2005-01-07 01:45:51 +00:00
|
|
|
/*-
|
1994-05-24 10:09:53 +00:00
|
|
|
* Copyright (c) 1980, 1986, 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.
|
|
|
|
*
|
2001-10-17 10:41:00 +00:00
|
|
|
* @(#)if.c 8.5 (Berkeley) 1/9/95
|
1999-08-28 01:08:13 +00:00
|
|
|
* $FreeBSD$
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
|
1997-12-16 17:40:42 +00:00
|
|
|
#include "opt_compat.h"
|
1999-12-07 17:39:16 +00:00
|
|
|
#include "opt_inet6.h"
|
1999-12-30 18:29:55 +00:00
|
|
|
#include "opt_inet.h"
|
1997-12-16 17:40:42 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/param.h>
|
2004-09-22 08:59:41 +00:00
|
|
|
#include <sys/types.h>
|
2001-09-29 05:55:04 +00:00
|
|
|
#include <sys/conf.h>
|
1997-09-02 01:19:47 +00:00
|
|
|
#include <sys/malloc.h>
|
2004-09-22 08:59:41 +00:00
|
|
|
#include <sys/sbuf.h>
|
2001-10-11 18:39:05 +00:00
|
|
|
#include <sys/bus.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/mbuf.h>
|
|
|
|
#include <sys/systm.h>
|
2006-11-06 13:42:10 +00:00
|
|
|
#include <sys/priv.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <sys/proc.h>
|
|
|
|
#include <sys/socket.h>
|
|
|
|
#include <sys/socketvar.h>
|
|
|
|
#include <sys/protosw.h>
|
|
|
|
#include <sys/kernel.h>
|
2008-12-09 20:05:58 +00:00
|
|
|
#include <sys/lock.h>
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
#include <sys/refcount.h>
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
#include <sys/module.h>
|
2008-12-09 20:05:58 +00:00
|
|
|
#include <sys/rwlock.h>
|
1997-03-24 11:33:46 +00:00
|
|
|
#include <sys/sockio.h>
|
1995-09-22 17:57:48 +00:00
|
|
|
#include <sys/syslog.h>
|
1995-12-20 21:53:53 +00:00
|
|
|
#include <sys/sysctl.h>
|
2004-07-27 23:20:45 +00:00
|
|
|
#include <sys/taskqueue.h>
|
2003-10-17 15:46:31 +00:00
|
|
|
#include <sys/domain.h>
|
2001-02-21 06:39:57 +00:00
|
|
|
#include <sys/jail.h>
|
2002-09-24 17:35:08 +00:00
|
|
|
#include <machine/stdarg.h>
|
This main goals of this project are:
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,
The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.
Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:
- Kip Macy revised the locking code completely, thus completing
the last piece of the puzzle, Kip has also been conducting
active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
me maintaining that branch before the svn conversion
2008-12-15 06:10:57 +00:00
|
|
|
#include <vm/uma.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
#include <net/if.h>
|
2008-11-06 15:26:09 +00:00
|
|
|
#include <net/if_arp.h>
|
Major overhaul of pseudo-interface cloning. Highlights include:
- Split the code out into if_clone.[ch].
- Locked struct if_clone. [1]
- Add a per-cloner match function rather then simply matching names of
the form <name><unit> and <name>.
- Use the match function to allow creation of <interface>.<tag>
vlan interfaces. The old way is preserved unchanged!
- Also the match function to allow creation of stf(4) interfaces named
stf0, stf, or 6to4. This is the only major user visible change in
that "ifconfig stf" creates the interface stf rather then stf0 and
does not print "stf0" to stdout.
- Allow destroy functions to fail so they can refuse to delete
interfaces. Currently, we forbid the deletion of interfaces which
were created in the init function, particularly lo0, pflog0, and
pfsync0. In the case of lo0 this was a panic implementation so it
does not count as a user visiable change. :-)
- Since most interfaces do not need the new functionality, an family of
wrapper functions, ifc_simple_*(), were created to wrap old style
cloner functions.
- The IF_CLONE_INITIALIZER macro is replaced with a new incompatible
IFC_CLONE_INITIALIZER and ifc_simple consumers use IFC_SIMPLE_DECLARE
instead.
Submitted by: Maurycy Pawlowski-Wieronski <maurycy at fouk.org> [1]
Reviewed by: andre, mlaier
Discussed on: net
2004-06-22 20:13:25 +00:00
|
|
|
#include <net/if_clone.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
#include <net/if_dl.h>
|
2000-08-15 00:48:38 +00:00
|
|
|
#include <net/if_types.h>
|
2001-07-02 20:49:25 +00:00
|
|
|
#include <net/if_var.h>
|
1994-10-08 01:40:23 +00:00
|
|
|
#include <net/radix.h>
|
1999-12-17 06:46:07 +00:00
|
|
|
#include <net/route.h>
|
2008-12-02 21:37:28 +00:00
|
|
|
#include <net/vnet.h>
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1999-12-30 18:29:55 +00:00
|
|
|
#if defined(INET) || defined(INET6)
|
1999-11-22 02:45:11 +00:00
|
|
|
/*XXX*/
|
|
|
|
#include <netinet/in.h>
|
1999-12-30 18:29:55 +00:00
|
|
|
#include <netinet/in_var.h>
|
2010-08-11 20:18:19 +00:00
|
|
|
#include <netinet/ip_carp.h>
|
2000-02-01 15:49:37 +00:00
|
|
|
#ifdef INET6
|
2000-07-16 01:46:42 +00:00
|
|
|
#include <netinet6/in6_var.h>
|
|
|
|
#include <netinet6/in6_ifattach.h>
|
2000-02-01 15:49:37 +00:00
|
|
|
#endif
|
1999-11-22 02:45:11 +00:00
|
|
|
#endif
|
2002-02-26 01:11:08 +00:00
|
|
|
#ifdef INET
|
|
|
|
#include <netinet/if_ether.h>
|
|
|
|
#endif
|
1999-11-22 02:45:11 +00:00
|
|
|
|
2006-10-22 11:52:19 +00:00
|
|
|
#include <security/mac/mac_framework.h>
|
|
|
|
|
2010-10-21 16:20:48 +00:00
|
|
|
#ifdef COMPAT_FREEBSD32
|
|
|
|
#include <sys/mount.h>
|
|
|
|
#include <compat/freebsd32/freebsd32.h>
|
|
|
|
#endif
|
|
|
|
|
2009-06-29 19:46:29 +00:00
|
|
|
struct ifindex_entry {
|
|
|
|
struct ifnet *ife_ifnet;
|
|
|
|
};
|
|
|
|
|
2005-03-12 12:58:03 +00:00
|
|
|
SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
|
|
|
|
SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
|
|
|
|
|
2010-05-03 07:32:50 +00:00
|
|
|
TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
|
2011-01-12 19:53:50 +00:00
|
|
|
SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
|
2010-05-03 07:32:50 +00:00
|
|
|
&ifqmaxlen, 0, "max send queue size");
|
|
|
|
|
2005-03-12 12:58:03 +00:00
|
|
|
/* Log link state change events */
|
|
|
|
static int log_link_state_change = 1;
|
|
|
|
|
|
|
|
SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
|
|
|
|
&log_link_state_change, 0,
|
|
|
|
"log interface link state change events");
|
|
|
|
|
2010-01-27 00:30:07 +00:00
|
|
|
/* Interface description */
|
|
|
|
static unsigned int ifdescr_maxlen = 1024;
|
|
|
|
SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
|
|
|
|
&ifdescr_maxlen, 0,
|
|
|
|
"administrative maximum length for interface description");
|
|
|
|
|
|
|
|
MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
|
|
|
|
|
|
|
|
/* global sx for non-critical path ifdescr */
|
|
|
|
static struct sx ifdescr_sx;
|
|
|
|
SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
|
|
|
|
|
2005-06-05 03:13:13 +00:00
|
|
|
void (*bstp_linkstate_p)(struct ifnet *ifp, int state);
|
2005-01-08 12:42:03 +00:00
|
|
|
void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
|
2007-04-17 00:35:11 +00:00
|
|
|
void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
|
2010-08-11 20:18:19 +00:00
|
|
|
/* These are external hooks for CARP. */
|
2010-08-11 00:51:50 +00:00
|
|
|
void (*carp_linkstate_p)(struct ifnet *ifp);
|
2010-08-11 20:18:19 +00:00
|
|
|
#if defined(INET) || defined(INET6)
|
|
|
|
struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
|
|
|
|
int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
|
|
|
|
struct sockaddr *sa, struct rtentry *rt);
|
|
|
|
#endif
|
|
|
|
#ifdef INET
|
|
|
|
int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
|
|
|
|
u_int8_t **);
|
|
|
|
#endif
|
|
|
|
#ifdef INET6
|
|
|
|
struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
|
|
|
|
caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
|
|
|
|
const struct in6_addr *taddr);
|
|
|
|
#endif
|
2005-01-08 12:42:03 +00:00
|
|
|
|
2004-06-15 01:45:19 +00:00
|
|
|
struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
|
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
/*
|
|
|
|
* XXX: Style; these should be sorted alphabetically, and unprototyped
|
|
|
|
* static functions should be prototyped. Currently they are sorted by
|
|
|
|
* declaration order.
|
|
|
|
*/
|
2003-10-17 15:46:31 +00:00
|
|
|
static void if_attachdomain(void *);
|
|
|
|
static void if_attachdomain1(struct ifnet *);
|
2001-09-06 00:44:45 +00:00
|
|
|
static int ifconf(u_long, caddr_t);
|
2007-03-20 00:36:10 +00:00
|
|
|
static void if_freemulti(struct ifmultiaddr *);
|
2001-09-06 02:40:43 +00:00
|
|
|
static void if_init(void *);
|
2009-08-24 12:52:05 +00:00
|
|
|
static void if_grow(void);
|
2004-04-18 18:59:44 +00:00
|
|
|
static void if_route(struct ifnet *, int flag, int fam);
|
2005-07-14 13:56:51 +00:00
|
|
|
static int if_setflag(struct ifnet *, int, int, int *, int);
|
2008-11-22 05:55:56 +00:00
|
|
|
static int if_transmit(struct ifnet *ifp, struct mbuf *m);
|
2004-04-18 18:59:44 +00:00
|
|
|
static void if_unroute(struct ifnet *, int flag, int fam);
|
2001-10-17 18:07:05 +00:00
|
|
|
static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
|
2001-09-06 00:44:45 +00:00
|
|
|
static int if_rtdel(struct radix_node *, void *);
|
2001-09-29 05:55:04 +00:00
|
|
|
static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
|
2007-03-20 00:36:10 +00:00
|
|
|
static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
|
2005-04-20 09:30:54 +00:00
|
|
|
static void do_link_state_change(void *, int);
|
2006-06-19 22:20:45 +00:00
|
|
|
static int if_getgroup(struct ifgroupreq *, struct ifnet *);
|
|
|
|
static int if_getgroupmembers(struct ifgroupreq *);
|
2009-04-10 19:16:14 +00:00
|
|
|
static void if_delgroups(struct ifnet *);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
static void if_attach_internal(struct ifnet *, int);
|
|
|
|
static void if_detach_internal(struct ifnet *, int);
|
2008-11-22 05:55:56 +00:00
|
|
|
|
1999-11-22 02:45:11 +00:00
|
|
|
#ifdef INET6
|
|
|
|
/*
|
|
|
|
* XXX: declare here to avoid to include many inet6 related files..
|
|
|
|
* should be more generalized?
|
|
|
|
*/
|
2002-03-19 21:54:18 +00:00
|
|
|
extern void nd6_setmtu(struct ifnet *);
|
1999-11-22 02:45:11 +00:00
|
|
|
#endif
|
|
|
|
|
2010-04-29 11:52:42 +00:00
|
|
|
VNET_DEFINE(int, if_index);
|
|
|
|
int ifqmaxlen = IFQ_MAXLEN;
|
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(struct ifnethead, ifnet); /* depend on static init XXX */
|
|
|
|
VNET_DEFINE(struct ifgrouphead, ifg_head);
|
2010-04-29 11:52:42 +00:00
|
|
|
|
2010-11-22 19:32:54 +00:00
|
|
|
static VNET_DEFINE(int, if_indexlim) = 8;
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
/* Table of ifnet by index. */
|
2010-10-25 08:30:19 +00:00
|
|
|
VNET_DEFINE(struct ifindex_entry *, ifindex_table);
|
Build on Jeff Roberson's linker-set based dynamic per-CPU allocator
(DPCPU), as suggested by Peter Wemm, and implement a new per-virtual
network stack memory allocator. Modify vnet to use the allocator
instead of monolithic global container structures (vinet, ...). This
change solves many binary compatibility problems associated with
VIMAGE, and restores ELF symbols for virtualized global variables.
Each virtualized global variable exists as a "reference copy", and also
once per virtual network stack. Virtualized global variables are
tagged at compile-time, placing the in a special linker set, which is
loaded into a contiguous region of kernel memory. Virtualized global
variables in the base kernel are linked as normal, but those in modules
are copied and relocated to a reserved portion of the kernel's vnet
region with the help of a the kernel linker.
Virtualized global variables exist in per-vnet memory set up when the
network stack instance is created, and are initialized statically from
the reference copy. Run-time access occurs via an accessor macro, which
converts from the current vnet and requested symbol to a per-vnet
address. When "options VIMAGE" is not compiled into the kernel, normal
global ELF symbols will be used instead and indirection is avoided.
This change restores static initialization for network stack global
variables, restores support for non-global symbols and types, eliminates
the need for many subsystem constructors, eliminates large per-subsystem
structures that caused many binary compatibility issues both for
monitoring applications (netstat) and kernel modules, removes the
per-function INIT_VNET_*() macros throughout the stack, eliminates the
need for vnet_symmap ksym(2) munging, and eliminates duplicate
definitions of virtualized globals under VIMAGE_GLOBALS.
Bump __FreeBSD_version and update UPDATING.
Portions submitted by: bz
Reviewed by: bz, zec
Discussed with: gnn, jamie, jeff, jhb, julian, sam
Suggested by: peter
Approved by: re (kensmith)
2009-07-14 22:48:30 +00:00
|
|
|
|
2009-07-16 21:13:04 +00:00
|
|
|
#define V_if_indexlim VNET(if_indexlim)
|
|
|
|
#define V_ifindex_table VNET(ifindex_table)
|
2008-11-19 09:39:34 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
/*
|
|
|
|
* The global network interface list (V_ifnet) and related state (such as
|
|
|
|
* if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
|
|
|
|
* an rwlock. Either may be acquired shared to stablize the list, but both
|
|
|
|
* must be acquired writable to modify the list. This model allows us to
|
|
|
|
* both stablize the interface list during interrupt thread processing, but
|
|
|
|
* also to stablize it over long-running ioctls, without introducing priority
|
|
|
|
* inversions and deadlocks.
|
|
|
|
*/
|
|
|
|
struct rwlock ifnet_rwlock;
|
|
|
|
struct sx ifnet_sxlock;
|
|
|
|
|
2009-08-26 11:13:10 +00:00
|
|
|
/*
|
|
|
|
* The allocation of network interfaces is a rather non-atomic affair; we
|
|
|
|
* need to select an index before we are ready to expose the interface for
|
|
|
|
* use, so will use this pointer value to indicate reservation.
|
|
|
|
*/
|
|
|
|
#define IFNET_HOLD (void *)(uintptr_t)(-1)
|
|
|
|
|
2005-06-10 16:49:24 +00:00
|
|
|
static if_com_alloc_t *if_com_alloc[256];
|
|
|
|
static if_com_free_t *if_com_free[256];
|
2001-09-06 00:44:45 +00:00
|
|
|
|
2005-06-10 16:49:24 +00:00
|
|
|
MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
|
2001-09-06 00:44:45 +00:00
|
|
|
MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
|
|
|
|
MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
|
2001-07-02 20:49:25 +00:00
|
|
|
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
struct ifnet *
|
2008-12-18 04:50:44 +00:00
|
|
|
ifnet_byindex_locked(u_short idx)
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
{
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
|
|
|
|
if (idx > V_if_index)
|
|
|
|
return (NULL);
|
2009-08-26 11:13:10 +00:00
|
|
|
if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
|
|
|
|
return (NULL);
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
return (V_ifindex_table[idx].ife_ifnet);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct ifnet *
|
|
|
|
ifnet_byindex(u_short idx)
|
|
|
|
{
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
struct ifnet *ifp;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
ifp = ifnet_byindex_locked(idx);
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2008-12-18 04:50:44 +00:00
|
|
|
return (ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct ifnet *
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
ifnet_byindex_ref(u_short idx)
|
2008-12-18 04:50:44 +00:00
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
2008-12-18 04:50:44 +00:00
|
|
|
ifp = ifnet_byindex_locked(idx);
|
2009-04-23 09:32:30 +00:00
|
|
|
if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
return (NULL);
|
|
|
|
}
|
|
|
|
if_ref(ifp);
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
return (ifp);
|
|
|
|
}
|
|
|
|
|
2009-08-25 20:21:16 +00:00
|
|
|
/*
|
|
|
|
* Allocate an ifindex array entry; return 0 on success or an error on
|
|
|
|
* failure.
|
|
|
|
*/
|
|
|
|
static int
|
2009-08-26 11:13:10 +00:00
|
|
|
ifindex_alloc_locked(u_short *idxp)
|
2009-08-25 20:21:16 +00:00
|
|
|
{
|
|
|
|
u_short idx;
|
|
|
|
|
|
|
|
IFNET_WLOCK_ASSERT();
|
|
|
|
|
2011-01-24 22:21:58 +00:00
|
|
|
retry:
|
2009-08-25 20:21:16 +00:00
|
|
|
/*
|
2009-08-26 11:13:10 +00:00
|
|
|
* Try to find an empty slot below V_if_index. If we fail, take the
|
2009-08-25 20:21:16 +00:00
|
|
|
* next slot.
|
|
|
|
*/
|
|
|
|
for (idx = 1; idx <= V_if_index; idx++) {
|
2009-08-26 11:13:10 +00:00
|
|
|
if (V_ifindex_table[idx].ife_ifnet == NULL)
|
2009-08-25 20:21:16 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Catch if_index overflow. */
|
|
|
|
if (idx < 1)
|
|
|
|
return (ENOSPC);
|
2011-01-24 22:21:58 +00:00
|
|
|
if (idx >= V_if_indexlim) {
|
|
|
|
if_grow();
|
|
|
|
goto retry;
|
|
|
|
}
|
2009-08-25 20:21:16 +00:00
|
|
|
if (idx > V_if_index)
|
|
|
|
V_if_index = idx;
|
|
|
|
*idxp = idx;
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2009-08-26 11:13:10 +00:00
|
|
|
static void
|
|
|
|
ifindex_free_locked(u_short idx)
|
|
|
|
{
|
|
|
|
|
|
|
|
IFNET_WLOCK_ASSERT();
|
|
|
|
|
|
|
|
V_ifindex_table[idx].ife_ifnet = NULL;
|
|
|
|
while (V_if_index > 0 &&
|
|
|
|
V_ifindex_table[V_if_index].ife_ifnet == NULL)
|
|
|
|
V_if_index--;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
ifindex_free(u_short idx)
|
|
|
|
{
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
ifindex_free_locked(idx);
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
}
|
|
|
|
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
static void
|
2009-08-23 20:40:19 +00:00
|
|
|
ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
IFNET_WLOCK_ASSERT();
|
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
V_ifindex_table[idx].ife_ifnet = ifp;
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
}
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
static void
|
|
|
|
ifnet_setbyindex(u_short idx, struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
ifnet_setbyindex_locked(idx, ifp);
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
}
|
|
|
|
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
struct ifaddr *
|
|
|
|
ifaddr_byindex(u_short idx)
|
|
|
|
{
|
|
|
|
struct ifaddr *ifa;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
2008-12-18 04:50:44 +00:00
|
|
|
ifa = ifnet_byindex_locked(idx)->if_addr;
|
2009-06-23 20:19:09 +00:00
|
|
|
if (ifa != NULL)
|
|
|
|
ifa_ref(ifa);
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
return (ifa);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Network interface utility routines.
