatson submitted the original netisr patches and came up with some of the original ideas on how to combat this issue.
2007-08-06 22:06:36 +00:00
/*
* Get ready to inject the packet into ip_output().
* Just in case socket options were specified on the
* divert socket, we duplicate them. This is done
* to avoid having to hold the PCB locks over the call
* to ip_output(), as doing this results in a number of
* lock ordering complexities.
*
* Note that we set the multicast options argument for
* ip_output() to NULL since it should be invariant that
* they are not present.
*/
KASSERT(inp->inp_moptions == NULL,
("multicast options set on a divert socket"));
options = NULL;
/*
* XXXCSJP: It is unclear to me whether or not it makes
* sense for divert sockets to have options. However,
* for now we will duplicate them with the INP locks
* held so we can use them in ip_output() without
* requring a reference to the pcb.
*/
if (inp->inp_options != NULL) {
options = m_dup(inp->inp_options, M_DONTWAIT);
if (options == NULL)
error = ENOBUFS;
}
INP_RUNLOCK(inp);
INP_INFO_WUNLOCK(&V_divcbinfo);
Over the past couple of years, there have been a number of reports relating the use of divert sockets to dead locks. A number of LORs have been reported between divert and a number of other network subsystems including: IPSEC, Pfil, multicast, ipfw and others. Other dead locks could occur because of recursive entry into the IP stack. This change should take care of most if not all of these issues. A summary of the changes follow: - We disallow multicast operations on divert sockets. It really doesn't make semantic sense to allow this, since typically you would set multicast parameters on multicast end points. NOTE: As a part of this change, we actually dis-allow multicast options on any socket that IS a divert socket OR IS NOT a SOCK_RAW or SOCK_DGRAM family - We check to see if there are any socket options that have been specified on the socket, and if there was (which is very un-common and also probably doesnt make sense to support) we duplicate the mbuf carrying the options. - We then drop the INP/INFO locks over the call to ip_output(). It should be noted that since we no longer support multicast operations on divert sockets and we have duplicated any socket options, we no longer need the reference to the pcb to be coherent. - Finally, we replaced the call to ip_input() to use netisr queuing. This should remove the recursive entry into the IP stack from divert. By dropping the locks over the call to ip_output() we eliminate all the lock ordering issues above. By switching over to netisr on the inbound path, we can no longer recursively enter the ip_input() code via divert. I have tested this change by using the following command: ipfwpcap -r 8000 - | tcpdump -r - -nn -v This should exercise the input and re-injection (outbound) path, which is very similar to the work load performed by natd(8). Additionally, I have run some ospf daemons which have a heavy reliance on raw sockets and multicast. Approved by: re@ (kensmith) MFC after: 1 month LOR: 163 LOR: 181 LOR: 202 LOR: 203 Discussed with: julian, andre et al (on freebsd-net) In collaboration with: bms [1], rwatson [2] [1] bms helped out with the multicast decisions [2] rwatson submitted the original netisr patches and came up with some of the original ideas on how to combat this issue.
2007-08-06 22:06:36 +00:00
if (error == ENOBUFS) {
m_freem(m);
return (error);
}
error = ip_output(m, options, NULL,
((so->so_options & SO_DONTROUTE) ?
IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST |
IP_RAWOUTPUT, NULL, NULL);
if (options != NULL)
m_freem(options);
}
} else {
dt->info |= IP_FW_DIVERT_LOOPBACK_FLAG;
if (m->m_pkthdr.rcvif == NULL) {
/*
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
* No luck with the name, check by IP address.
* Clear the port and the ifname to make sure
* there are no distractions for ifa_ifwithaddr.