|
|
|
|
*
|
|
|
|
* Routines with ifa_ifwith* names take sockaddr *'s as
|
|
|
|
* parameters.
|
|
|
|
*/
|
2007-05-16 19:59:01 +00:00
|
|
|
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
static void
|
|
|
|
vnet_if_init(const void *unused __unused)
|
|
|
|
{
|
|
|
|
|
|
|
|
TAILQ_INIT(&V_ifnet);
|
|
|
|
TAILQ_INIT(&V_ifg_head);
|
2011-01-24 22:21:58 +00:00
|
|
|
IFNET_WLOCK();
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
if_grow(); /* create initial table */
|
2011-01-24 22:21:58 +00:00
|
|
|
IFNET_WUNLOCK();
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
vnet_if_clone_init();
|
|
|
|
}
|
2011-01-24 22:21:58 +00:00
|
|
|
VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
NULL);
|
|
|
|
|
1995-08-28 09:19:25 +00:00
|
|
|
/* ARGSUSED*/
|
2001-09-06 02:40:43 +00:00
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_init(void *dummy __unused)
|
2001-09-06 02:40:43 +00:00
|
|
|
{
|
|
|
|
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
IFNET_LOCK_INIT();
|
|
|
|
if_clone_init();
|
|
|
|
}
|
2011-01-24 22:21:58 +00:00
|
|
|
SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
|
First pass at separating per-vnet initializer functions
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
2009-04-06 22:29:41 +00:00
|
|
|
|
2001-09-06 02:40:43 +00:00
|
|
|
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
#ifdef VIMAGE
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
static void
|
|
|
|
vnet_if_uninit(const void *unused __unused)
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
{
|
|
|
|
|
2011-02-11 13:27:00 +00:00
|
|
|
VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
|
|
|
|
"not empty", __func__, __LINE__, &V_ifnet));
|
|
|
|
VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
|
|
|
|
"not empty", __func__, __LINE__, &V_ifg_head));
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
|
|
|
|
free((caddr_t)V_ifindex_table, M_IFNET);
|
|
|
|
}
|
Introduce and use a sysinit-based initialization scheme for virtual
network stacks, VNET_SYSINIT:
- Add VNET_SYSINIT and VNET_SYSUNINIT macros to declare events that will
occur each time a network stack is instantiated and destroyed. In the
!VIMAGE case, these are simply mapped into regular SYSINIT/SYSUNINIT.
For the VIMAGE case, we instead use SYSINIT's to track their order and
properties on registration, using them for each vnet when created/
destroyed, or immediately on module load for already-started vnets.
- Remove vnet_modinfo mechanism that existed to serve this purpose
previously, as well as its dependency scheme: we now just use the
SYSINIT ordering scheme.
- Implement VNET_DOMAIN_SET() to allow protocol domains to declare that
they want init functions to be called for each virtual network stack
rather than just once at boot, compiling down to DOMAIN_SET() in the
non-VIMAGE case.
- Walk all virtualized kernel subsystems and make use of these instead
of modinfo or DOMAIN_SET() for init/uninit events. In some cases,
convert modular components from using modevent to using sysinit (where
appropriate). In some cases, do minor rejuggling of SYSINIT ordering
to make room for or better manage events.
Portions submitted by: jhb (VNET_SYSINIT), bz (cleanup)
Discussed with: jhb, bz, julian, zec
Reviewed by: bz
Approved by: re (VIMAGE blanket)
2009-07-23 20:46:49 +00:00
|
|
|
VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
|
|
|
|
vnet_if_uninit, NULL);
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
#endif
|
|
|
|
|
2009-08-24 12:52:05 +00:00
|
|
|
static void
|
2001-09-06 02:40:43 +00:00
|
|
|
if_grow(void)
|
|
|
|
{
|
2011-01-24 22:21:58 +00:00
|
|
|
int oldlim;
|
2001-09-06 02:40:43 +00:00
|
|
|
u_int n;
|
|
|
|
struct ifindex_entry *e;
|
|
|
|
|
2011-01-24 22:21:58 +00:00
|
|
|
IFNET_WLOCK_ASSERT();
|
|
|
|
oldlim = V_if_indexlim;
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
n = (oldlim << 1) * sizeof(*e);
|
2005-06-10 16:49:24 +00:00
|
|
|
e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
|
2011-01-24 22:21:58 +00:00
|
|
|
IFNET_WLOCK();
|
|
|
|
if (V_if_indexlim != oldlim) {
|
|
|
|
free(e, M_IFNET);
|
|
|
|
return;
|
|
|
|
}
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if (V_ifindex_table != NULL) {
|
|
|
|
memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
|
|
|
|
free((caddr_t)V_ifindex_table, M_IFNET);
|
2001-09-06 02:40:43 +00:00
|
|
|
}
|
2011-01-24 22:21:58 +00:00
|
|
|
V_if_indexlim <<= 1;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
V_ifindex_table = e;
|
2001-09-06 02:40:43 +00:00
|
|
|
}
|
|
|
|
|
2005-06-10 16:49:24 +00:00
|
|
|
/*
|
2007-05-16 19:59:01 +00:00
|
|
|
* Allocate a struct ifnet and an index for an interface. A layer 2
|
|
|
|
* common structure will also be allocated if an allocation routine is
|
|
|
|
* registered for the passed type.
|
2005-06-10 16:49:24 +00:00
|
|
|
*/
|
2009-04-23 10:59:40 +00:00
|
|
|
struct ifnet *
|
2005-06-10 16:49:24 +00:00
|
|
|
if_alloc(u_char type)
|
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
2009-08-25 20:21:16 +00:00
|
|
|
u_short idx;
|
2005-06-10 16:49:24 +00:00
|
|
|
|
|
|
|
ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
|
2009-08-25 20:21:16 +00:00
|
|
|
IFNET_WLOCK();
|
2009-08-26 11:13:10 +00:00
|
|
|
if (ifindex_alloc_locked(&idx) != 0) {
|
2009-08-25 20:21:16 +00:00
|
|
|
IFNET_WUNLOCK();
|
2005-08-18 18:36:40 +00:00
|
|
|
free(ifp, M_IFNET);
|
|
|
|
return (NULL);
|
|
|
|
}
|
2009-08-26 11:13:10 +00:00
|
|
|
ifnet_setbyindex_locked(idx, IFNET_HOLD);
|
2009-08-25 20:21:16 +00:00
|
|
|
IFNET_WUNLOCK();
|
|
|
|
ifp->if_index = idx;
|
2005-06-10 16:49:24 +00:00
|
|
|
ifp->if_type = type;
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
ifp->if_alloctype = type;
|
2005-06-10 16:49:24 +00:00
|
|
|
if (if_com_alloc[type] != NULL) {
|
|
|
|
ifp->if_l2com = if_com_alloc[type](type, ifp);
|
2005-06-12 00:53:03 +00:00
|
|
|
if (ifp->if_l2com == NULL) {
|
2005-06-10 16:49:24 +00:00
|
|
|
free(ifp, M_IFNET);
|
2009-08-26 11:13:10 +00:00
|
|
|
ifindex_free(idx);
|
2005-06-12 00:53:03 +00:00
|
|
|
return (NULL);
|
|
|
|
}
|
2005-06-10 16:49:24 +00:00
|
|
|
}
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
|
|
|
|
IF_ADDR_LOCK_INIT(ifp);
|
2009-04-23 10:59:40 +00:00
|
|
|
TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
|
|
|
|
ifp->if_afdata_initialized = 0;
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
IF_AFDATA_LOCK_INIT(ifp);
|
2009-04-23 10:59:40 +00:00
|
|
|
TAILQ_INIT(&ifp->if_addrhead);
|
|
|
|
TAILQ_INIT(&ifp->if_prefixhead);
|
|
|
|
TAILQ_INIT(&ifp->if_multiaddrs);
|
|
|
|
TAILQ_INIT(&ifp->if_groups);
|
|
|
|
#ifdef MAC
|
|
|
|
mac_ifnet_init(ifp);
|
|
|
|
#endif
|
2009-06-15 19:50:03 +00:00
|
|
|
ifq_init(&ifp->if_snd, ifp);
|
2009-04-23 10:59:40 +00:00
|
|
|
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
refcount_init(&ifp->if_refcount, 1); /* Index reference. */
|
Introduce locking around use of ifindex_table, whose use was previously
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
2008-06-26 23:05:28 +00:00
|
|
|
ifnet_setbyindex(ifp->if_index, ifp);
|
2005-06-10 16:49:24 +00:00
|
|
|
return (ifp);
|
|
|
|
}
|
|
|
|
|
2007-05-16 19:59:01 +00:00
|
|
|
/*
|
2009-04-23 09:32:30 +00:00
|
|
|
* Do the actual work of freeing a struct ifnet, associated index, and layer
|
|
|
|
* 2 common structure. This call is made when the last reference to an
|
|
|
|
* interface is released.
|
2007-05-16 19:59:01 +00:00
|
|
|
*/
|
2009-04-23 09:32:30 +00:00
|
|
|
static void
|
|
|
|
if_free_internal(struct ifnet *ifp)
|
2005-06-10 16:49:24 +00:00
|
|
|
{
|
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
KASSERT((ifp->if_flags & IFF_DYING),
|
|
|
|
("if_free_internal: interface not dying"));
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
|
|
|
|
("%s: freeing unallocated ifnet", ifp->if_xname));
|
|
|
|
|
2009-08-26 11:13:10 +00:00
|
|
|
ifindex_free_locked(ifp->if_index);
|
2009-04-23 09:32:30 +00:00
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
if (if_com_free[ifp->if_alloctype] != NULL)
|
|
|
|
if_com_free[ifp->if_alloctype](ifp->if_l2com,
|
|
|
|
ifp->if_alloctype);
|
|
|
|
|
2009-04-23 10:59:40 +00:00
|
|
|
#ifdef MAC
|
|
|
|
mac_ifnet_destroy(ifp);
|
|
|
|
#endif /* MAC */
|
2010-01-27 00:30:07 +00:00
|
|
|
if (ifp->if_description != NULL)
|
|
|
|
free(ifp->if_description, M_IFDESCR);
|
2009-04-23 10:59:40 +00:00
|
|
|
IF_AFDATA_DESTROY(ifp);
|
2009-04-23 09:32:30 +00:00
|
|
|
IF_ADDR_LOCK_DESTROY(ifp);
|
2009-06-15 19:50:03 +00:00
|
|
|
ifq_delete(&ifp->if_snd);
|
2009-04-23 09:32:30 +00:00
|
|
|
free(ifp, M_IFNET);
|
2005-06-10 16:49:24 +00:00
|
|
|
}
|
|
|
|
|
2007-05-16 19:59:01 +00:00
|
|
|
/*
|
2009-04-23 09:32:30 +00:00
|
|
|
* This version should only be called by intefaces that switch their type
|
|
|
|
* after calling if_alloc(). if_free_type() will go away again now that we
|
|
|
|
* have if_alloctype to cache the original allocation type. For now, assert
|
|
|
|
* that they match, since we require that in practice.
|
2007-05-16 19:59:01 +00:00
|
|
|
*/
|
2005-06-10 16:49:24 +00:00
|
|
|
void
|
|
|
|
if_free_type(struct ifnet *ifp, u_char type)
|
|
|
|
{
|
|
|
|
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
KASSERT(ifp->if_alloctype == type,
|
|
|
|
("if_free_type: type (%d) != alloctype (%d)", type,
|
|
|
|
ifp->if_alloctype));
|
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
ifp->if_flags |= IFF_DYING; /* XXX: Locking */
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
if (!refcount_release(&ifp->if_refcount))
|
2005-06-10 16:49:24 +00:00
|
|
|
return;
|
2009-04-23 09:32:30 +00:00
|
|
|
if_free_internal(ifp);
|
|
|
|
}
|
2005-06-10 16:49:24 +00:00
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
/*
|
|
|
|
* This is the normal version of if_free(), used by device drivers to free a
|
|
|
|
* detached network interface. The contents of if_free_type() will move into
|
|
|
|
* here when if_free_type() goes away.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_free(struct ifnet *ifp)
|
|
|
|
{
|
2005-06-10 16:49:24 +00:00
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
if_free_type(ifp, ifp->if_alloctype);
|
2009-03-14 17:54:58 +00:00
|
|
|
}
|
2005-06-10 16:49:24 +00:00
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
/*
|
|
|
|
* Interfaces to keep an ifnet type-stable despite the possibility of the
|
|
|
|
* driver calling if_free(). If there are additional references, we defer
|
|
|
|
* freeing the underlying data structure.
|
|
|
|
*/
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
void
|
|
|
|
if_ref(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
/* We don't assert the ifnet list lock here, but arguably should. */
|
|
|
|
refcount_acquire(&ifp->if_refcount);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
if_rele(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
2009-04-23 09:32:30 +00:00
|
|
|
if (!refcount_release(&ifp->if_refcount))
|
|
|
|
return;
|
|
|
|
if_free_internal(ifp);
|
Start to address a number of races relating to use of ifnet pointers
after the corresponding interface has been destroyed:
(1) Add an ifnet refcount, ifp->if_refcount. Initialize it to 1 in
if_alloc(), and modify if_free_type() to decrement and check the
refcount.
(2) Add new if_ref() and if_rele() interfaces to allow kernel code
walking global interface lists to release IFNET_[RW]LOCK() yet
keep the ifnet stable. Currently, if_rele() is a no-op wrapper
around if_free(), but this may change in the future.
(3) Add new ifnet field, if_alloctype, which caches the type passed
to if_alloc(), but unlike if_type, won't be changed by drivers.
This allows asynchronous free's of the interface after the
driver has released it to still use the right type. Use that
instead of the type passed to if_free_type(), but assert that
they are the same (might have to rethink this if that doesn't
work out).
(4) Add a new ifnet_byindex_ref(), which looks up an interface by
index and returns a reference rather than a pointer to it.
(5) Fix if_alloc() to fully initialize the if_addr_mtx before hooking
up the ifnet to global lists.
(6) Modify sysctls in if_mib.c to use ifnet_byindex_ref() and release
the ifnet when done.
When this change is MFC'd, it will need to replace if_ispare fields
rather than adding new fields in order to avoid breaking the binary
interface. Once this change is MFC'd, if_free_type() should be
removed, as its 'type' argument is now optional.
This refcount is not appropriate for counting mbuf pkthdr references,
and also not for counting entry into the device driver via ifnet
function pointers. An rmlock may be appropriate for the latter.
Rather, this is about ensuring data structure stability when reaching
an ifnet via global ifnet lists and tables followed by copy in or out
of userspace.
MFC after: 3 weeks
Reported by: mdtancsa
Reviewed by: brooks
2009-04-21 22:43:32 +00:00
|
|
|
}
|
|
|
|
|
2008-11-22 05:55:56 +00:00
|
|
|
void
|
2009-06-15 19:50:03 +00:00
|
|
|
ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
|
2008-11-22 05:55:56 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
|
|
|
|
|
|
|
|
if (ifq->ifq_maxlen == 0)
|
|
|
|
ifq->ifq_maxlen = ifqmaxlen;
|
|
|
|
|
|
|
|
ifq->altq_type = 0;
|
|
|
|
ifq->altq_disc = NULL;
|
|
|
|
ifq->altq_flags &= ALTQF_CANTCHANGE;
|
|
|
|
ifq->altq_tbr = NULL;
|
|
|
|
ifq->altq_ifp = ifp;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
2009-06-15 19:50:03 +00:00
|
|
|
ifq_delete(struct ifaltq *ifq)
|
2008-11-22 05:55:56 +00:00
|
|
|
{
|
|
|
|
mtx_destroy(&ifq->ifq_mtx);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
2007-05-16 19:59:01 +00:00
|
|
|
* Perform generic interface initalization tasks and attach the interface
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
* to the list of "active" interfaces. If vmove flag is set on entry
|
|
|
|
* to if_attach_internal(), perform only a limited subset of initialization
|
|
|
|
* tasks, given that we are moving from one vnet to another an ifnet which
|
|
|
|
* has already been fully initialized.
|
2007-05-16 19:59:01 +00:00
|
|
|
*
|
|
|
|
* XXX:
|
|
|
|
* - The decision to return void and thus require this function to
|
|
|
|
* succeed is questionable.
|
|
|
|
* - We should probably do more sanity checking. For instance we don't
|
|
|
|
* do anything to insure if_xname is unique or non-empty.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
|
|
|
void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_attach(struct ifnet *ifp)
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
if_attach_internal(ifp, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
if_attach_internal(struct ifnet *ifp, int vmove)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
|
|
|
unsigned socksize, ifasize;
|
1996-01-24 21:12:23 +00:00
|
|
|
int namelen, masklen;
|
2003-10-23 13:49:10 +00:00
|
|
|
struct sockaddr_dl *sdl;
|
|
|
|
struct ifaddr *ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2005-06-10 16:49:24 +00:00
|
|
|
if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
|
|
|
|
panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
|
|
|
|
ifp->if_xname);
|
|
|
|
|
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
|
|
|
|
ifp->if_vnet = curvnet;
|
Introduce an infrastructure for dismantling vnet instances.
Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state. The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.
While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers. Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.
Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels. Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.
Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)
2009-06-08 17:15:40 +00:00
|
|
|
if (ifp->if_home_vnet == NULL)
|
|
|
|
ifp->if_home_vnet = curvnet;
|
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
|
|
|
#endif
|
|
|
|
|
2006-06-19 22:20:45 +00:00
|
|
|
if_addgroup(ifp, IFG_ALL);
|
|
|
|
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2005-02-25 19:46:41 +00:00
|
|
|
ifp->if_data.ifi_epoch = time_uptime;
|
2005-06-10 16:49:24 +00:00
|
|
|
ifp->if_data.ifi_datalen = sizeof(struct if_data);
|
2009-04-23 10:59:40 +00:00
|
|
|
|
2009-04-16 23:05:10 +00:00
|
|
|
KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
|
|
|
|
(ifp->if_transmit != NULL && ifp->if_qflush != NULL),
|
|
|
|
("transmit and qflush must both either be set or both be NULL"));
|
|
|
|
if (ifp->if_transmit == NULL) {
|
|
|
|
ifp->if_transmit = if_transmit;
|
|
|
|
ifp->if_qflush = if_qflush;
|
|
|
|
}
|
|
|
|
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
if (!vmove) {
|
2002-07-31 16:16:03 +00:00
|
|
|
#ifdef MAC
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
mac_ifnet_create(ifp);
|
2002-07-31 16:16:03 +00:00
|
|
|
#endif
|
|
|
|
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
/*
|
|
|
|
* Create a Link Level name for this device.
|
|
|
|
*/
|
|
|
|
namelen = strlen(ifp->if_xname);
|
|
|
|
/*
|
|
|
|
* Always save enough space for any possiable name so we
|
|
|
|
* can do a rename in place later.
|
|
|
|
*/
|
|
|
|
masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
|
|
|
|
socksize = masklen + ifp->if_addrlen;
|
|
|
|
if (socksize < sizeof(*sdl))
|
|
|
|
socksize = sizeof(*sdl);
|
|
|
|
socksize = roundup2(socksize, sizeof(long));
|
|
|
|
ifasize = sizeof(*ifa) + 2 * socksize;
|
|
|
|
ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
|
2009-06-21 19:30:33 +00:00
|
|
|
ifa_init(ifa);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
sdl = (struct sockaddr_dl *)(ifa + 1);
|
|
|
|
sdl->sdl_len = socksize;
|
|
|
|
sdl->sdl_family = AF_LINK;
|
|
|
|
bcopy(ifp->if_xname, sdl->sdl_data, namelen);
|
|
|
|
sdl->sdl_nlen = namelen;
|
|
|
|
sdl->sdl_index = ifp->if_index;
|
|
|
|
sdl->sdl_type = ifp->if_type;
|
|
|
|
ifp->if_addr = ifa;
|
|
|
|
ifa->ifa_ifp = ifp;
|
|
|
|
ifa->ifa_rtrequest = link_rtrequest;
|
|
|
|
ifa->ifa_addr = (struct sockaddr *)sdl;
|
|
|
|
sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
|
|
|
|
ifa->ifa_netmask = (struct sockaddr *)sdl;
|
|
|
|
sdl->sdl_len = masklen;
|
|
|
|
while (namelen != 0)
|
|
|
|
sdl->sdl_data[--namelen] = 0xff;
|
|
|
|
TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
|
|
|
|
/* Reliably crash if used uninitialized. */
|
|
|
|
ifp->if_broadcastaddr = NULL;
|
|
|
|
}
|
2009-08-24 10:14:09 +00:00
|
|
|
#ifdef VIMAGE
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* Update the interface index in the link layer address
|
|
|
|
* of the interface.
|
|
|
|
*/
|
|
|
|
for (ifa = ifp->if_addr; ifa != NULL;
|
|
|
|
ifa = TAILQ_NEXT(ifa, ifa_link)) {
|
|
|
|
if (ifa->ifa_addr->sa_family == AF_LINK) {
|
|
|
|
sdl = (struct sockaddr_dl *)ifa->ifa_addr;
|
|
|
|
sdl->sdl_index = ifp->if_index;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
2008-05-17 03:38:13 +00:00
|
|
|
|
2006-06-21 06:02:35 +00:00
|
|
|
IFNET_WLOCK();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
#ifdef VIMAGE
|
2009-07-19 17:40:45 +00:00
|
|
|
curvnet->vnet_ifcnt++;
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
#endif
|
2006-06-21 06:02:35 +00:00
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
2004-11-30 22:38:37 +00:00
|
|
|
if (domain_init_status >= 2)
|
2003-10-17 15:46:31 +00:00
|
|
|
if_attachdomain1(ifp);
|
|
|
|
|
2004-02-26 04:27:55 +00:00
|
|
|
EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
if (IS_DEFAULT_VNET(curvnet))
|
|
|
|
devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
|
2004-02-26 04:27:55 +00:00
|
|
|
|
2002-01-18 14:33:04 +00:00
|
|
|
/* Announce the interface. */
|
|
|
|
rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1999-04-16 21:22:55 +00:00
|
|
|
|
2003-10-17 15:46:31 +00:00
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_attachdomain(void *dummy)
|
2003-10-17 15:46:31 +00:00
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
int s;
|
|
|
|
|
|
|
|
s = splnet();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link)
|
2003-10-17 15:46:31 +00:00
|
|
|
if_attachdomain1(ifp);
|
|
|
|
splx(s);
|
|
|
|
}
|
2004-11-30 22:38:37 +00:00
|
|
|
SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
|
2003-10-17 15:46:31 +00:00
|
|
|
if_attachdomain, NULL);
|
|
|
|
|
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_attachdomain1(struct ifnet *ifp)
|
2003-10-17 15:46:31 +00:00
|
|
|
{
|
|
|
|
struct domain *dp;
|
|
|
|
int s;
|
|
|
|
|
|
|
|
s = splnet();
|
|
|
|
|
2003-10-24 16:57:59 +00:00
|
|
|
/*
|
|
|
|
* Since dp->dom_ifattach calls malloc() with M_WAITOK, we
|
|
|
|
* cannot lock ifp->if_afdata initialization, entirely.
|
|
|
|
*/
|
|
|
|
if (IF_AFDATA_TRYLOCK(ifp) == 0) {
|
|
|
|
splx(s);
|
|
|
|
return;
|
|
|
|
}
|
2004-11-30 22:38:37 +00:00
|
|
|
if (ifp->if_afdata_initialized >= domain_init_status) {
|
2003-10-24 16:57:59 +00:00
|
|
|
IF_AFDATA_UNLOCK(ifp);
|
|
|
|
splx(s);
|
2004-11-23 23:31:33 +00:00
|
|
|
printf("if_attachdomain called more than once on %s\n",
|
|
|
|
ifp->if_xname);
|
2003-10-24 16:57:59 +00:00
|
|
|
return;
|
|
|
|
}
|
2004-11-30 22:38:37 +00:00
|
|
|
ifp->if_afdata_initialized = domain_init_status;
|
2003-10-24 16:57:59 +00:00
|
|
|
IF_AFDATA_UNLOCK(ifp);
|
|
|
|
|
2003-10-17 15:46:31 +00:00
|
|
|
/* address family dependent data region */
|
|
|
|
bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
|
|
|
|
for (dp = domains; dp; dp = dp->dom_next) {
|
|
|
|
if (dp->dom_ifattach)
|
|
|
|
ifp->if_afdata[dp->dom_family] =
|
|
|
|
(*dp->dom_ifattach)(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
splx(s);
|
|
|
|
}
|
|
|
|
|
2005-05-25 13:52:03 +00:00
|
|
|
/*
|
2007-03-20 00:36:10 +00:00
|
|
|
* Remove any unicast or broadcast network addresses from an interface.
|
2005-05-25 13:52:03 +00:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_purgeaddrs(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
struct ifaddr *ifa, *next;
|
|
|
|
|
|
|
|
TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
|
2006-06-29 19:22:05 +00:00
|
|
|
if (ifa->ifa_addr->sa_family == AF_LINK)
|
2005-05-25 13:52:03 +00:00
|
|
|
continue;
|
|
|
|
#ifdef INET
|
|
|
|
/* XXX: Ugly!! ad hoc just for INET */
|
2006-06-29 19:22:05 +00:00
|
|
|
if (ifa->ifa_addr->sa_family == AF_INET) {
|
2005-05-25 13:52:03 +00:00
|
|
|
struct ifaliasreq ifr;
|
|
|
|
|
|
|
|
bzero(&ifr, sizeof(ifr));
|
|
|
|
ifr.ifra_addr = *ifa->ifa_addr;
|
|
|
|
if (ifa->ifa_dstaddr)
|
|
|
|
ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
|
|
|
|
if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
|
|
|
|
NULL) == 0)
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
#endif /* INET */
|
|
|
|
#ifdef INET6
|
2006-06-29 19:22:05 +00:00
|
|
|
if (ifa->ifa_addr->sa_family == AF_INET6) {
|
2005-05-25 13:52:03 +00:00
|
|
|
in6_purgeaddr(ifa);
|
|
|
|
/* ifp_addrhead is already updated */
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
#endif /* INET6 */
|
|
|
|
TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
|
2009-06-21 19:30:33 +00:00
|
|
|
ifa_free(ifa);
|
2005-05-25 13:52:03 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
/*
|
2010-01-24 16:17:58 +00:00
|
|
|
* Remove any multicast network addresses from an interface when an ifnet
|
|
|
|
* is going away.
|
2007-03-20 00:36:10 +00:00
|
|
|
*/
|
2010-01-24 16:17:58 +00:00
|
|
|
static void
|
2007-03-20 00:36:10 +00:00
|
|
|
if_purgemaddrs(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
struct ifmultiaddr *ifma;
|
|
|
|
struct ifmultiaddr *next;
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
|
|
|
|
if_delmulti_locked(ifp, ifma, 1);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
|
|
|
|
1999-04-16 21:22:55 +00:00
|
|
|
/*
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
* Detach an interface, removing it from the list of "active" interfaces.
|
|
|
|
* If vmove flag is set on entry to if_detach_internal(), perform only a
|
|
|
|
* limited subset of cleanup tasks, given that we are moving an ifnet from
|
|
|
|
* one vnet to another, where it must be fully operational.
|
2005-09-18 17:36:28 +00:00
|
|
|
*
|
|
|
|
* XXXRW: There are some significant questions about event ordering, and
|
|
|
|
* how to prevent things from starting to use the interface during detach.
|
1999-04-16 21:22:55 +00:00
|
|
|
*/
|
|
|
|
void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_detach(struct ifnet *ifp)
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
if_detach_internal(ifp, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
if_detach_internal(struct ifnet *ifp, int vmove)
|
1999-04-16 21:22:55 +00:00
|
|
|
{
|
2005-05-25 13:52:03 +00:00
|
|
|
struct ifaddr *ifa;
|
1999-12-17 06:46:07 +00:00
|
|
|
struct radix_node_head *rnh;
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
int i, j;
|
2003-10-17 15:46:31 +00:00
|
|
|
struct domain *dp;
|
2004-08-06 09:08:33 +00:00
|
|
|
struct ifnet *iter;
|
2006-06-21 06:02:35 +00:00
|
|
|
int found = 0;
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(iter, &V_ifnet, if_link)
|
2006-06-21 06:02:35 +00:00
|
|
|
if (iter == ifp) {
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_REMOVE(&V_ifnet, ifp, if_link);
|
2006-06-21 06:02:35 +00:00
|
|
|
found = 1;
|
|
|
|
break;
|
|
|
|
}
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
#ifdef VIMAGE
|
|
|
|
if (found)
|
2009-07-19 17:40:45 +00:00
|
|
|
curvnet->vnet_ifcnt--;
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
#endif
|
2006-06-21 06:02:35 +00:00
|
|
|
IFNET_WUNLOCK();
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
if (!found) {
|
|
|
|
if (vmove)
|
2010-02-20 21:43:36 +00:00
|
|
|
panic("%s: ifp=%p not on the ifnet tailq %p",
|
|
|
|
__func__, ifp, &V_ifnet);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
else
|
|
|
|
return; /* XXX this should panic as well? */
|
|
|
|
}
|
1999-04-16 21:22:55 +00:00
|
|
|
|
2005-04-20 09:30:54 +00:00
|
|
|
/*
|
|
|
|
* Remove/wait for pending events.
|
|
|
|
*/
|
|
|
|
taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
|
|
|
|
|
1999-04-16 21:22:55 +00:00
|
|
|
/*
|
|
|
|
* Remove routes and flush queues.
|
|
|
|
*/
|
|
|
|
if_down(ifp);
|
2004-06-13 17:29:10 +00:00
|
|
|
#ifdef ALTQ
|
|
|
|
if (ALTQ_IS_ENABLED(&ifp->if_snd))
|
|
|
|
altq_disable(&ifp->if_snd);
|
|
|
|
if (ALTQ_IS_ATTACHED(&ifp->if_snd))
|
|
|
|
altq_detach(&ifp->if_snd);
|
|
|
|
#endif
|
1999-04-16 21:22:55 +00:00
|
|
|
|
2005-05-25 13:52:03 +00:00
|
|
|
if_purgeaddrs(ifp);
|
1999-04-16 21:22:55 +00:00
|
|
|
|
2005-09-18 17:36:28 +00:00
|
|
|
#ifdef INET
|
|
|
|
in_ifdetach(ifp);
|
|
|
|
#endif
|
|
|
|
|
2001-06-11 12:39:29 +00:00
|
|
|
#ifdef INET6
|
|
|
|
/*
|
|
|
|
* Remove all IPv6 kernel structs related to ifp. This should be done
|
|
|
|
* before removing routing entries below, since IPv6 interface direct
|
|
|
|
* routes are expected to be removed by the IPv6-specific kernel API.
|
|
|
|
* Otherwise, the kernel will detect some inconsistency and bark it.
|
|
|
|
*/
|
|
|
|
in6_ifdetach(ifp);
|
|
|
|
#endif
|
2007-03-20 00:36:10 +00:00
|
|
|
if_purgemaddrs(ifp);
|
|
|
|
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
if (!vmove) {
|
|
|
|
/*
|
|
|
|
* Prevent further calls into the device driver via ifnet.
|
|
|
|
*/
|
|
|
|
if_dead(ifp);
|
2004-04-19 17:28:15 +00:00
|
|
|
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
/*
|
|
|
|
* Remove link ifaddr pointer and maybe decrement if_index.
|
|
|
|
* Clean up all addresses.
|
|
|
|
*/
|
|
|
|
ifp->if_addr = NULL;
|
|
|
|
|
|
|
|
/* We can now free link ifaddr. */
|
|
|
|
if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
|
|
|
|
ifa = TAILQ_FIRST(&ifp->if_addrhead);
|
|
|
|
TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
|
2009-06-21 19:30:33 +00:00
|
|
|
ifa_free(ifa);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
}
|
2004-08-06 09:08:33 +00:00
|
|
|
}
|
2003-10-16 13:38:29 +00:00
|
|
|
|
1999-12-17 06:46:07 +00:00
|
|
|
/*
|
|
|
|
* Delete all remaining routes using this interface
|
|
|
|
* Unfortuneatly the only way to do this is to slog through
|
|
|
|
* the entire routing table looking for routes which point
|
|
|
|
* to this interface...oh well...
|
|
|
|
*/
|
|
|
|
for (i = 1; i <= AF_MAX; i++) {
|
2008-11-24 17:34:00 +00:00
|
|
|
for (j = 0; j < rt_numfibs; j++) {
|
2009-06-01 15:49:42 +00:00
|
|
|
rnh = rt_tables_get_rnh(j, i);
|
|
|
|
if (rnh == NULL)
|
2008-11-24 17:34:00 +00:00
|
|
|
continue;
|
|
|
|
RADIX_NODE_HEAD_LOCK(rnh);
|
|
|
|
(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
|
|
|
|
RADIX_NODE_HEAD_UNLOCK(rnh);
|
|
|
|
}
|
1999-12-17 06:46:07 +00:00
|
|
|
}
|
|
|
|
|
2002-01-18 14:33:04 +00:00
|
|
|
/* Announce that the interface is gone. */
|
|
|
|
rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
|
2005-07-14 20:26:43 +00:00
|
|
|
EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
if (IS_DEFAULT_VNET(curvnet))
|
|
|
|
devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
|
2009-04-10 19:16:14 +00:00
|
|
|
if_delgroups(ifp);
|
2002-01-18 14:33:04 +00:00
|
|
|
|
2010-04-11 11:51:44 +00:00
|
|
|
/*
|
|
|
|
* We cannot hold the lock over dom_ifdetach calls as they might
|
|
|
|
* sleep, for example trying to drain a callout, thus open up the
|
|
|
|
* theoretical race with re-attaching.
|
|
|
|
*/
|
2003-10-24 16:57:59 +00:00
|
|
|
IF_AFDATA_LOCK(ifp);
|
2010-04-11 11:51:44 +00:00
|
|
|
i = ifp->if_afdata_initialized;
|
|
|
|
ifp->if_afdata_initialized = 0;
|
|
|
|
IF_AFDATA_UNLOCK(ifp);
|
|
|
|
for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
|
2003-10-17 15:46:31 +00:00
|
|
|
if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
|
|
|
|
(*dp->dom_ifdetach)(ifp,
|
|
|
|
ifp->if_afdata[dp->dom_family]);
|
|
|
|
}
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
}
|
|
|
|
|
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
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
/*
|
|
|
|
* if_vmove() performs a limited version of if_detach() in current
|
|
|
|
* vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
|
|
|
|
* An attempt is made to shrink if_index in current vnet, find an
|
|
|
|
* unused if_index in target vnet and calls if_grow() if necessary,
|
|
|
|
* and finally find an unused if_xname for the target vnet.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
|
|
|
|
{
|
2009-08-25 20:21:16 +00:00
|
|
|
u_short idx;
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Detach from current vnet, but preserve LLADDR info, do not
|
|
|
|
* mark as dead etc. so that the ifnet can be reattached later.
|
|
|
|
*/
|
|
|
|
if_detach_internal(ifp, 1);
|
|
|
|
|
|
|
|
/*
|
2009-08-23 20:40:19 +00:00
|
|
|
* Unlink the ifnet from ifindex_table[] in current vnet, and shrink
|
|
|
|
* the if_index for that vnet if possible.
|
|
|
|
*
|
|
|
|
* NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
|
|
|
|
* or we'd lock on one vnet and unlock on another.