*/
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
struct ifaddr *ifa;
bzero(sin->sin_zero, sizeof(sin->sin_zero));
sin->sin_port = 0;
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
ifa = ifa_ifwithaddr((struct sockaddr *) sin);
if (ifa == NULL) {
error = EADDRNOTAVAIL;
goto cantsend;
}
m->m_pkthdr.rcvif = ifa->ifa_ifp;
ifa_free(ifa);
}
#ifdef MAC
mac_socket_create_mbuf(so, m);
#endif
Over the past couple of years, there have been a number of reports relating the use of divert sockets to dead locks. A number of LORs have been reported between divert and a number of other network subsystems including: IPSEC, Pfil, multicast, ipfw and others. Other dead locks could occur because of recursive entry into the IP stack. This change should take care of most if not all of these issues. A summary of the changes follow: - We disallow multicast operations on divert sockets. It really doesn't make semantic sense to allow this, since typically you would set multicast parameters on multicast end points. NOTE: As a part of this change, we actually dis-allow multicast options on any socket that IS a divert socket OR IS NOT a SOCK_RAW or SOCK_DGRAM family - We check to see if there are any socket options that have been specified on the socket, and if there was (which is very un-common and also probably doesnt make sense to support) we duplicate the mbuf carrying the options. - We then drop the INP/INFO locks over the call to ip_output(). It should be noted that since we no longer support multicast operations on divert sockets and we have duplicated any socket options, we no longer need the reference to the pcb to be coherent. - Finally, we replaced the call to ip_input() to use netisr queuing. This should remove the recursive entry into the IP stack from divert. By dropping the locks over the call to ip_output() we eliminate all the lock ordering issues above. By switching over to netisr on the inbound path, we can no longer recursively enter the ip_input() code via divert. I have tested this change by using the following command: ipfwpcap -r 8000 - | tcpdump -r - -nn -v This should exercise the input and re-injection (outbound) path, which is very similar to the work load performed by natd(8). Additionally, I have run some ospf daemons which have a heavy reliance on raw sockets and multicast. Approved by: re@ (kensmith) MFC after: 1 month LOR: 163 LOR: 181 LOR: 202 LOR: 203 Discussed with: julian, andre et al (on freebsd-net) In collaboration with: bms [1], rwatson [2] [1] bms helped out with the multicast decisions [2] rwatson submitted the original netisr patches and came up with some of the original ideas on how to combat this issue.
2007-08-06 22:06:36 +00:00
/* Send packet to input processing via netisr */
Reimplement the netisr framework in order to support parallel netisr threads: - Support up to one netisr thread per CPU, each processings its own workstream, or set of per-protocol queues. Threads may be bound to specific CPUs, or allowed to migrate, based on a global policy. In the future it would be desirable to support topology-centric policies, such as "one netisr per package". - Allow each protocol to advertise an ordering policy, which can currently be one of: NETISR_POLICY_SOURCE: packets must maintain ordering with respect to an implicit or explicit source (such as an interface or socket). NETISR_POLICY_FLOW: make use of mbuf flow identifiers to place work, as well as allowing protocols to provide a flow generation function for mbufs without flow identifers (m2flow). Falls back on NETISR_POLICY_SOURCE if now flow ID is available. NETISR_POLICY_CPU: allow protocols to inspect and assign a CPU for each packet handled by netisr (m2cpuid). - Provide utility functions for querying the number of workstreams being used, as well as a mapping function from workstream to CPU ID, which protocols may use in work placement decisions. - Add explicit interfaces to get and set per-protocol queue limits, and get and