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
*/
|
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
|
|
|
IFNET_WLOCK();
|
2009-08-26 11:13:10 +00:00
|
|
|
ifindex_free_locked(ifp->if_index);
|
2010-08-13 18:17:32 +00:00
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Perform interface-specific reassignment tasks, if provided by
|
|
|
|
* the driver.
|
|
|
|
*/
|
|
|
|
if (ifp->if_reassign != NULL)
|
|
|
|
ifp->if_reassign(ifp, new_vnet, NULL);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Switch to the context of the target vnet.
|
|
|
|
*/
|
|
|
|
CURVNET_SET_QUIET(new_vnet);
|
|
|
|
|
2010-08-13 18:17:32 +00:00
|
|
|
IFNET_WLOCK();
|
2009-08-26 11:13:10 +00:00
|
|
|
if (ifindex_alloc_locked(&idx) != 0) {
|
2009-08-25 20:21:16 +00:00
|
|
|
IFNET_WUNLOCK();
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
panic("if_index overflow");
|
2009-08-25 20:21:16 +00:00
|
|
|
}
|
|
|
|
ifp->if_index = idx;
|
2009-08-23 20:40:19 +00:00
|
|
|
ifnet_setbyindex_locked(ifp->if_index, ifp);
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
if_attach_internal(ifp, 1);
|
|
|
|
|
|
|
|
CURVNET_RESTORE();
|
1999-12-17 06:46:07 +00:00
|
|
|
}
|
2009-07-26 11:29:26 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Move an ifnet to or from another child prison/vnet, specified by the jail id.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
|
|
|
|
{
|
|
|
|
struct prison *pr;
|
|
|
|
struct ifnet *difp;
|
|
|
|
|
|
|
|
/* Try to find the prison within our visibility. */
|
|
|
|
sx_slock(&allprison_lock);
|
|
|
|
pr = prison_find_child(td->td_ucred->cr_prison, jid);
|
|
|
|
sx_sunlock(&allprison_lock);
|
|
|
|
if (pr == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
prison_hold_locked(pr);
|
|
|
|
mtx_unlock(&pr->pr_mtx);
|
|
|
|
|
|
|
|
/* Do not try to move the iface from and to the same prison. */
|
|
|
|
if (pr->pr_vnet == ifp->if_vnet) {
|
|
|
|
prison_free(pr);
|
|
|
|
return (EEXIST);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Make sure the named iface does not exists in the dst. prison/vnet. */
|
|
|
|
/* XXX Lock interfaces to avoid races. */
|
2009-08-14 22:46:45 +00:00
|
|
|
CURVNET_SET_QUIET(pr->pr_vnet);
|
2009-07-26 11:29:26 +00:00
|
|
|
difp = ifunit(ifname);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
if (difp != NULL) {
|
|
|
|
prison_free(pr);
|
|
|
|
return (EEXIST);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Move the interface into the child jail/vnet. */
|
|
|
|
if_vmove(ifp, pr->pr_vnet);
|
|
|
|
|
|
|
|
/* Report the new if_xname back to the userland. */
|
|
|
|
sprintf(ifname, "%s", ifp->if_xname);
|
|
|
|
|
|
|
|
prison_free(pr);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
if_vmove_reclaim(struct thread *td, char *ifname, int jid)
|
|
|
|
{
|
|
|
|
struct prison *pr;
|
|
|
|
struct vnet *vnet_dst;
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
|
|
|
/* Try to find the prison within our visibility. */
|
|
|
|
sx_slock(&allprison_lock);
|
|
|
|
pr = prison_find_child(td->td_ucred->cr_prison, jid);
|
|
|
|
sx_sunlock(&allprison_lock);
|
|
|
|
if (pr == NULL)
|
|
|
|
return (ENXIO);
|
|
|
|
prison_hold_locked(pr);
|
|
|
|
mtx_unlock(&pr->pr_mtx);
|
|
|
|
|
|
|
|
/* Make sure the named iface exists in the source prison/vnet. */
|
|
|
|
CURVNET_SET(pr->pr_vnet);
|
|
|
|
ifp = ifunit(ifname); /* XXX Lock to avoid races. */
|
|
|
|
if (ifp == NULL) {
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
prison_free(pr);
|
|
|
|
return (ENXIO);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Do not try to move the iface from and to the same prison. */
|
|
|
|
vnet_dst = TD_TO_VNET(td);
|
|
|
|
if (vnet_dst == ifp->if_vnet) {
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
prison_free(pr);
|
|
|
|
return (EEXIST);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Get interface back from child jail/vnet. */
|
|
|
|
if_vmove(ifp, vnet_dst);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
|
|
|
|
/* Report the new if_xname back to the userland. */
|
|
|
|
sprintf(ifname, "%s", ifp->if_xname);
|
|
|
|
|
|
|
|
prison_free(pr);
|
|
|
|
return (0);
|
|
|
|
}
|
Introduce the if_vmove() function, which will be used in the future
for reassigning ifnets from one vnet to another.
if_vmove() works by calling a restricted subset of actions normally
executed by if_detach() on an ifnet in the current vnet, and then
switches to the target vnet and executes an appropriate subset of
if_attach() actions there.
if_attach() and if_detach() have become wrapper functions around
if_attach_internal() and if_detach_internal(), where the later
variants have an additional argument, a flag indicating whether a
full attach or detach sequence is to be executed, or only a
restricted subset suitable for moving an ifnet from one vnet to
another. Hence, if_vmove() will not call if_detach() and if_attach()
directly, but will call the if_detach_internal() and
if_attach_internal() variants instead, with the vmove flag set.
While here, staticize ifnet_setbyindex() since it is not referenced
from outside of sys/net/if.c.
Also rename ifccnt field in struct vimage to ifcnt, and do some minor
whitespace garbage collection where appropriate.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz, rwatson, brooks?
Approved by: julian (mentor)
2009-05-22 22:09:00 +00:00
|
|
|
#endif /* VIMAGE */
|
1999-12-17 06:46:07 +00:00
|
|
|
|
2006-06-19 22:20:45 +00:00
|
|
|
/*
|
|
|
|
* Add a group to an interface
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
if_addgroup(struct ifnet *ifp, const char *groupname)
|
|
|
|
{
|
|
|
|
struct ifg_list *ifgl;
|
|
|
|
struct ifg_group *ifg = NULL;
|
|
|
|
struct ifg_member *ifgm;
|
|
|
|
|
|
|
|
if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
|
|
|
|
groupname[strlen(groupname) - 1] <= '9')
|
|
|
|
return (EINVAL);
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
|
|
|
|
if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
return (EEXIST);
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
|
|
|
|
M_NOWAIT)) == NULL) {
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
|
|
|
|
M_TEMP, M_NOWAIT)) == NULL) {
|
|
|
|
free(ifgl, M_TEMP);
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
|
|
|
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
|
2006-06-19 22:20:45 +00:00
|
|
|
if (!strcmp(ifg->ifg_group, groupname))
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (ifg == NULL) {
|
|
|
|
if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
|
|
|
|
M_TEMP, M_NOWAIT)) == NULL) {
|
|
|
|
free(ifgl, M_TEMP);
|
|
|
|
free(ifgm, M_TEMP);
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
|
|
|
strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
|
|
|
|
ifg->ifg_refcnt = 0;
|
|
|
|
TAILQ_INIT(&ifg->ifg_members);
|
|
|
|
EVENTHANDLER_INVOKE(group_attach_event, ifg);
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
|
2006-06-19 22:20:45 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
ifg->ifg_refcnt++;
|
|
|
|
ifgl->ifgl_group = ifg;
|
|
|
|
ifgm->ifgm_ifp = ifp;
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
|
|
|
|
TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
EVENTHANDLER_INVOKE(group_change_event, groupname);
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove a group from an interface
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
if_delgroup(struct ifnet *ifp, const char *groupname)
|
|
|
|
{
|
|
|
|
struct ifg_list *ifgl;
|
|
|
|
struct ifg_member *ifgm;
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
|
|
|
|
if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
|
|
|
|
break;
|
|
|
|
if (ifgl == NULL) {
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
return (ENOENT);
|
|
|
|
}
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
|
|
|
TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
|
|
|
|
if (ifgm->ifgm_ifp == ifp)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (ifgm != NULL) {
|
|
|
|
TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
|
|
|
|
free(ifgm, M_TEMP);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (--ifgl->ifgl_group->ifg_refcnt == 0) {
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
|
2006-06-19 22:20:45 +00:00
|
|
|
EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
|
|
|
|
free(ifgl->ifgl_group, M_TEMP);
|
|
|
|
}
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
free(ifgl, M_TEMP);
|
|
|
|
|
|
|
|
EVENTHANDLER_INVOKE(group_change_event, groupname);
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2009-04-10 19:16:14 +00:00
|
|
|
/*
|
|
|
|
* Remove an interface from all groups
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
if_delgroups(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
struct ifg_list *ifgl;
|
|
|
|
struct ifg_member *ifgm;
|
|
|
|
char groupname[IFNAMSIZ];
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
while (!TAILQ_EMPTY(&ifp->if_groups)) {
|
|
|
|
ifgl = TAILQ_FIRST(&ifp->if_groups);
|
|
|
|
|
|
|
|
strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
|
|
|
TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
|
|
|
|
if (ifgm->ifgm_ifp == ifp)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (ifgm != NULL) {
|
|
|
|
TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
|
|
|
|
ifgm_next);
|
|
|
|
free(ifgm, M_TEMP);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (--ifgl->ifgl_group->ifg_refcnt == 0) {
|
|
|
|
TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
|
|
|
|
EVENTHANDLER_INVOKE(group_detach_event,
|
|
|
|
ifgl->ifgl_group);
|
|
|
|
free(ifgl->ifgl_group, M_TEMP);
|
|
|
|
}
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
|
|
|
|
free(ifgl, M_TEMP);
|
|
|
|
|
|
|
|
EVENTHANDLER_INVOKE(group_change_event, groupname);
|
|
|
|
|
|
|
|
IFNET_WLOCK();
|
|
|
|
}
|
|
|
|
IFNET_WUNLOCK();
|
|
|
|
}
|
|
|
|
|
2006-06-19 22:20:45 +00:00
|
|
|
/*
|
|
|
|
* Stores all groups from an interface in memory pointed
|
|
|
|
* to by data
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
int len, error;
|
|
|
|
struct ifg_list *ifgl;
|
|
|
|
struct ifg_req ifgrq, *ifgp;
|
|
|
|
struct ifgroupreq *ifgr = data;
|
|
|
|
|
|
|
|
if (ifgr->ifgr_len == 0) {
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
|
|
|
|
ifgr->ifgr_len += sizeof(struct ifg_req);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
len = ifgr->ifgr_len;
|
|
|
|
ifgp = ifgr->ifgr_groups;
|
|
|
|
/* XXX: wire */
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
|
|
|
|
if (len < sizeof(ifgrq)) {
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
bzero(&ifgrq, sizeof ifgrq);
|
|
|
|
strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
|
|
|
|
sizeof(ifgrq.ifgrq_group));
|
|
|
|
if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
len -= sizeof(ifgrq);
|
|
|
|
ifgp++;
|
|
|
|
}
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Stores all members of a group in memory pointed to by data
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
if_getgroupmembers(struct ifgroupreq *data)
|
|
|
|
{
|
|
|
|
struct ifgroupreq *ifgr = data;
|
|
|
|
struct ifg_group *ifg;
|
|
|
|
struct ifg_member *ifgm;
|
|
|
|
struct ifg_req ifgrq, *ifgp;
|
|
|
|
int len, error;
|
|
|
|
|
|
|
|
IFNET_RLOCK();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
|
2006-06-19 22:20:45 +00:00
|
|
|
if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
|
|
|
|
break;
|
|
|
|
if (ifg == NULL) {
|
|
|
|
IFNET_RUNLOCK();
|
|
|
|
return (ENOENT);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ifgr->ifgr_len == 0) {
|
|
|
|
TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
|
|
|
|
ifgr->ifgr_len += sizeof(ifgrq);
|
|
|
|
IFNET_RUNLOCK();
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
len = ifgr->ifgr_len;
|
|
|
|
ifgp = ifgr->ifgr_groups;
|
|
|
|
TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
|
|
|
|
if (len < sizeof(ifgrq)) {
|
|
|
|
IFNET_RUNLOCK();
|
|
|
|
return (EINVAL);
|
|
|
|
}
|
|
|
|
bzero(&ifgrq, sizeof ifgrq);
|
|
|
|
strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
|
|
|
|
sizeof(ifgrq.ifgrq_member));
|
|
|
|
if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
|
|
|
|
IFNET_RUNLOCK();
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
len -= sizeof(ifgrq);
|
|
|
|
ifgp++;
|
|
|
|
}
|
|
|
|
IFNET_RUNLOCK();
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
1999-12-17 06:46:07 +00:00
|
|
|
/*
|
|
|
|
* Delete Routes for a Network Interface
|
2003-10-23 13:49:10 +00:00
|
|
|
*
|
1999-12-17 06:46:07 +00:00
|
|
|
* Called for each routing entry via the rnh->rnh_walktree() call above
|
|
|
|
* to delete all route entries referencing a detaching network interface.
|
|
|
|
*
|
|
|
|
* Arguments:
|
|
|
|
* rn pointer to node in the routing table
|
|
|
|
* arg argument passed to rnh->rnh_walktree() - detaching interface
|
|
|
|
*
|
|
|
|
* Returns:
|
|
|
|
* 0 successful
|
|
|
|
* errno failed - reason indicated
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
static int
|
2003-10-23 13:49:10 +00:00
|
|
|
if_rtdel(struct radix_node *rn, void *arg)
|
1999-12-17 06:46:07 +00:00
|
|
|
{
|
|
|
|
struct rtentry *rt = (struct rtentry *)rn;
|
|
|
|
struct ifnet *ifp = arg;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (rt->rt_ifp == ifp) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Protect (sorta) against walktree recursion problems
|
|
|
|
* with cloned routes
|
|
|
|
*/
|
|
|
|
if ((rt->rt_flags & RTF_UP) == 0)
|
|
|
|
return (0);
|
|
|
|
|
Add code to allow the system to handle multiple routing tables.
This particular implementation is designed to be fully backwards compatible
and to be MFC-able to 7.x (and 6.x)
Currently the only protocol that can make use of the multiple tables is IPv4
Similar functionality exists in OpenBSD and Linux.
From my notes:
-----
One thing where FreeBSD has been falling behind, and which by chance I
have some time to work on is "policy based routing", which allows
different
packet streams to be routed by more than just the destination address.
Constraints:
------------
I want to make some form of this available in the 6.x tree
(and by extension 7.x) , but FreeBSD in general needs it so I might as
well do it in -current and back port the portions I need.
One of the ways that this can be done is to have the ability to
instantiate multiple kernel routing tables (which I will now
refer to as "Forwarding Information Bases" or "FIBs" for political
correctness reasons). Which FIB a particular packet uses to make
the next hop decision can be decided by a number of mechanisms.
The policies these mechanisms implement are the "Policies" referred
to in "Policy based routing".
One of the constraints I have if I try to back port this work to
6.x is that it must be implemented as a EXTENSION to the existing
ABIs in 6.x so that third party applications do not need to be
recompiled in timespan of the branch.
This first version will not have some of the bells and whistles that
will come with later versions. It will, for example, be limited to 16
tables in the first commit.
Implementation method, Compatible version. (part 1)
-------------------------------
For this reason I have implemented a "sufficient subset" of a
multiple routing table solution in Perforce, and back-ported it
to 6.x. (also in Perforce though not always caught up with what I
have done in -current/P4). The subset allows a number of FIBs
to be defined at compile time (8 is sufficient for my purposes in 6.x)
and implements the changes needed to allow IPV4 to use them. I have not
done the changes for ipv6 simply because I do not need it, and I do not
have enough knowledge of ipv6 (e.g. neighbor discovery) needed to do it.
Other protocol families are left untouched and should there be
users with proprietary protocol families, they should continue to work
and be oblivious to the existence of the extra FIBs.
To understand how this is done, one must know that the current FIB
code starts everything off with a single dimensional array of
pointers to FIB head structures (One per protocol family), each of
which in turn points to the trie of routes available to that family.
The basic change in the ABI compatible version of the change is to
extent that array to be a 2 dimensional array, so that
instead of protocol family X looking at rt_tables[X] for the
table it needs, it looks at rt_tables[Y][X] when for all
protocol families except ipv4 Y is always 0.
Code that is unaware of the change always just sees the first row
of the table, which of course looks just like the one dimensional
array that existed before.
The entry points rtrequest(), rtalloc(), rtalloc1(), rtalloc_ign()
are all maintained, but refer only to the first row of the array,
so that existing callers in proprietary protocols can continue to
do the "right thing".
Some new entry points are added, for the exclusive use of ipv4 code
called in_rtrequest(), in_rtalloc(), in_rtalloc1() and in_rtalloc_ign(),
which have an extra argument which refers the code to the correct row.
In addition, there are some new entry points (currently called
rtalloc_fib() and friends) that check the Address family being
looked up and call either rtalloc() (and friends) if the protocol
is not IPv4 forcing the action to row 0 or to the appropriate row
if it IS IPv4 (and that info is available). These are for calling
from code that is not specific to any particular protocol. The way
these are implemented would change in the non ABI preserving code
to be added later.
One feature of the first version of the code is that for ipv4,
the interface routes show up automatically on all the FIBs, so
that no matter what FIB you select you always have the basic
direct attached hosts available to you. (rtinit() does this
automatically).
You CAN delete an interface route from one FIB should you want
to but by default it's there. ARP information is also available
in each FIB. It's assumed that the same machine would have the
same MAC address, regardless of which FIB you are using to get
to it.
This brings us as to how the correct FIB is selected for an outgoing
IPV4 packet.
Firstly, all packets have a FIB associated with them. if nothing
has been done to change it, it will be FIB 0. The FIB is changed
in the following ways.
Packets fall into one of a number of classes.
1/ locally generated packets, coming from a socket/PCB.
Such packets select a FIB from a number associated with the
socket/PCB. This in turn is inherited from the process,
but can be changed by a socket option. The process in turn
inherits it on fork. I have written a utility call setfib
that acts a bit like nice..
setfib -3 ping target.example.com # will use fib 3 for ping.
It is an obvious extension to make it a property of a jail
but I have not done so. It can be achieved by combining the setfib and
jail commands.
2/ packets received on an interface for forwarding.
By default these packets would use table 0,
(or possibly a number settable in a sysctl(not yet)).
but prior to routing the firewall can inspect them (see below).
(possibly in the future you may be able to associate a FIB
with packets received on an interface.. An ifconfig arg, but not yet.)
3/ packets inspected by a packet classifier, which can arbitrarily
associate a fib with it on a packet by packet basis.
A fib assigned to a packet by a packet classifier
(such as ipfw) would over-ride a fib associated by
a more default source. (such as cases 1 or 2).
4/ a tcp listen socket associated with a fib will generate
accept sockets that are associated with that same fib.
5/ Packets generated in response to some other packet (e.g. reset
or icmp packets). These should use the FIB associated with the
packet being reponded to.
6/ Packets generated during encapsulation.
gif, tun and other tunnel interfaces will encapsulate using the FIB
that was in effect withthe proces that set up the tunnel.
thus setfib 1 ifconfig gif0 [tunnel instructions]
will set the fib for the tunnel to use to be fib 1.
Routing messages would be associated with their
process, and thus select one FIB or another.
messages from the kernel would be associated with the fib they
refer to and would only be received by a routing socket associated
with that fib. (not yet implemented)
In addition Netstat has been edited to be able to cope with the
fact that the array is now 2 dimensional. (It looks in system
memory using libkvm (!)). Old versions of netstat see only the first FIB.
In addition two sysctls are added to give:
a) the number of FIBs compiled in (active)
b) the default FIB of the calling process.
Early testing experience:
-------------------------
Basically our (IronPort's) appliance does this functionality already
using ipfw fwd but that method has some drawbacks.
For example,
It can't fully simulate a routing table because it can't influence the
socket's choice of local address when a connect() is done.
Testing during the generating of these changes has been
remarkably smooth so far. Multiple tables have co-existed
with no notable side effects, and packets have been routes
accordingly.
ipfw has grown 2 new keywords:
setfib N ip from anay to any
count ip from any to any fib N
In pf there seems to be a requirement to be able to give symbolic names to the
fibs but I do not have that capacity. I am not sure if it is required.
SCTP has interestingly enough built in support for this, called VRFs
in Cisco parlance. it will be interesting to see how that handles it
when it suddenly actually does something.