clear drop counters, which query data or apply changes across all workstreams. - Add a more extensible netisr registration interface, in which protocols declare 'struct netisr_handler' structures for each registered NETISR_ type. These include name, handler function, optional mbuf to flow ID function, optional mbuf to CPU ID function, queue limit, and ordering policy. Padding is present to allow these to be expanded in the future. If no queue limit is declared, then a default is used. - Queue limits are now per-workstream, and raised from the previous IFQ_MAXLEN default of 50 to 256. - All protocols are updated to use the new registration interface, and with the exception of netnatm, default queue limits. Most protocols register as NETISR_POLICY_SOURCE, except IPv4 and IPv6, which use NETISR_POLICY_FLOW, and will therefore take advantage of driver- generated flow IDs if present. - Formalize a non-packet based interface between interface polling and the netisr, rather than having polling pretend to be two protocols. Provide two explicit hooks in the netisr worker for start and end events for runs: netisr_poll() and netisr_pollmore(), as well as a function, netisr_sched_poll(), to allow the polling code to schedule netisr execution. DEVICE_POLLING still embeds single-netisr assumptions in its implementation, so for now if it is compiled into the kernel, a single and un-bound netisr thread is enforced regardless of tunable configuration. In the default configuration, the new netisr implementation maintains the same basic assumptions as the previous implementation: a single, un-bound worker thread processes all deferred work, and direct dispatch is enabled by default wherever possible. Performance measurement shows a marginal performance improvement over the old implementation due to the use of batched dequeue. An rmlock is used to synchronize use and registration/unregistration using the framework; currently, synchronized use is disabled (replicating current netisr policy) due to a measurable 3%-6% hit in ping-pong micro-benchmarking. It will be enabled once further rmlock optimization has taken place. However, in practice, netisrs are rarely registered or unregistered at runtime. A new man page for netisr will follow, but since one doesn't currently exist, it hasn't been updated. This change is not appropriate for MFC, although the polling shutdown handler should be merged to 7-STABLE. Bump __FreeBSD_version. Reviewed by: bz
2009-06-01 10:41:38 +00:00
netisr_queue_src(NETISR_IP, (uintptr_t)so, m);
}
return error;
cantsend:
m_freem(m);
return error;
}
static int
div_attach(struct socket *so, int proto, struct thread *td)
{
struct inpcb *inp;
int error;
inp = sotoinpcb(so);
KASSERT(inp == NULL, ("div_attach: inp != NULL"));
if (td != NULL) {
error = priv_check(td, PRIV_NETINET_DIVERT);
if (error)
return (error);
}
error = soreserve(so, div_sendspace, div_recvspace);
if (error)
return error;
INP_INFO_WLOCK(&V_divcbinfo);
error = in_pcballoc(so, &V_divcbinfo);
if (error) {
INP_INFO_WUNLOCK(&V_divcbinfo);
return error;
}
inp = (struct inpcb *)so->so_pcb;
INP_INFO_WUNLOCK(&V_divcbinfo);
inp->inp_ip_p = proto;
inp->inp_vflag |= INP_IPV4;
inp->inp_flags |= INP_HDRINCL;
INP_WUNLOCK(inp);
return 0;
}
static void
div_detach(struct socket *so)
{
struct inpcb *inp;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("div_detach: inp == NULL"));
INP_INFO_WLOCK(&V_divcbinfo);
INP_WLOCK(inp);
in_pcbdetach(inp);
in_pcbfree(inp);
INP_INFO_WUNLOCK(&V_divcbinfo);
}
static int
div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
{
struct inpcb *inp;
int error;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("div_bind: inp == NULL"));
2002-06-23 09:13:46 +00:00
/* in_pcbbind assumes that nam is a sockaddr_in
* and in_pcbbind requires a valid address. Since divert
* sockets don't we need to make sure the address is
* filled in properly.
* XXX -- divert should not be abusing in_pcbind
* and should probably have its own family.