Where to next:
--------------------
After committing the ABI compatible version and MFCing it, I'd
like to proceed in a forward direction in -current. this will
result in some roto-tilling in the routing code.
Firstly: the current code's idea of having a separate tree per
protocol family, all of the same format, and pointed to by the
1 dimensional array is a bit silly. Especially when one considers that
there is code that makes assumptions about every protocol having the
same internal structures there. Some protocols don't WANT that
sort of structure. (for example the whole idea of a netmask is foreign
to appletalk). This needs to be made opaque to the external code.
My suggested first change is to add routing method pointers to the
'domain' structure, along with information pointing the data.
instead of having an array of pointers to uniform structures,
there would be an array pointing to the 'domain' structures
for each protocol address domain (protocol family),
and the methods this reached would be called. The methods would have
an argument that gives FIB number, but the protocol would be free
to ignore it.
When the ABI can be changed it raises the possibilty of the
addition of a fib entry into the "struct route". Currently,
the structure contains the sockaddr of the desination, and the resulting
fib entry. To make this work fully, one could add a fib number
so that given an address and a fib, one can find the third element, the
fib entry.
Interaction with the ARP layer/ LL layer would need to be
revisited as well. Qing Li has been working on this already.
This work was sponsored by Ironport Systems/Cisco
Reviewed by: several including rwatson, bz and mlair (parts each)
Obtained from: Ironport systems/Cisco
2008-05-09 23:03:00 +00:00
|
|
|
err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
|
2008-12-18 09:59:24 +00:00
|
|
|
rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
|
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
|
|
|
(struct rtentry **) NULL, rt->rt_fibnum);
|
1999-12-17 06:46:07 +00:00
|
|
|
if (err) {
|
|
|
|
log(LOG_WARNING, "if_rtdel: error %d\n", err);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return (0);
|
1999-04-16 21:22:55 +00:00
|
|
|
}
|
|
|
|
|
2009-06-26 00:36:47 +00:00
|
|
|
/*
|
|
|
|
* Wrapper functions for struct ifnet address list locking macros. These are
|
|
|
|
* used by kernel modules to avoid encoding programming interface or binary
|
|
|
|
* interface assumptions that may be violated when kernel-internal locking
|
|
|
|
* approaches change.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_addr_rlock(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
if_addr_runlock(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
if_maddr_rlock(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
if_maddr_runlock(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
|
|
|
|
2009-06-21 19:30:33 +00:00
|
|
|
/*
|
|
|
|
* Reference count functions for ifaddrs.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
ifa_init(struct ifaddr *ifa)
|
|
|
|
{
|
|
|
|
|
|
|
|
mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
|
|
|
|
refcount_init(&ifa->ifa_refcnt, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
ifa_ref(struct ifaddr *ifa)
|
|
|
|
{
|
|
|
|
|
|
|
|
refcount_acquire(&ifa->ifa_refcnt);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
ifa_free(struct ifaddr *ifa)
|
|
|
|
{
|
|
|
|
|
|
|
|
if (refcount_release(&ifa->ifa_refcnt)) {
|
|
|
|
mtx_destroy(&ifa->ifa_mtx);
|
|
|
|
free(ifa, M_IFADDR);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-09-15 19:18:34 +00:00
|
|
|
int
|
|
|
|
ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
struct rtentry *rt = NULL;
|
|
|
|
struct rt_addrinfo info;
|
|
|
|
static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
|
|
|
|
|
|
|
|
bzero(&info, sizeof(info));
|
|
|
|
info.rti_ifp = V_loif;
|
|
|
|
info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
|
|
|
|
info.rti_info[RTAX_DST] = ia;
|
|
|
|
info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
|
|
|
|
error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
|
|
|
|
|
|
|
|
if (error == 0 && rt != NULL) {
|
|
|
|
RT_LOCK(rt);
|
|
|
|
((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
|
2009-09-20 17:22:19 +00:00
|
|
|
ifa->ifa_ifp->if_type;
|
2009-09-15 19:18:34 +00:00
|
|
|
((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
|
2009-09-20 17:22:19 +00:00
|
|
|
ifa->ifa_ifp->if_index;
|
2009-09-15 19:18:34 +00:00
|
|
|
RT_REMREF(rt);
|
|
|
|
RT_UNLOCK(rt);
|
|
|
|
} else if (error != 0)
|
|
|
|
log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
struct rt_addrinfo info;
|
|
|
|
struct sockaddr_dl null_sdl;
|
|
|
|
|
|
|
|
bzero(&null_sdl, sizeof(null_sdl));
|
|
|
|
null_sdl.sdl_len = sizeof(null_sdl);
|
|
|
|
null_sdl.sdl_family = AF_LINK;
|
|
|
|
null_sdl.sdl_type = ifa->ifa_ifp->if_type;
|
|
|
|
null_sdl.sdl_index = ifa->ifa_ifp->if_index;
|
|
|
|
bzero(&info, sizeof(info));
|
|
|
|
info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
|
|
|
|
info.rti_info[RTAX_DST] = ia;
|
|
|
|
info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
|
|
|
|
error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
|
|
|
|
|
|
|
|
if (error != 0)
|
|
|
|
log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
|
|
|
|
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
2007-02-22 00:14:02 +00:00
|
|
|
/*
|
|
|
|
* XXX: Because sockaddr_dl has deeper structure than the sockaddr
|
|
|
|
* structs used to represent other address families, it is necessary
|
|
|
|
* to perform a different comparison.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#define sa_equal(a1, a2) \
|
|
|
|
(bcmp((a1), (a2), ((a1))->sa_len) == 0)
|
|
|
|
|
|
|
|
#define sa_dl_equal(a1, a2) \
|
|
|
|
((((struct sockaddr_dl *)(a1))->sdl_len == \
|
|
|
|
((struct sockaddr_dl *)(a2))->sdl_len) && \
|
|
|
|
(bcmp(LLADDR((struct sockaddr_dl *)(a1)), \
|
|
|
|
LLADDR((struct sockaddr_dl *)(a2)), \
|
|
|
|
((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
|
2002-12-18 11:46:59 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Locate an interface based on a complete address.
|
|
|
|
*/
|
|
|
|
/*ARGSUSED*/
|
2009-06-22 10:59:34 +00:00
|
|
|
static struct ifaddr *
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-06 00:44:45 +00:00
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifaddr *ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
2009-04-21 19:06:47 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
|
|
|
IF_ADDR_LOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
|
|
|
if (ifa->ifa_addr->sa_family != addr->sa_family)
|
|
|
|
continue;
|
2009-04-21 19:06:47 +00:00
|
|
|
if (sa_equal(addr, ifa->ifa_addr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
if (getref)
|
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
/* IP6 doesn't have broadcast */
|
|
|
|
if ((ifp->if_flags & IFF_BROADCAST) &&
|
|
|
|
ifa->ifa_broadaddr &&
|
|
|
|
ifa->ifa_broadaddr->sa_len != 0 &&
|
2009-04-21 19:06:47 +00:00
|
|
|
sa_equal(ifa->ifa_broadaddr, addr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
if (getref)
|
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
}
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
ifa = NULL;
|
|
|
|
done:
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2001-09-06 00:44:45 +00:00
|
|
|
return (ifa);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
|
2009-06-22 10:59:34 +00:00
|
|
|
struct ifaddr *
|
|
|
|
ifa_ifwithaddr(struct sockaddr *addr)
|
|
|
|
{
|
|
|
|
|
2009-06-23 20:19:09 +00:00
|
|
|
return (ifa_ifwithaddr_internal(addr, 1));
|
2009-06-22 10:59:34 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
ifa_ifwithaddr_check(struct sockaddr *addr)
|
|
|
|
{
|
|
|
|
|
2009-06-23 20:19:09 +00:00
|
|
|
return (ifa_ifwithaddr_internal(addr, 0) != NULL);
|
2009-06-22 10:59:34 +00:00
|
|
|
}
|
|
|
|
|
2006-09-06 17:12:10 +00:00
|
|
|
/*
|
|
|
|
* Locate an interface based on the broadcast address.
|
|
|
|
*/
|
|
|
|
/* ARGSUSED */
|
|
|
|
struct ifaddr *
|
|
|
|
ifa_ifwithbroadaddr(struct sockaddr *addr)
|
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifaddr *ifa;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
2009-04-21 19:06:47 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
|
|
|
IF_ADDR_LOCK(ifp);
|
2006-09-06 17:12:10 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
|
|
|
if (ifa->ifa_addr->sa_family != addr->sa_family)
|
|
|
|
continue;
|
|
|
|
if ((ifp->if_flags & IFF_BROADCAST) &&
|
|
|
|
ifa->ifa_broadaddr &&
|
|
|
|
ifa->ifa_broadaddr->sa_len != 0 &&
|
2009-04-21 19:06:47 +00:00
|
|
|
sa_equal(ifa->ifa_broadaddr, addr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2006-09-06 17:12:10 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
2006-09-06 17:12:10 +00:00
|
|
|
}
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
2006-09-06 17:12:10 +00:00
|
|
|
ifa = NULL;
|
|
|
|
done:
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2006-09-06 17:12:10 +00:00
|
|
|
return (ifa);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Locate the point to point interface with a given destination address.
|
|
|
|
*/
|
|
|
|
/*ARGSUSED*/
|
|
|
|
struct ifaddr *
|
2003-10-23 13:49:10 +00:00
|
|
|
ifa_ifwithdstaddr(struct sockaddr *addr)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-06 00:44:45 +00:00
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifaddr *ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
2001-09-06 00:44:45 +00:00
|
|
|
if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
|
|
|
|
continue;
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2001-02-04 16:08:18 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ifa->ifa_addr->sa_family != addr->sa_family)
|
|
|
|
continue;
|
2008-08-24 11:03:43 +00:00
|
|
|
if (ifa->ifa_dstaddr != NULL &&
|
2009-04-21 19:06:47 +00:00
|
|
|
sa_equal(addr, ifa->ifa_dstaddr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
}
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
ifa = NULL;
|
|
|
|
done:
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2001-09-06 00:44:45 +00:00
|
|
|
return (ifa);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find an interface on a specific network. If many, choice
|
|
|
|
* is most specific found.
|
|
|
|
*/
|
|
|
|
struct ifaddr *
|
2010-05-25 20:42:35 +00:00
|
|
|
ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifaddr *ifa;
|
2009-06-23 20:19:09 +00:00
|
|
|
struct ifaddr *ifa_maybe = NULL;
|
1994-05-24 10:09:53 +00:00
|
|
|
u_int af = addr->sa_family;
|
|
|
|
char *addr_data = addr->sa_data, *cplim;
|
|
|
|
|
1997-08-22 22:47:27 +00:00
|
|
|
/*
|
|
|
|
* AF_LINK addresses can be looked up directly by their index number,
|
|
|
|
* so do that if we can.
|
|
|
|
*/
|
1994-05-24 10:09:53 +00:00
|
|
|
if (af == AF_LINK) {
|
2003-10-04 03:44:50 +00:00
|
|
|
struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
|
2001-09-06 02:40:43 +00:00
|
|
|
return (ifaddr_byindex(sdl->sdl_index));
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
1997-08-22 22:47:27 +00:00
|
|
|
|
1999-11-22 02:45:11 +00:00
|
|
|
/*
|
2009-06-23 20:19:09 +00:00
|
|
|
* Scan though each interface, looking for ones that have addresses
|
|
|
|
* in this address family. Maintain a reference on ifa_maybe once
|
|
|
|
* we find one, as we release the IF_ADDR_LOCK() that kept it stable
|
|
|
|
* when we move onto the next interface.
|
1997-08-22 22:47:27 +00:00
|
|
|
*/
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2001-02-04 16:08:18 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
2003-10-23 13:49:10 +00:00
|
|
|
char *cp, *cp2, *cp3;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
1995-06-28 05:31:03 +00:00
|
|
|
if (ifa->ifa_addr->sa_family != af)
|
1997-08-22 22:47:27 +00:00
|
|
|
next: continue;
|
2010-05-25 20:42:35 +00:00
|
|
|
if (af == AF_INET &&
|
|
|
|
ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
|
1997-08-22 22:47:27 +00:00
|
|
|
/*
|
1999-11-22 02:45:11 +00:00
|
|
|
* This is a bit broken as it doesn't
|
|
|
|
* take into account that the remote end may
|
1997-08-22 22:47:27 +00:00
|
|
|
* be a single node in the network we are
|
|
|
|
* looking for.
|
1999-11-22 02:45:11 +00:00
|
|
|
* The trouble is that we don't know the
|
1997-08-22 22:47:27 +00:00
|
|
|
* netmask for the remote end.
|
|
|
|
*/
|
2008-08-24 11:03:43 +00:00
|
|
|
if (ifa->ifa_dstaddr != NULL &&
|
2009-04-21 19:06:47 +00:00
|
|
|
sa_equal(addr, ifa->ifa_dstaddr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
1995-06-15 00:19:56 +00:00
|
|
|
} else {
|
1997-08-28 01:17:12 +00:00
|
|
|
/*
|
|
|
|
* if we have a special address handler,
|
|
|
|
* then use it instead of the generic one.
|
|
|
|
*/
|
2003-10-23 13:49:10 +00:00
|
|
|
if (ifa->ifa_claim_addr) {
|
2009-04-21 19:06:47 +00:00
|
|
|
if ((*ifa->ifa_claim_addr)(ifa, addr)) {
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 00:44:45 +00:00
|
|
|
goto done;
|
2009-04-21 19:06:47 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
continue;
|
1997-08-28 01:17:12 +00:00
|
|
|
}
|
|
|
|
|
1997-08-22 22:47:27 +00:00
|
|
|
/*
|
|
|
|
* Scan all the bits in the ifa's address.
|
|
|
|
* If a bit dissagrees with what we are
|
|
|
|
* looking for, mask it with the netmask
|
|
|
|
* to see if it really matters.
|
|
|
|
* (A byte at a time)
|
|
|
|
*/
|
1995-06-28 05:31:03 +00:00
|
|
|
if (ifa->ifa_netmask == 0)
|
|
|
|
continue;
|
1995-05-27 04:37:24 +00:00
|
|
|
cp = addr_data;
|
|
|
|
cp2 = ifa->ifa_addr->sa_data;
|
|
|
|
cp3 = ifa->ifa_netmask->sa_data;
|
1997-08-22 22:47:27 +00:00
|
|
|
cplim = ifa->ifa_netmask->sa_len
|
|
|
|
+ (char *)ifa->ifa_netmask;
|
1995-05-27 04:37:24 +00:00
|
|
|
while (cp3 < cplim)
|
|
|
|
if ((*cp++ ^ *cp2++) & *cp3++)
|
1997-08-22 22:47:27 +00:00
|
|
|
goto next; /* next address! */
|
|
|
|
/*
|
|
|
|
* If the netmask of what we just found
|
|
|
|
* is more specific than what we had before
|
|
|
|
* (if we had one) then remember the new one
|
|
|
|
* before continuing to search
|
|
|
|
* for an even better one.
|
|
|
|
*/
|
2009-06-23 20:19:09 +00:00
|
|
|
if (ifa_maybe == NULL ||
|
1995-05-27 04:37:24 +00:00
|
|
|
rn_refines((caddr_t)ifa->ifa_netmask,
|
2009-06-23 20:19:09 +00:00
|
|
|
(caddr_t)ifa_maybe->ifa_netmask)) {
|
|
|
|
if (ifa_maybe != NULL)
|
|
|
|
ifa_free(ifa_maybe);
|
1995-05-27 04:37:24 +00:00
|
|
|
ifa_maybe = ifa;
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_ref(ifa_maybe);
|
|
|
|
}
|
1995-05-27 04:37:24 +00:00
|
|
|
}
|
|
|
|
}
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
1995-05-27 04:37:24 +00:00
|
|
|
}
|
2001-09-06 00:44:45 +00:00
|
|
|
ifa = ifa_maybe;
|
2009-06-23 20:19:09 +00:00
|
|
|
ifa_maybe = NULL;
|
2001-09-06 00:44:45 +00:00
|
|
|
done:
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2009-06-23 20:19:09 +00:00
|
|
|
if (ifa_maybe != NULL)
|
|
|
|
ifa_free(ifa_maybe);
|
2001-09-06 00:44:45 +00:00
|
|
|
return (ifa);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find an interface address specific to an interface best matching
|
|
|
|
* a given address.
|
|
|
|
*/
|
|
|
|
struct ifaddr *
|
2003-10-23 13:49:10 +00:00
|
|
|
ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct ifaddr *ifa;
|
|
|
|
char *cp, *cp2, *cp3;
|
|
|
|
char *cplim;
|
2009-06-23 20:19:09 +00:00
|
|
|
struct ifaddr *ifa_maybe = NULL;
|
1994-05-24 10:09:53 +00:00
|
|
|
u_int af = addr->sa_family;
|
|
|
|
|
|
|
|
if (af >= AF_MAX)
|
2010-07-27 11:54:01 +00:00
|
|
|
return (NULL);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2001-02-04 16:08:18 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ifa->ifa_addr->sa_family != af)
|
|
|
|
continue;
|
2009-06-23 20:19:09 +00:00
|
|
|
if (ifa_maybe == NULL)
|
1996-08-07 04:09:05 +00:00
|
|
|
ifa_maybe = ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ifa->ifa_netmask == 0) {
|
2005-07-19 10:03:47 +00:00
|
|
|
if (sa_equal(addr, ifa->ifa_addr) ||
|
|
|
|
(ifa->ifa_dstaddr &&
|
|
|
|
sa_equal(addr, ifa->ifa_dstaddr)))
|
2001-09-07 05:32:54 +00:00
|
|
|
goto done;
|
1994-05-24 10:09:53 +00:00
|
|
|
continue;
|
|
|
|
}
|
1995-05-27 04:37:24 +00:00
|
|
|
if (ifp->if_flags & IFF_POINTOPOINT) {
|
2005-07-19 10:03:47 +00:00
|
|
|
if (sa_equal(addr, ifa->ifa_dstaddr))
|
2001-09-07 05:39:47 +00:00
|
|
|
goto done;
|
1995-06-15 00:19:56 +00:00
|
|
|
} else {
|
1995-05-27 04:37:24 +00:00
|
|
|
cp = addr->sa_data;
|
|
|
|
cp2 = ifa->ifa_addr->sa_data;
|
|
|
|
cp3 = ifa->ifa_netmask->sa_data;
|
|
|
|
cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
|
|
|
|
for (; cp3 < cplim; cp3++)
|
|
|
|
if ((*cp++ ^ *cp2++) & *cp3)
|
|
|
|
break;
|
|
|
|
if (cp3 == cplim)
|
2001-09-07 05:32:54 +00:00
|
|
|
goto done;
|
1995-05-27 04:37:24 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2001-09-06 02:40:43 +00:00
|
|
|
ifa = ifa_maybe;
|
|
|
|
done:
|
2009-06-23 20:19:09 +00:00
|
|
|
if (ifa != NULL)
|
|
|
|
ifa_ref(ifa);
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2001-09-06 02:40:43 +00:00
|
|
|
return (ifa);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
This main goals of this project are:
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,
The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.
Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:
- Kip Macy revised the locking code completely, thus completing
the last piece of the puzzle, Kip has also been conducting
active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
me maintaining that branch before the svn conversion
2008-12-15 06:10:57 +00:00
|
|
|
#include <net/if_llatbl.h>
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Default action when installing a route with a Link Level gateway.
|
|
|
|
* Lookup an appropriate real ifa to point to.
|
|
|
|
* This should be moved to /sys/net/link.c eventually.
|
|
|
|
*/
|
1995-12-09 20:47:15 +00:00
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct ifaddr *ifa, *oifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
struct sockaddr *dst;
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
2003-10-04 03:44:50 +00:00
|
|
|
RT_LOCK_ASSERT(rt);
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
|
|
|
|
((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
|
|
|
|
return;
|
1994-10-08 01:40:23 +00:00
|
|
|
ifa = ifaof_ifpforaddr(dst, ifp);
|
|
|
|
if (ifa) {
|
2003-10-04 03:44:50 +00:00
|
|
|
oifa = rt->rt_ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
rt->rt_ifa = ifa;
|
2009-06-21 19:30:33 +00:00
|
|
|
ifa_free(oifa);
|
1994-05-24 10:09:53 +00:00
|
|
|
if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
|
2001-10-17 18:07:05 +00:00
|
|
|
ifa->ifa_rtrequest(cmd, rt, info);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark an interface down and notify protocols of
|
|
|
|
* the transition.
|
|
|
|
* NOTE: must be called at splnet or eqivalent.
|
|
|
|
*/
|
2004-04-18 18:59:44 +00:00
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_unroute(struct ifnet *ifp, int flag, int fam)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct ifaddr *ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
|
|
|
|
|
1998-12-16 18:30:43 +00:00
|
|
|
ifp->if_flags &= ~flag;
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
1998-12-16 18:30:43 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
|
|
|
|
if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
|
|
|
|
pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
|
2008-11-22 05:55:56 +00:00
|
|
|
ifp->if_qflush(ifp);
|
2009-04-16 23:05:10 +00:00
|
|
|
|
2005-02-22 13:04:05 +00:00
|
|
|
if (ifp->if_carp)
|
2010-08-11 00:51:50 +00:00
|
|
|
(*carp_linkstate_p)(ifp);
|
1994-05-24 10:09:53 +00:00
|
|
|
rt_ifmsg(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark an interface up and notify protocols of
|
|
|
|
* the transition.
|
|
|
|
* NOTE: must be called at splnet or eqivalent.
|
|
|
|
*/
|
2004-04-18 18:59:44 +00:00
|
|
|
static void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_route(struct ifnet *ifp, int flag, int fam)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct ifaddr *ifa;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
|
|
|
|
|
1998-12-16 18:30:43 +00:00
|
|
|
ifp->if_flags |= flag;
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
1998-12-16 18:30:43 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
|
|
|
|
if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
|
|
|
|
pfctlinput(PRC_IFUP, ifa->ifa_addr);
|
2005-02-22 13:04:05 +00:00
|
|
|
if (ifp->if_carp)
|
2010-08-11 00:51:50 +00:00
|
|
|
(*carp_linkstate_p)(ifp);
|
1994-05-24 10:09:53 +00:00
|
|
|
rt_ifmsg(ifp);
|
1999-11-22 02:45:11 +00:00
|
|
|
#ifdef INET6
|
|
|
|
in6_if_up(ifp);
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
2009-12-31 20:29:58 +00:00
|
|
|
void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
|
Merge the //depot/user/yar/vlan branch into CVS. It contains some collective
work by yar, thompsa and myself. The checksum offloading part also involves
work done by Mihail Balikov.
The most important changes:
o Instead of global linked list of all vlan softc use a per-trunk
hash. The size of hash is dynamically adjusted, depending on
number of entries. This changes struct ifnet, replacing counter
of vlans with a pointer to trunk structure. This change is an
improvement for setups with big number of VLANs, several interfaces
and several CPUs. It is a small regression for a setup with a single
VLAN interface.
An alternative to dynamic hash is a per-trunk static array with
4096 entries, which is a compile time option - VLAN_ARRAY. In my
experiments the array is not an improvement, probably because such
a big trunk structure doesn't fit into CPU cache.
o Introduce an UMA zone for VLAN tags. Since drivers depend on it,
the zone is declared in kern_mbuf.c, not in optional vlan(4) driver.
This change is a big improvement for any setup utilizing vlan(4).
o Use rwlock(9) instead of mutex(9) for locking. We are the first
ones to do this! :)
o Some drivers can do hardware VLAN tagging + hardware checksum
offloading. Add an infrastructure for this. Whenever vlan(4) is
attached to a parent or parent configuration is changed, the flags
on vlan(4) interface are updated.
In collaboration with: yar, thompsa
In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:15 +00:00
|
|
|
void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
|
2011-03-21 09:40:01 +00:00
|
|
|
struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
|
|
|
|
struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
|
|
|
|
int (*vlan_tag_p)(struct ifnet *, uint16_t *);
|
|
|
|
int (*vlan_setcookie_p)(struct ifnet *, void *);
|
|
|
|
void *(*vlan_cookie_p)(struct ifnet *);
|
2004-12-08 05:45:59 +00:00
|
|
|
|
|
|
|
/*
|
2005-04-20 09:30:54 +00:00
|
|
|
* Handle a change in the interface link state. To avoid LORs
|
|
|
|
* between driver lock and upper layer locks, as well as possible
|
|
|
|
* recursions, we post event to taskqueue, and all job
|
|
|
|
* is done in static do_link_state_change().
|
2004-12-08 05:45:59 +00:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_link_state_change(struct ifnet *ifp, int link_state)
|
|
|
|
{
|
2005-02-22 14:21:59 +00:00
|
|
|
/* Return if state hasn't changed. */
|
|
|
|
if (ifp->if_link_state == link_state)
|
|
|
|
return;
|
|
|
|
|
|
|
|
ifp->if_link_state = link_state;
|
|
|
|
|
2005-04-20 09:30:54 +00:00
|
|
|
taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
do_link_state_change(void *arg, int pending)
|
|
|
|
{
|
|
|
|
struct ifnet *ifp = (struct ifnet *)arg;
|
|
|
|
int link_state = ifp->if_link_state;
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
CURVNET_SET(ifp->if_vnet);
|
2005-04-20 09:30:54 +00:00
|
|
|
|
2004-12-08 05:45:59 +00:00
|
|
|
/* Notify that the link state has changed. */
|
2005-02-22 14:21:59 +00:00
|
|
|
rt_ifmsg(ifp);
|
Merge the //depot/user/yar/vlan branch into CVS. It contains some collective
work by yar, thompsa and myself. The checksum offloading part also involves
work done by Mihail Balikov.
The most important changes:
o Instead of global linked list of all vlan softc use a per-trunk
hash. The size of hash is dynamically adjusted, depending on
number of entries. This changes struct ifnet, replacing counter
of vlans with a pointer to trunk structure. This change is an
improvement for setups with big number of VLANs, several interfaces
and several CPUs. It is a small regression for a setup with a single
VLAN interface.
An alternative to dynamic hash is a per-trunk static array with
4096 entries, which is a compile time option - VLAN_ARRAY. In my
experiments the array is not an improvement, probably because such
a big trunk structure doesn't fit into CPU cache.
o Introduce an UMA zone for VLAN tags. Since drivers depend on it,
the zone is declared in kern_mbuf.c, not in optional vlan(4) driver.
This change is a big improvement for any setup utilizing vlan(4).
o Use rwlock(9) instead of mutex(9) for locking. We are the first
ones to do this! :)
o Some drivers can do hardware VLAN tagging + hardware checksum
offloading. Add an infrastructure for this. Whenever vlan(4) is
attached to a parent or parent configuration is changed, the flags
on vlan(4) interface are updated.
In collaboration with: yar, thompsa
In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:15 +00:00
|
|
|
if (ifp->if_vlantrunk != NULL)
|
2009-12-31 20:29:58 +00:00
|
|
|
(*vlan_link_state_p)(ifp);
|
2005-02-22 14:21:59 +00:00
|
|
|
|
|
|
|
if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
|
|
|
|
IFP2AC(ifp)->ac_netgraph != NULL)
|
|
|
|
(*ng_ether_link_state_p)(ifp, link_state);
|
|
|
|
if (ifp->if_carp)
|
2010-08-11 00:51:50 +00:00
|
|
|
(*carp_linkstate_p)(ifp);
|
2005-06-05 03:13:13 +00:00
|
|
|
if (ifp->if_bridge) {
|
|
|
|
KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
|
|
|
|
(*bstp_linkstate_p)(ifp, link_state);
|
|
|
|
}
|
2007-04-17 00:35:11 +00:00
|
|
|
if (ifp->if_lagg) {
|
|
|
|
KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
|
|
|
|
(*lagg_linkstate_p)(ifp, link_state);
|
2007-04-10 00:27:25 +00:00
|
|
|
}
|
2005-06-05 03:13:13 +00:00
|
|
|
|
Change the curvnet variable from a global const struct vnet *,
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
|
|
|
if (IS_DEFAULT_VNET(curvnet))
|
|
|
|
devctl_notify("IFNET", ifp->if_xname,
|
|
|
|
(link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
|
|
|
|
NULL);
|
2005-04-20 09:30:54 +00:00
|
|
|
if (pending > 1)
|
|
|
|
if_printf(ifp, "%d link states coalesced\n", pending);
|
2005-03-12 12:58:03 +00:00
|
|
|
if (log_link_state_change)
|
|
|
|
log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
|
|
|
|
(link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
|
2011-03-21 09:40:01 +00:00
|
|
|
EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
|
Step 1.5 of importing the network stack virtualization infrastructure
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
2008-10-02 15:37:58 +00:00
|
|
|
CURVNET_RESTORE();
|
2004-12-08 05:45:59 +00:00
|
|
|
}
|
|
|
|
|
1998-12-16 18:30:43 +00:00
|
|
|
/*
|
|
|
|
* Mark an interface down and notify protocols of
|
|
|
|
* the transition.
|
|
|
|
* NOTE: must be called at splnet or eqivalent.
|
|
|
|
*/
|
|
|
|
void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_down(struct ifnet *ifp)
|
1998-12-16 18:30:43 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
if_unroute(ifp, IFF_UP, AF_UNSPEC);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark an interface up and notify protocols of
|
|
|
|
* the transition.
|
|
|
|
* NOTE: must be called at splnet or eqivalent.
|
|
|
|
*/
|
|
|
|
void
|
2003-10-23 13:49:10 +00:00
|
|
|
if_up(struct ifnet *ifp)
|
1998-12-16 18:30:43 +00:00
|
|
|
{
|
|
|
|
|
|
|
|
if_route(ifp, IFF_UP, AF_UNSPEC);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Flush an interface queue.
|
|
|
|
*/
|
2009-04-16 23:05:10 +00:00
|
|
|
void
|
2008-11-22 05:55:56 +00:00
|
|
|
if_qflush(struct ifnet *ifp)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2003-10-23 13:49:10 +00:00
|
|
|
struct mbuf *m, *n;
|
2008-11-22 05:55:56 +00:00
|
|
|
struct ifaltq *ifq;
|
|
|
|
|
|
|
|
ifq = &ifp->if_snd;
|
2004-09-01 19:56:47 +00:00
|
|
|
IFQ_LOCK(ifq);
|
2004-06-13 17:29:10 +00:00
|
|
|
#ifdef ALTQ
|
|
|
|
if (ALTQ_IS_ENABLED(ifq))
|
|
|
|
ALTQ_PURGE(ifq);
|
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
n = ifq->ifq_head;
|
1994-10-08 01:40:23 +00:00
|
|
|
while ((m = n) != 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
n = m->m_act;
|
|
|
|
m_freem(m);
|
|
|
|
}
|
|
|
|
ifq->ifq_head = 0;
|
|
|
|
ifq->ifq_tail = 0;
|
|
|
|
ifq->ifq_len = 0;
|
2004-09-01 19:56:47 +00:00
|
|
|
IFQ_UNLOCK(ifq);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2009-04-23 13:08:47 +00:00
|
|
|
* Map interface name to interface structure pointer, with or without
|
|
|
|
* returning a reference.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2009-04-23 13:08:47 +00:00
|
|
|
struct ifnet *
|
|
|
|
ifunit_ref(const char *name)
|
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
2009-04-23 13:08:47 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
2009-04-23 15:56:01 +00:00
|
|
|
if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
|
|
|
|
!(ifp->if_flags & IFF_DYING))
|
2009-04-23 13:08:47 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (ifp != NULL)
|
|
|
|
if_ref(ifp);
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2009-04-23 13:08:47 +00:00
|
|
|
return (ifp);
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
struct ifnet *
|
2001-07-02 20:49:25 +00:00
|
|
|
ifunit(const char *name)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
1999-12-13 15:57:11 +00:00
|
|
|
struct ifnet *ifp;
|
2001-10-17 18:58:14 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
2004-02-04 02:54:25 +00:00
|
|
|
if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
|
1994-05-24 10:09:53 +00:00
|
|
|
break;
|
|
|
|
}
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
1994-05-24 10:09:53 +00:00
|
|
|
return (ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2001-09-29 05:55:04 +00:00
|
|
|
* Hardware specific interface ioctls.
|
1994-05-24 10:09:53 +00:00
|
|
|
*/
|
2001-09-29 05:55:04 +00:00
|
|
|
static int
|
|
|
|
ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-29 05:55:04 +00:00
|
|
|
struct ifreq *ifr;
|
1999-06-19 18:42:31 +00:00
|
|
|
struct ifstat *ifs;
|
2001-09-29 05:55:04 +00:00
|
|
|
int error = 0;
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
int new_flags, temp_flags;
|
2004-02-04 02:54:25 +00:00
|
|
|
size_t namelen, onamelen;
|
2010-01-27 00:30:07 +00:00
|
|
|
size_t descrlen;
|
|
|
|
char *descrbuf, *odescrbuf;
|
2004-02-04 02:54:25 +00:00
|
|
|
char new_name[IFNAMSIZ];
|
|
|
|
struct ifaddr *ifa;
|
|
|
|
struct sockaddr_dl *sdl;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
ifr = (struct ifreq *)data;
|
|
|
|
switch (cmd) {
|
2001-10-17 19:40:44 +00:00
|
|
|
case SIOCGIFINDEX:
|
|
|
|
ifr->ifr_index = ifp->if_index;
|
|
|
|
break;
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
case SIOCGIFFLAGS:
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
temp_flags = ifp->if_flags | ifp->if_drv_flags;
|
|
|
|
ifr->ifr_flags = temp_flags & 0xffff;
|
|
|
|
ifr->ifr_flagshigh = temp_flags >> 16;
|
1994-05-24 10:09:53 +00:00
|
|
|
break;
|
|
|
|
|
2001-09-18 17:41:42 +00:00
|
|
|
case SIOCGIFCAP:
|
|
|
|
ifr->ifr_reqcap = ifp->if_capabilities;
|
|
|
|
ifr->ifr_curcap = ifp->if_capenable;
|
|
|
|
break;
|
|
|
|
|
2002-08-01 21:15:53 +00:00
|
|
|
#ifdef MAC
|
|
|
|
case SIOCGIFMAC:
|
2007-10-24 19:04:04 +00:00
|
|
|
error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
|
2002-08-01 21:15:53 +00:00
|
|
|
break;
|
|
|
|
#endif
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
case SIOCGIFMETRIC:
|
|
|
|
ifr->ifr_metric = ifp->if_metric;
|
|
|
|
break;
|
|
|
|
|
1994-08-08 10:49:26 +00:00
|
|
|
case SIOCGIFMTU:
|
|
|
|
ifr->ifr_mtu = ifp->if_mtu;
|
|
|
|
break;
|
|
|
|
|
1994-12-21 22:57:05 +00:00
|
|
|
case SIOCGIFPHYS:
|
|
|
|
ifr->ifr_phys = ifp->if_physical;
|
|
|
|
break;
|
2010-01-27 00:30:07 +00:00
|
|
|
|
|
|
|
case SIOCGIFDESCR:
|
|
|
|
error = 0;
|
|
|
|
sx_slock(&ifdescr_sx);
|
2010-04-14 22:02:19 +00:00
|
|
|
if (ifp->if_description == NULL)
|
2010-01-27 00:30:07 +00:00
|
|
|
error = ENOMSG;
|
2010-04-14 22:02:19 +00:00
|
|
|
else {
|
2010-01-27 00:30:07 +00:00
|
|
|
/* space for terminating nul */
|
|
|
|
descrlen = strlen(ifp->if_description) + 1;
|
|
|
|
if (ifr->ifr_buffer.length < descrlen)
|
2010-04-14 22:02:19 +00:00
|
|
|
ifr->ifr_buffer.buffer = NULL;
|
2010-01-27 00:30:07 +00:00
|
|
|
else
|
|
|
|
error = copyout(ifp->if_description,
|
|
|
|
ifr->ifr_buffer.buffer, descrlen);
|
|
|
|
ifr->ifr_buffer.length = descrlen;
|
|
|
|
}
|
|
|
|
sx_sunlock(&ifdescr_sx);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SIOCSIFDESCR:
|
|
|
|
error = priv_check(td, PRIV_NET_SETIFDESCR);
|
|
|
|
if (error)
|
|
|
|
return (error);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy only (length-1) bytes to make sure that
|
|
|
|
* if_description is always nul terminated. The
|
|
|
|
* length parameter is supposed to count the
|
|
|
|
* terminating nul in.
|
|
|
|
*/
|
|
|
|
if (ifr->ifr_buffer.length > ifdescr_maxlen)
|
|
|
|
return (ENAMETOOLONG);
|
|
|
|
else if (ifr->ifr_buffer.length == 0)
|
|
|
|
descrbuf = NULL;
|
|
|
|
else {
|
|
|
|
descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
|
|
|
|
M_WAITOK | M_ZERO);
|
|
|
|
error = copyin(ifr->ifr_buffer.buffer, descrbuf,
|
|
|
|
ifr->ifr_buffer.length - 1);
|
|
|
|
if (error) {
|
|
|
|
free(descrbuf, M_IFDESCR);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
sx_xlock(&ifdescr_sx);
|
|
|
|
odescrbuf = ifp->if_description;
|
|
|
|
ifp->if_description = descrbuf;
|
|
|
|
sx_xunlock(&ifdescr_sx);
|
|
|
|
|
|
|
|
getmicrotime(&ifp->if_lastchange);
|
|
|
|
free(odescrbuf, M_IFDESCR);
|
|
|
|
break;
|
1994-12-21 22:57:05 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
case SIOCSIFFLAGS:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFFLAGS);
|
1994-10-08 01:40:23 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
/*
|
|
|
|
* Currently, no driver owned flags pass the IFF_CANTCHANGE
|
|
|
|
* check, so we don't need special handling here yet.
|
|
|
|
*/
|
2002-08-18 07:05:00 +00:00
|
|
|
new_flags = (ifr->ifr_flags & 0xffff) |
|
|
|
|
(ifr->ifr_flagshigh << 16);
|
1999-06-06 09:17:51 +00:00
|
|
|
if (ifp->if_flags & IFF_SMART) {
|
|
|
|
/* Smart drivers twiddle their own routes */
|
1999-06-06 09:28:01 +00:00
|
|
|
} else if (ifp->if_flags & IFF_UP &&
|
2002-08-18 07:05:00 +00:00
|
|
|
(new_flags & IFF_UP) == 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
int s = splimp();
|
|
|
|
if_down(ifp);
|
|
|
|
splx(s);
|
2002-08-18 07:05:00 +00:00
|
|
|
} else if (new_flags & IFF_UP &&
|
1999-06-06 09:17:51 +00:00
|
|
|
(ifp->if_flags & IFF_UP) == 0) {
|
1994-05-24 10:09:53 +00:00
|
|
|
int s = splimp();
|
|
|
|
if_up(ifp);
|
|
|
|
splx(s);
|
|
|
|
}
|
2005-10-03 01:47:43 +00:00
|
|
|
/* See if permanently promiscuous mode bit is about to flip */
|
|
|
|
if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
|
|
|
|
if (new_flags & IFF_PPROMISC)
|
|
|
|
ifp->if_flags |= IFF_PROMISC;
|
|
|
|
else if (ifp->if_pcount == 0)
|
|
|
|
ifp->if_flags &= ~IFF_PROMISC;
|
|
|
|
log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
|
|
|
|
ifp->if_xname,
|
|
|
|
(new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
|
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
|
2002-08-18 07:05:00 +00:00
|
|
|
(new_flags &~ IFF_CANTCHANGE);
|
2004-10-19 18:11:55 +00:00
|
|
|
if (ifp->if_ioctl) {
|
1994-05-24 10:09:53 +00:00
|
|
|
(void) (*ifp->if_ioctl)(ifp, cmd, data);
|
2004-10-19 18:11:55 +00:00
|
|
|
}
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
1994-05-24 10:09:53 +00:00
|
|
|
break;
|
|
|
|
|
2001-09-18 17:41:42 +00:00
|
|
|
case SIOCSIFCAP:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFCAP);
|
2001-09-18 17:41:42 +00:00
|
|
|
if (error)
|
|
|
|
return (error);
|
2004-02-21 12:48:25 +00:00
|
|
|
if (ifp->if_ioctl == NULL)
|
|
|
|
return (EOPNOTSUPP);
|
2001-09-18 17:41:42 +00:00
|
|
|
if (ifr->ifr_reqcap & ~ifp->if_capabilities)
|
|
|
|
return (EINVAL);
|
2004-02-21 12:48:25 +00:00
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
|
|
|
if (error == 0)
|
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2001-09-18 17:41:42 +00:00
|
|
|
break;
|
|
|
|
|
2002-08-01 21:15:53 +00:00
|
|
|
#ifdef MAC
|
|
|
|
case SIOCSIFMAC:
|
2007-10-24 19:04:04 +00:00
|
|
|
error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
|
2002-08-01 21:15:53 +00:00
|
|
|
break;
|
|
|
|
#endif
|
|
|
|
|
2004-02-04 02:54:25 +00:00
|
|
|
case SIOCSIFNAME:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFNAME);
|
|
|
|
if (error)
|
2004-02-04 02:54:25 +00:00
|
|
|
return (error);
|
|
|
|
error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
|
2004-03-13 02:35:03 +00:00
|
|
|
if (error != 0)
|
2004-02-04 02:54:25 +00:00
|
|
|
return (error);
|
2004-03-13 02:35:03 +00:00
|
|
|
if (new_name[0] == '\0')
|
|
|
|
return (EINVAL);
|
2004-02-04 02:54:25 +00:00
|
|
|
if (ifunit(new_name) != NULL)
|
|
|
|
return (EEXIST);
|
2009-12-29 13:35:18 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX: Locking. Nothing else seems to lock if_flags,
|
|
|
|
* and there are numerous other races with the
|
|
|
|
* ifunit() checks not being atomic with namespace
|
|
|
|
* changes (renames, vmoves, if_attach, etc).
|
|
|
|
*/
|
|
|
|
ifp->if_flags |= IFF_RENAMING;
|
2004-02-04 02:54:25 +00:00
|
|
|
|
|
|
|
/* Announce the departure of the interface. */
|
|
|
|
rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
|
2005-07-14 20:26:43 +00:00
|
|
|
EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
|
2004-02-04 02:54:25 +00:00
|
|
|
|
2004-09-18 05:02:08 +00:00
|
|
|
log(LOG_INFO, "%s: changing name to '%s'\n",
|
|
|
|
ifp->if_xname, new_name);
|
|
|
|
|
2004-02-04 02:54:25 +00:00
|
|
|
strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
|
2005-11-11 16:04:59 +00:00
|
|
|
ifa = ifp->if_addr;
|
2004-02-04 02:54:25 +00:00
|
|
|
IFA_LOCK(ifa);
|
|
|
|
sdl = (struct sockaddr_dl *)ifa->ifa_addr;
|
|
|
|
namelen = strlen(new_name);
|
|
|
|
onamelen = sdl->sdl_nlen;
|
|
|
|
/*
|
|
|
|
* Move the address if needed. This is safe because we
|
|
|
|
* allocate space for a name of length IFNAMSIZ when we
|
|
|
|
* create this in if_attach().
|
|
|
|
*/
|
|
|
|
if (namelen != onamelen) {
|
|
|
|
bcopy(sdl->sdl_data + onamelen,
|
|
|
|
sdl->sdl_data + namelen, sdl->sdl_alen);
|
|
|
|
}
|
|
|
|
bcopy(new_name, sdl->sdl_data, namelen);
|
|
|
|
sdl->sdl_nlen = namelen;
|
|
|
|
sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
|
|
|
|
bzero(sdl->sdl_data, onamelen);
|
|
|
|
while (namelen != 0)
|
|
|
|
sdl->sdl_data[--namelen] = 0xff;
|
|
|
|
IFA_UNLOCK(ifa);
|
|
|
|
|
2004-02-26 04:27:55 +00:00
|
|
|
EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
|
2004-02-04 02:54:25 +00:00
|
|
|
/* Announce the return of the interface. */
|
|
|
|
rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
|
2009-12-29 13:35:18 +00:00
|
|
|
|
|
|
|
ifp->if_flags &= ~IFF_RENAMING;
|
2004-02-04 02:54:25 +00:00
|
|
|
break;
|
|
|
|
|
2009-06-15 18:59:29 +00:00
|
|
|
#ifdef VIMAGE
|
|
|
|
case SIOCSIFVNET:
|
|
|
|
error = priv_check(td, PRIV_NET_SETIFVNET);
|
|
|
|
if (error)
|
|
|
|
return (error);
|
2009-07-26 11:29:26 +00:00
|
|
|
error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
|
2009-06-15 18:59:29 +00:00
|
|
|
break;
|
|
|
|
#endif
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
case SIOCSIFMETRIC:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFMETRIC);
|
1994-10-08 01:40:23 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
ifp->if_metric = ifr->ifr_metric;
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
1994-05-24 10:09:53 +00:00
|
|
|
break;
|
|
|
|
|
1994-12-21 22:57:05 +00:00
|
|
|
case SIOCSIFPHYS:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFPHYS);
|
1996-06-10 23:07:36 +00:00
|
|
|
if (error)
|
2004-02-21 12:56:09 +00:00
|
|
|
return (error);
|
|
|
|
if (ifp->if_ioctl == NULL)
|
|
|
|
return (EOPNOTSUPP);
|
1996-06-10 23:07:36 +00:00
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
|
|
|
if (error == 0)
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2004-02-21 12:56:09 +00:00
|
|
|
break;
|
1994-12-21 22:57:05 +00:00
|
|
|
|
1994-08-08 10:49:26 +00:00
|
|
|
case SIOCSIFMTU:
|
1999-11-22 02:45:11 +00:00
|
|
|
{
|
|
|
|
u_long oldmtu = ifp->if_mtu;
|
|
|
|
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETIFMTU);
|
1994-10-08 01:40:23 +00:00
|
|
|
if (error)
|
1994-08-08 10:49:26 +00:00
|
|
|
return (error);
|
1999-08-06 13:53:03 +00:00
|
|
|
if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
|
1994-08-08 10:58:30 +00:00
|
|
|
return (EINVAL);
|
2001-09-29 05:55:04 +00:00
|
|
|
if (ifp->if_ioctl == NULL)
|
|
|
|
return (EOPNOTSUPP);
|
1996-06-10 23:07:36 +00:00
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
2000-01-24 08:53:39 +00:00
|
|
|
if (error == 0) {
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2000-01-24 08:53:39 +00:00
|
|
|
rt_ifmsg(ifp);
|
|
|
|
}
|
1999-11-22 02:45:11 +00:00
|
|
|
/*
|
|
|
|
* If the link MTU changed, do network layer specific procedure.
|
|
|
|
*/
|
|
|
|
if (ifp->if_mtu != oldmtu) {
|
|
|
|
#ifdef INET6
|
|
|
|
nd6_setmtu(ifp);
|
|
|
|
#endif
|
|
|
|
}
|
2001-09-29 05:55:04 +00:00
|
|
|
break;
|
1999-11-22 02:45:11 +00:00
|
|
|
}
|
1994-08-08 10:49:26 +00:00
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
case SIOCADDMULTI:
|
|
|
|
case SIOCDELMULTI:
|
2006-11-06 13:42:10 +00:00
|
|
|
if (cmd == SIOCADDMULTI)
|
|
|
|
error = priv_check(td, PRIV_NET_ADDMULTI);
|
|
|
|
else
|
|
|
|
error = priv_check(td, PRIV_NET_DELMULTI);
|
1994-10-08 01:40:23 +00:00
|
|
|
if (error)
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
1997-01-13 21:26:53 +00:00
|
|
|
|
|
|
|
/* Don't allow group membership on non-multicast interfaces. */
|
|
|
|
if ((ifp->if_flags & IFF_MULTICAST) == 0)
|
2001-09-29 05:55:04 +00:00
|
|
|
return (EOPNOTSUPP);
|
1997-01-13 21:26:53 +00:00
|
|
|
|
|
|
|
/* Don't let users screw up protocols' entries. */
|
|
|
|
if (ifr->ifr_addr.sa_family != AF_LINK)
|
2001-09-29 05:55:04 +00:00
|
|
|
return (EINVAL);
|
1997-01-13 21:26:53 +00:00
|
|
|
|
|
|
|
if (cmd == SIOCADDMULTI) {
|
|
|
|
struct ifmultiaddr *ifma;
|
2007-03-20 00:36:10 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Userland is only permitted to join groups once
|
|
|
|
* via the if_addmulti() KPI, because it cannot hold
|
|
|
|
* struct ifmultiaddr * between calls. It may also
|
|
|
|
* lose a race while we check if the membership
|
|
|
|
* already exists.
|
|
|
|
*/
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
ifma = if_findmulti(ifp, &ifr->ifr_addr);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
if (ifma != NULL)
|
|
|
|
error = EADDRINUSE;
|
|
|
|
else
|
|
|
|
error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
|
1997-01-13 21:26:53 +00:00
|
|
|
} else {
|
|
|
|
error = if_delmulti(ifp, &ifr->ifr_addr);
|
|
|
|
}
|
|
|
|
if (error == 0)
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2001-09-29 05:55:04 +00:00
|
|
|
break;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2000-10-04 23:16:29 +00:00
|
|
|
case SIOCSIFPHYADDR:
|
|
|
|
case SIOCDIFPHYADDR:
|
|
|
|
#ifdef INET6
|
|
|
|
case SIOCSIFPHYADDR_IN6:
|
|
|
|
#endif
|
2001-06-11 12:39:29 +00:00
|
|
|
case SIOCSLIFPHYADDR:
|
2003-10-23 13:49:10 +00:00
|
|
|
case SIOCSIFMEDIA:
|
1997-10-07 07:40:35 +00:00
|
|
|
case SIOCSIFGENERIC:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_HWIOCTL);
|
1997-05-03 21:07:13 +00:00
|
|
|
if (error)
|
|
|
|
return (error);
|
2001-09-29 05:55:04 +00:00
|
|
|
if (ifp->if_ioctl == NULL)
|
1997-05-03 21:07:13 +00:00
|
|
|
return (EOPNOTSUPP);
|
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
|
|
|
if (error == 0)
|
1998-04-06 11:43:12 +00:00
|
|
|
getmicrotime(&ifp->if_lastchange);
|
2001-09-29 05:55:04 +00:00
|
|
|
break;
|
1997-05-03 21:07:13 +00:00
|
|
|
|
1999-06-19 18:42:31 +00:00
|
|
|
case SIOCGIFSTATUS:
|
|
|
|
ifs = (struct ifstat *)data;
|
|
|
|
ifs->ascii[0] = '\0';
|
2003-10-23 13:49:10 +00:00
|
|
|
|
2001-06-11 12:39:29 +00:00
|
|
|
case SIOCGIFPSRCADDR:
|
|
|
|
case SIOCGIFPDSTADDR:
|
|
|
|
case SIOCGLIFPHYADDR:
|
1997-05-03 21:07:13 +00:00
|
|
|
case SIOCGIFMEDIA:
|
1997-10-07 07:40:35 +00:00
|
|
|
case SIOCGIFGENERIC:
|
2004-02-21 12:56:09 +00:00
|
|
|
if (ifp->if_ioctl == NULL)
|
1997-05-03 21:07:13 +00:00
|
|
|
return (EOPNOTSUPP);
|
2001-09-29 05:55:04 +00:00
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
|
|
|
break;
|
1997-05-03 21:07:13 +00:00
|
|
|
|
2000-06-16 20:14:43 +00:00
|
|
|
case SIOCSIFLLADDR:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_SETLLADDR);
|
2000-06-16 20:14:43 +00:00
|
|
|
if (error)
|
|
|
|
return (error);
|
2001-09-29 05:55:04 +00:00
|
|
|
error = if_setlladdr(ifp,
|
2000-08-15 00:48:38 +00:00
|
|
|
ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
|
2010-01-18 20:34:00 +00:00
|
|
|
EVENTHANDLER_INVOKE(iflladdr_event, ifp);
|
2001-09-29 05:55:04 +00:00
|
|
|
break;
|
2000-08-15 00:48:38 +00:00
|
|
|
|
2006-06-19 22:20:45 +00:00
|
|
|
case SIOCAIFGROUP:
|
|
|
|
{
|
|
|
|
struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
|
|
|
|
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_ADDIFGROUP);
|
2006-06-19 22:20:45 +00:00
|
|
|
if (error)
|
|
|
|
return (error);
|
|
|
|
if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
|
|
|
|
return (error);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case SIOCGIFGROUP:
|
|
|
|
if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
|
|
|
|
return (error);
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SIOCDIFGROUP:
|
|
|
|
{
|
|
|
|
struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
|
|
|
|
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_DELIFGROUP);
|
2006-06-19 22:20:45 +00:00
|
|
|
if (error)
|
|
|
|
return (error);
|
|
|
|
if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
|
|
|
|
return (error);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
default:
|
2001-09-29 05:55:04 +00:00
|
|
|
error = ENOIOCTL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
2010-10-21 16:20:48 +00:00
|
|
|
#ifdef COMPAT_FREEBSD32
|
|
|
|
struct ifconf32 {
|
|
|
|
int32_t ifc_len;
|
|
|
|
union {
|
|
|
|
uint32_t ifcu_buf;
|
|
|
|
uint32_t ifcu_req;
|
|
|
|
} ifc_ifcu;
|
|
|
|
};
|
|
|
|
#define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32)
|
|
|
|
#endif
|
|
|
|
|
2001-09-29 05:55:04 +00:00
|
|
|
/*
|
|
|
|
* Interface ioctls.