*/
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
if (nam->sa_family != AF_INET)
return EAFNOSUPPORT;
((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
INP_INFO_WLOCK(&V_divcbinfo);
INP_WLOCK(inp);
error = in_pcbbind(inp, nam, td->td_ucred);
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_divcbinfo);
return error;
}
static int
div_shutdown(struct socket *so)
{
struct inpcb *inp;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("div_shutdown: inp == NULL"));
INP_WLOCK(inp);
socantsendmore(so);
INP_WUNLOCK(inp);
return 0;
}
static int
div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
struct mbuf *control, struct thread *td)
{
/* Packet must have a header (but that's about it) */
if (m->m_len < sizeof (struct ip) &&
(m = m_pullup(m, sizeof (struct ip))) == 0) {
KMOD_IPSTAT_INC(ips_toosmall);
m_freem(m);
return EINVAL;
}
/* Send packet */
Remove (almost all) global variables that were used to hold packet forwarding state ("annotations") during ip processing. The code is considerably cleaner now. The variables removed by this change are: ip_divert_cookie used by divert sockets ip_fw_fwd_addr used for transparent ip redirection last_pkt used by dynamic pipes in dummynet Removal of the first two has been done by carrying the annotations into volatile structs prepended to the mbuf chains, and adding appropriate code to add/remove annotations in the routines which make use of them, i.e. ip_input(), ip_output(), tcp_input(), bdg_forward(), ether_demux(), ether_output_frame(), div_output(). On passing, remove a bug in divert handling of fragmented packet. Now it is the fragment at offset 0 which sets the divert status of the whole packet, whereas formerly it was the last incoming fragment to decide. Removal of last_pkt required a change in the interface of ip_fw_chk() and dummynet_io(). On passing, use the same mechanism for dummynet annotations and for divert/forward annotations. option IPFIREWALL_FORWARD is effectively useless, the code to implement it is very small and is now in by default to avoid the obfuscation of conditionally compiled code. NOTES: * there is at least one global variable left, sro_fwd, in ip_output(). I am not sure if/how this can be removed. * I have deliberately avoided gratuitous style changes in this commit to avoid cluttering the diffs. Minor stule cleanup will likely be necessary * this commit only focused on the IP layer. I am sure there is a number of global variables used in the TCP and maybe UDP stack. * despite the number of files touched, there are absolutely no API's or data structures changed by this commit (except the interfaces of ip_fw_chk() and dummynet_io(), which are internal anyways), so an MFC is quite safe and unintrusive (and desirable, given the improved readability of the code). MFC after: 10 days
2002-06-22 11:51:02 +00:00
return div_output(so, m, (struct sockaddr_in *)nam, control);
}
void
div_ctlinput(int cmd, struct sockaddr *sa, void *vip)
{
struct in_addr faddr;
faddr = ((struct sockaddr_in *)sa)->sin_addr;
if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
return;
if (PRC_IS_REDIRECT(cmd))
return;
}
static int
div_pcblist(SYSCTL_HANDLER_ARGS)
{
int error, i, n;
struct inpcb *inp, **inp_list;
inp_gen_t gencnt;
struct xinpgen xig;
/*
* The process of preparing the TCB list is too time-consuming and
* resource-intensive to repeat twice on every request.
*/
if (req->oldptr == 0) {
n = V_divcbinfo.ipi_count;
req->oldidx = 2 * (sizeof xig)
+ (n + n/8) * sizeof(struct xinpcb);
return 0;
}
if (req->newptr != 0)
return EPERM;
/*
* OK, now we're committed to doing something.
*/
INP_INFO_RLOCK(&V_divcbinfo);
gencnt = V_divcbinfo.ipi_gencnt;
n = V_divcbinfo.ipi_count;
INP_INFO_RUNLOCK(&V_divcbinfo);
error = sysctl_wire_old_buffer(req,
2 * sizeof(xig) + n*sizeof(struct xinpcb));
if (error != 0)
return (error);
xig.xig_len = sizeof xig;
xig.xig_count = n;
xig.xig_gen = gencnt;
xig.xig_sogen = so_gencnt;
error = SYSCTL_OUT(req, &xig, sizeof xig);
if (error)
return error;
inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0)
return ENOMEM;
INP_INFO_RLOCK(&V_divcbinfo);
for (inp = LIST_FIRST(V_divcbinfo.ipi_listhead), i = 0; inp && i < n;
inp = LIST_NEXT(inp, inp_list)) {
INP_RLOCK(inp);
if (inp->inp_gencnt <= gencnt &&
cr_canseeinpcb(req->td->td_ucred, inp) == 0)
inp_list[i++] = inp;
INP_RUNLOCK(inp);
}
INP_INFO_RUNLOCK(&V_divcbinfo);
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
INP_RLOCK(inp);
if (inp->inp_gencnt <= gencnt) {
struct xinpcb xi;
bzero(&xi, sizeof(xi));
xi.xi_len = sizeof xi;
/* XXX should avoid extra copy */
bcopy(inp, &xi.xi_inp, sizeof *inp);
if (inp->inp_socket)
sotoxsocket(inp->inp_socket, &xi.xi_socket);
INP_RUNLOCK(inp);
error = SYSCTL_OUT(req, &xi, sizeof xi);
} else
INP_RUNLOCK(inp);
}
if (!error) {
/*
* Give the user an updated idea of our state.
* If the generation differs from what we told
* her before, she knows that something happened
* while we were processing this request, and it
* might be necessary to retry.
*/
INP_INFO_RLOCK(&V_divcbinfo);
xig.xig_gen = V_divcbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = V_divcbinfo.ipi_count;
INP_INFO_RUNLOCK(&V_divcbinfo);
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
free(inp_list, M_TEMP);
return error;
}
#ifdef SYSCTL_NODE
SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert, CTLFLAG_RW, 0, "IPDIVERT");
SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
div_pcblist, "S,xinpcb", "List of active divert sockets");
#endif
struct pr_usrreqs div_usrreqs = {
.pru_attach = div_attach,
.pru_bind = div_bind,
.pru_control = in_control,
.pru_detach = div_detach,
.pru_peeraddr = in_getpeeraddr,
.pru_send = div_send,
.pru_shutdown = div_shutdown,
.pru_sockaddr = in_getsockaddr,
2004-11-11 19:02:01 +00:00
.pru_sosetlabel = in_pcbsosetlabel
};
struct protosw div_protosw = {
.pr_type = SOCK_RAW,
.pr_protocol = IPPROTO_DIVERT,
.pr_flags = PR_ATOMIC|PR_ADDR,
.pr_input = div_input,
.pr_ctlinput = div_ctlinput,
.pr_ctloutput = ip_ctloutput,
.pr_init = div_init,
#ifdef VIMAGE
.pr_destroy = div_destroy,
#endif
.pr_usrreqs = &div_usrreqs
};
static int
div_modevent(module_t mod, int type, void *unused)
{
int err = 0;
#ifndef VIMAGE
int n;
#endif
switch (type) {
case MOD_LOAD:
/*
* Protocol will be initialized by pf_proto_register().
* We don't have to register ip_protox because we are not
* a true IP protocol that goes over the wire.
*/
err = pf_proto_register(PF_INET, &div_protosw);
if (err != 0)
return (err);
ip_divert_ptr = divert_packet;
ip_divert_event_tag = EVENTHANDLER_REGISTER(maxsockets_change,
div_zone_change, NULL, EVENTHANDLER_PRI_ANY);
break;
case MOD_QUIESCE:
/*
* IPDIVERT may normally not be unloaded because of the
* potential race conditions. Tell kldunload we can't be
* unloaded unless the unload is forced.
*/
err = EPERM;
break;
case MOD_UNLOAD:
#ifdef VIMAGE
err = EPERM;
break;
#else
/*
* Forced unload.
*
* Module ipdivert can only be unloaded if no sockets are
* connected. Maybe this can be changed later to forcefully
* disconnect any open sockets.
*
* XXXRW: Note that there is a slight race here, as a new
* socket open request could be spinning on the lock and then
* we destroy the lock.
*/
INP_INFO_WLOCK(&V_divcbinfo);
n = V_divcbinfo.ipi_count;
if (n != 0) {
err = EBUSY;
INP_INFO_WUNLOCK(&V_divcbinfo);
break;
}
ip_divert_ptr = NULL;
err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW);
INP_INFO_WUNLOCK(&V_divcbinfo);
div_destroy();
EVENTHANDLER_DEREGISTER(maxsockets_change, ip_divert_event_tag);
break;
#endif /* !VIMAGE */
default:
err = EOPNOTSUPP;
break;
}
return err;
}
static moduledata_t ipdivertmod = {
"ipdivert",
div_modevent,
0
};
DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY);
MODULE_DEPEND(dummynet, ipfw, 2, 2, 2);
MODULE_VERSION(ipdivert, 1);