|
|
|
|
*/
|
|
|
|
int
|
2003-10-23 13:49:10 +00:00
|
|
|
ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
|
2001-09-29 05:55:04 +00:00
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifreq *ifr;
|
|
|
|
int error;
|
2002-08-18 07:05:00 +00:00
|
|
|
int oif_flags;
|
2001-09-29 05:55:04 +00:00
|
|
|
|
2011-02-16 21:29:13 +00:00
|
|
|
CURVNET_SET(so->so_vnet);
|
2001-09-29 05:55:04 +00:00
|
|
|
switch (cmd) {
|
|
|
|
case SIOCGIFCONF:
|
|
|
|
case OSIOCGIFCONF:
|
2011-02-16 21:29:13 +00:00
|
|
|
error = ifconf(cmd, data);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2010-10-21 16:20:48 +00:00
|
|
|
|
|
|
|
#ifdef COMPAT_FREEBSD32
|
2006-02-02 19:58:37 +00:00
|
|
|
case SIOCGIFCONF32:
|
2010-10-21 16:20:48 +00:00
|
|
|
{
|
|
|
|
struct ifconf32 *ifc32;
|
|
|
|
struct ifconf ifc;
|
|
|
|
|
|
|
|
ifc32 = (struct ifconf32 *)data;
|
|
|
|
ifc.ifc_len = ifc32->ifc_len;
|
|
|
|
ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
|
|
|
|
|
2011-02-16 21:29:13 +00:00
|
|
|
error = ifconf(SIOCGIFCONF, (void *)&ifc);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2010-10-21 16:20:48 +00:00
|
|
|
}
|
2006-02-02 19:58:37 +00:00
|
|
|
#endif
|
2001-09-29 05:55:04 +00:00
|
|
|
}
|
|
|
|
ifr = (struct ifreq *)data;
|
|
|
|
|
|
|
|
switch (cmd) {
|
2009-05-31 12:10:04 +00:00
|
|
|
#ifdef VIMAGE
|
2009-06-15 18:59:29 +00:00
|
|
|
case SIOCSIFRVNET:
|
|
|
|
error = priv_check(td, PRIV_NET_SETIFVNET);
|
2011-02-16 21:29:13 +00:00
|
|
|
if (error == 0)
|
|
|
|
error = if_vmove_reclaim(td, ifr->ifr_name,
|
|
|
|
ifr->ifr_jid);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2009-05-31 12:10:04 +00:00
|
|
|
#endif
|
2001-09-29 05:55:04 +00:00
|
|
|
case SIOCIFCREATE:
|
2006-07-09 06:04:01 +00:00
|
|
|
case SIOCIFCREATE2:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_IFCREATE);
|
2011-02-16 21:29:13 +00:00
|
|
|
if (error == 0)
|
|
|
|
error = if_clone_create(ifr->ifr_name,
|
|
|
|
sizeof(ifr->ifr_name),
|
|
|
|
cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2001-09-29 05:55:04 +00:00
|
|
|
case SIOCIFDESTROY:
|
2006-11-06 13:42:10 +00:00
|
|
|
error = priv_check(td, PRIV_NET_IFDESTROY);
|
2011-02-16 21:29:13 +00:00
|
|
|
if (error == 0)
|
|
|
|
error = if_clone_destroy(ifr->ifr_name);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2003-10-23 13:49:10 +00:00
|
|
|
|
2001-09-29 05:55:04 +00:00
|
|
|
case SIOCIFGCLONERS:
|
2011-02-16 21:29:13 +00:00
|
|
|
error = if_clone_list((struct if_clonereq *)data);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2006-06-19 22:20:45 +00:00
|
|
|
case SIOCGIFGMEMB:
|
2011-02-16 21:29:13 +00:00
|
|
|
error = if_getgroupmembers((struct ifgroupreq *)data);
|
|
|
|
CURVNET_RESTORE();
|
|
|
|
return (error);
|
2001-09-29 05:55:04 +00:00
|
|
|
}
|
|
|
|
|
2009-04-23 13:08:47 +00:00
|
|
|
ifp = ifunit_ref(ifr->ifr_name);
|
2011-02-16 21:29:13 +00:00
|
|
|
if (ifp == NULL) {
|
|
|
|
CURVNET_RESTORE();
|
2001-09-29 05:55:04 +00:00
|
|
|
return (ENXIO);
|
2011-02-16 21:29:13 +00:00
|
|
|
}
|
2001-09-29 05:55:04 +00:00
|
|
|
|
|
|
|
error = ifhwioctl(cmd, ifp, data, td);
|
2009-04-23 13:08:47 +00:00
|
|
|
if (error != ENOIOCTL) {
|
|
|
|
if_rele(ifp);
|
2011-02-16 21:29:13 +00:00
|
|
|
CURVNET_RESTORE();
|
2001-09-29 05:55:04 +00:00
|
|
|
return (error);
|
2009-04-23 13:08:47 +00:00
|
|
|
}
|
2001-09-29 05:55:04 +00:00
|
|
|
|
|
|
|
oif_flags = ifp->if_flags;
|
2009-04-23 13:08:47 +00:00
|
|
|
if (so->so_proto == NULL) {
|
|
|
|
if_rele(ifp);
|
2011-02-16 21:29:13 +00:00
|
|
|
CURVNET_RESTORE();
|
2001-09-29 05:55:04 +00:00
|
|
|
return (EOPNOTSUPP);
|
2009-04-23 13:08:47 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
#ifndef COMPAT_43
|
2001-09-29 05:55:04 +00:00
|
|
|
error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
|
1996-07-11 16:32:50 +00:00
|
|
|
data,
|
2001-09-12 08:38:13 +00:00
|
|
|
ifp, td));
|
2009-03-20 13:41:23 +00:00
|
|
|
if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
|
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
1994-05-24 10:09:53 +00:00
|
|
|
#else
|
2001-09-29 05:55:04 +00:00
|
|
|
{
|
2009-06-21 10:29:31 +00:00
|
|
|
u_long ocmd = cmd;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
|
|
|
switch (cmd) {
|
|
|
|
|
|
|
|
case SIOCSIFDSTADDR:
|
|
|
|
case SIOCSIFADDR:
|
|
|
|
case SIOCSIFBRDADDR:
|
|
|
|
case SIOCSIFNETMASK:
|
|
|
|
#if BYTE_ORDER != BIG_ENDIAN
|
|
|
|
if (ifr->ifr_addr.sa_family == 0 &&
|
|
|
|
ifr->ifr_addr.sa_len < 16) {
|
|
|
|
ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
|
|
|
|
ifr->ifr_addr.sa_len = 16;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
if (ifr->ifr_addr.sa_len == 0)
|
|
|
|
ifr->ifr_addr.sa_len = 16;
|
|
|
|
#endif
|
|
|
|
break;
|
|
|
|
|
|
|
|
case OSIOCGIFADDR:
|
|
|
|
cmd = SIOCGIFADDR;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case OSIOCGIFDSTADDR:
|
|
|
|
cmd = SIOCGIFDSTADDR;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case OSIOCGIFBRDADDR:
|
|
|
|
cmd = SIOCGIFBRDADDR;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case OSIOCGIFNETMASK:
|
|
|
|
cmd = SIOCGIFNETMASK;
|
|
|
|
}
|
1996-07-11 16:32:50 +00:00
|
|
|
error = ((*so->so_proto->pr_usrreqs->pru_control)(so,
|
|
|
|
cmd,
|
|
|
|
data,
|
2001-09-12 08:38:13 +00:00
|
|
|
ifp, td));
|
2009-03-20 13:41:23 +00:00
|
|
|
if (error == EOPNOTSUPP && ifp != NULL &&
|
|
|
|
ifp->if_ioctl != NULL)
|
|
|
|
error = (*ifp->if_ioctl)(ifp, cmd, data);
|
1994-05-24 10:09:53 +00:00
|
|
|
switch (ocmd) {
|
|
|
|
|
|
|
|
case OSIOCGIFADDR:
|
|
|
|
case OSIOCGIFDSTADDR:
|
|
|
|
case OSIOCGIFBRDADDR:
|
|
|
|
case OSIOCGIFNETMASK:
|
|
|
|
*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
|
|
|
|
|
1999-11-22 02:45:11 +00:00
|
|
|
}
|
2001-09-29 05:55:04 +00:00
|
|
|
}
|
1999-11-22 02:45:11 +00:00
|
|
|
#endif /* COMPAT_43 */
|
|
|
|
|
2001-09-29 05:55:04 +00:00
|
|
|
if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
|
1999-11-22 02:45:11 +00:00
|
|
|
#ifdef INET6
|
2001-09-29 05:55:04 +00:00
|
|
|
if (ifp->if_flags & IFF_UP) {
|
|
|
|
int s = splimp();
|
|
|
|
in6_if_up(ifp);
|
|
|
|
splx(s);
|
1999-11-22 02:45:11 +00:00
|
|
|
}
|
2001-09-29 05:55:04 +00:00
|
|
|
#endif
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2009-04-23 13:08:47 +00:00
|
|
|
if_rele(ifp);
|
2011-02-16 21:29:13 +00:00
|
|
|
CURVNET_RESTORE();
|
2001-09-29 05:55:04 +00:00
|
|
|
return (error);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
|
|
|
|
1995-09-22 17:57:48 +00:00
|
|
|
/*
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
* The code common to handling reference counted flags,
|
2005-07-14 13:56:51 +00:00
|
|
|
* e.g., in ifpromisc() and if_allmulti().
|
2005-10-03 02:14:51 +00:00
|
|
|
* The "pflag" argument can specify a permanent mode flag to check,
|
2005-07-14 13:56:51 +00:00
|
|
|
* such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
*
|
|
|
|
* Only to be used on stack-owned flags, not driver-owned flags.
|
1995-09-22 17:57:48 +00:00
|
|
|
*/
|
2005-07-14 13:56:51 +00:00
|
|
|
static int
|
|
|
|
if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
|
1995-09-22 17:57:48 +00:00
|
|
|
{
|
|
|
|
struct ifreq ifr;
|
1997-02-14 15:30:54 +00:00
|
|
|
int error;
|
2005-07-14 13:56:51 +00:00
|
|
|
int oldflags, oldcount;
|
1995-09-22 17:57:48 +00:00
|
|
|
|
2005-10-03 02:14:51 +00:00
|
|
|
/* Sanity checks to catch programming errors */
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
|
2005-10-03 02:14:51 +00:00
|
|
|
("%s: setting driver-owned flag %d", __func__, flag));
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
|
2005-10-03 02:14:51 +00:00
|
|
|
if (onswitch)
|
|
|
|
KASSERT(*refcount >= 0,
|
|
|
|
("%s: increment negative refcount %d for flag %d",
|
|
|
|
__func__, *refcount, flag));
|
|
|
|
else
|
|
|
|
KASSERT(*refcount > 0,
|
|
|
|
("%s: decrement non-positive refcount %d for flag %d",
|
|
|
|
__func__, *refcount, flag));
|
2005-07-14 13:56:51 +00:00
|
|
|
|
|
|
|
/* In case this mode is permanent, just touch refcount */
|
|
|
|
if (ifp->if_flags & pflag) {
|
|
|
|
*refcount += onswitch ? 1 : -1;
|
2002-08-19 15:16:38 +00:00
|
|
|
return (0);
|
|
|
|
}
|
2005-07-14 13:56:51 +00:00
|
|
|
|
|
|
|
/* Save ifnet parameters for if_ioctl() may fail */
|
|
|
|
oldcount = *refcount;
|
|
|
|
oldflags = ifp->if_flags;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* See if we aren't the only and touching refcount is enough.
|
|
|
|
* Actually toggle interface flag if we are the first or last.
|
|
|
|
*/
|
|
|
|
if (onswitch) {
|
|
|
|
if ((*refcount)++)
|
1995-09-22 17:57:48 +00:00
|
|
|
return (0);
|
2005-07-14 13:56:51 +00:00
|
|
|
ifp->if_flags |= flag;
|
1995-09-22 17:57:48 +00:00
|
|
|
} else {
|
2005-07-14 13:56:51 +00:00
|
|
|
if (--(*refcount))
|
1995-09-22 17:57:48 +00:00
|
|
|
return (0);
|
2005-07-14 13:56:51 +00:00
|
|
|
ifp->if_flags &= ~flag;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Call down the driver since we've changed interface flags */
|
|
|
|
if (ifp->if_ioctl == NULL) {
|
|
|
|
error = EOPNOTSUPP;
|
|
|
|
goto recover;
|
1995-09-22 17:57:48 +00:00
|
|
|
}
|
2002-08-18 07:05:00 +00:00
|
|
|
ifr.ifr_flags = ifp->if_flags & 0xffff;
|
|
|
|
ifr.ifr_flagshigh = ifp->if_flags >> 16;
|
1997-02-14 15:30:54 +00:00
|
|
|
error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
|
2005-07-14 13:56:51 +00:00
|
|
|
if (error)
|
|
|
|
goto recover;
|
|
|
|
/* Notify userland that interface flags have changed */
|
|
|
|
rt_ifmsg(ifp);
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
recover:
|
|
|
|
/* Recover after driver error */
|
|
|
|
*refcount = oldcount;
|
|
|
|
ifp->if_flags = oldflags;
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set/clear promiscuous mode on interface ifp based on the truth value
|
|
|
|
* of pswitch. The calls are reference counted so that only the first
|
|
|
|
* "on" request actually has an effect, as does the final "off" request.
|
|
|
|
* Results are undefined if the "off" and "on" requests are not matched.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
ifpromisc(struct ifnet *ifp, int pswitch)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
int oldflags = ifp->if_flags;
|
|
|
|
|
|
|
|
error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
|
|
|
|
&ifp->if_pcount, pswitch);
|
|
|
|
/* If promiscuous mode status has changed, log a message */
|
|
|
|
if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
|
2003-10-31 18:32:15 +00:00
|
|
|
log(LOG_INFO, "%s: promiscuous mode %s\n",
|
|
|
|
ifp->if_xname,
|
2001-04-27 22:20:22 +00:00
|
|
|
(ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
|
2005-07-14 13:56:51 +00:00
|
|
|
return (error);
|
1995-09-22 17:57:48 +00:00
|
|
|
}
|
|
|
|
|
1994-05-24 10:09:53 +00:00
|
|
|
/*
|
|
|
|
* Return interface configuration
|
|
|
|
* of system. List may be used
|
|
|
|
* in later ioctl's (above) to get
|
|
|
|
* other information.
|
|
|
|
*/
|
|
|
|
/*ARGSUSED*/
|
1995-12-09 20:47:15 +00:00
|
|
|
static int
|
2003-10-23 13:49:10 +00:00
|
|
|
ifconf(u_long cmd, caddr_t data)
|
1994-05-24 10:09:53 +00:00
|
|
|
{
|
2001-09-06 00:44:45 +00:00
|
|
|
struct ifconf *ifc = (struct ifconf *)data;
|
|
|
|
struct ifnet *ifp;
|
|
|
|
struct ifaddr *ifa;
|
2004-09-22 08:59:41 +00:00
|
|
|
struct ifreq ifr;
|
|
|
|
struct sbuf *sb;
|
|
|
|
int error, full = 0, valid_len, max_len;
|
|
|
|
|
|
|
|
/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
|
|
|
|
max_len = MAXPHYS - 1;
|
|
|
|
|
2005-02-12 17:51:12 +00:00
|
|
|
/* Prevent hostile input from being able to crash the system */
|
|
|
|
if (ifc->ifc_len <= 0)
|
|
|
|
return (EINVAL);
|
|
|
|
|
2004-09-22 08:59:41 +00:00
|
|
|
again:
|
|
|
|
if (ifc->ifc_len <= max_len) {
|
|
|
|
max_len = ifc->ifc_len;
|
|
|
|
full = 1;
|
|
|
|
}
|
|
|
|
sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
|
|
|
|
max_len = 0;
|
|
|
|
valid_len = 0;
|
1994-05-24 10:09:53 +00:00
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
|
2003-10-31 18:32:15 +00:00
|
|
|
int addrs;
|
1995-05-30 08:16:23 +00:00
|
|
|
|
2005-04-15 01:52:40 +00:00
|
|
|
/*
|
|
|
|
* Zero the ifr_name buffer to make sure we don't
|
|
|
|
* disclose the contents of the stack.
|
|
|
|
*/
|
|
|
|
memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
|
|
|
|
|
2003-10-31 18:32:15 +00:00
|
|
|
if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
|
2005-09-04 17:32:47 +00:00
|
|
|
>= sizeof(ifr.ifr_name)) {
|
|
|
|
sbuf_delete(sb);
|
|
|
|
IFNET_RUNLOCK();
|
2004-09-22 08:59:41 +00:00
|
|
|
return (ENAMETOOLONG);
|
2005-09-04 17:32:47 +00:00
|
|
|
}
|
1994-10-05 20:11:28 +00:00
|
|
|
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
addrs = 0;
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2001-09-07 05:32:54 +00:00
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
|
|
|
struct sockaddr *sa = ifa->ifa_addr;
|
|
|
|
|
2009-02-05 14:06:09 +00:00
|
|
|
if (prison_if(curthread->td_ucred, sa) != 0)
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
continue;
|
|
|
|
addrs++;
|
1994-05-24 10:09:53 +00:00
|
|
|
#ifdef COMPAT_43
|
|
|
|
if (cmd == OSIOCGIFCONF) {
|
|
|
|
struct osockaddr *osa =
|
|
|
|
(struct osockaddr *)&ifr.ifr_addr;
|
|
|
|
ifr.ifr_addr = *sa;
|
|
|
|
osa->sa_family = sa->sa_family;
|
2004-09-22 08:59:41 +00:00
|
|
|
sbuf_bcat(sb, &ifr, sizeof(ifr));
|
|
|
|
max_len += sizeof(ifr);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else
|
|
|
|
#endif
|
|
|
|
if (sa->sa_len <= sizeof(*sa)) {
|
|
|
|
ifr.ifr_addr = *sa;
|
2004-09-22 08:59:41 +00:00
|
|
|
sbuf_bcat(sb, &ifr, sizeof(ifr));
|
|
|
|
max_len += sizeof(ifr);
|
1994-05-24 10:09:53 +00:00
|
|
|
} else {
|
2004-09-22 08:59:41 +00:00
|
|
|
sbuf_bcat(sb, &ifr,
|
|
|
|
offsetof(struct ifreq, ifr_addr));
|
|
|
|
max_len += offsetof(struct ifreq, ifr_addr);
|
|
|
|
sbuf_bcat(sb, sa, sa->sa_len);
|
|
|
|
max_len += sa->sa_len;
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2004-09-22 08:59:41 +00:00
|
|
|
|
2010-09-10 16:42:16 +00:00
|
|
|
if (sbuf_error(sb) == 0)
|
2004-09-22 08:59:41 +00:00
|
|
|
valid_len = sbuf_len(sb);
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2009-04-21 19:06:47 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2004-09-22 08:59:41 +00:00
|
|
|
if (addrs == 0) {
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
|
2004-09-22 08:59:41 +00:00
|
|
|
sbuf_bcat(sb, &ifr, sizeof(ifr));
|
|
|
|
max_len += sizeof(ifr);
|
|
|
|
|
2010-09-10 16:42:16 +00:00
|
|
|
if (sbuf_error(sb) == 0)
|
2004-09-22 08:59:41 +00:00
|
|
|
valid_len = sbuf_len(sb);
|
This Implements the mumbled about "Jail" feature.
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
|
|
|
}
|
1994-05-24 10:09:53 +00:00
|
|
|
}
|
2002-12-22 05:35:03 +00:00
|
|
|
IFNET_RUNLOCK();
|
2004-09-22 08:59:41 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If we didn't allocate enough space (uncommon), try again. If
|
|
|
|
* we have already allocated as much space as we are allowed,
|
|
|
|
* return what we've got.
|
|
|
|
*/
|
|
|
|
if (valid_len != max_len && !full) {
|
|
|
|
sbuf_delete(sb);
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
|
|
|
ifc->ifc_len = valid_len;
|
2004-09-22 12:53:27 +00:00
|
|
|
sbuf_finish(sb);
|
2004-09-22 08:59:41 +00:00
|
|
|
error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
|
|
|
|
sbuf_delete(sb);
|
1994-05-24 10:09:53 +00:00
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
1997-01-07 19:15:32 +00:00
|
|
|
/*
|
2005-02-22 15:29:29 +00:00
|
|
|
* Just like ifpromisc(), but for all-multicast-reception mode.
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
|
|
|
int
|
2003-10-23 13:49:10 +00:00
|
|
|
if_allmulti(struct ifnet *ifp, int onswitch)
|
1997-01-07 19:15:32 +00:00
|
|
|
{
|
1997-02-14 15:30:54 +00:00
|
|
|
|
2005-07-14 13:56:51 +00:00
|
|
|
return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
|
|
|
|
2007-03-20 03:15:43 +00:00
|
|
|
struct ifmultiaddr *
|
2005-08-02 23:23:26 +00:00
|
|
|
if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
|
|
|
|
{
|
|
|
|
struct ifmultiaddr *ifma;
|
|
|
|
|
|
|
|
IF_ADDR_LOCK_ASSERT(ifp);
|
|
|
|
|
|
|
|
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
|
2007-02-22 00:14:02 +00:00
|
|
|
if (sa->sa_family == AF_LINK) {
|
|
|
|
if (sa_dl_equal(ifma->ifma_addr, sa))
|
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
if (sa_equal(ifma->ifma_addr, sa))
|
|
|
|
break;
|
|
|
|
}
|
2005-08-02 23:23:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return ifma;
|
|
|
|
}
|
|
|
|
|
1997-01-07 19:15:32 +00:00
|
|
|
/*
|
2005-08-02 23:23:26 +00:00
|
|
|
* Allocate a new ifmultiaddr and initialize based on passed arguments. We
|
|
|
|
* make copies of passed sockaddrs. The ifmultiaddr will not be added to
|
|
|
|
* the ifnet multicast address list here, so the caller must do that and
|
|
|
|
* other setup work (such as notifying the device driver). The reference
|
|
|
|
* count is initialized to 1.
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
2005-08-02 23:23:26 +00:00
|
|
|
static struct ifmultiaddr *
|
|
|
|
if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
|
|
|
|
int mflags)
|
1997-01-07 19:15:32 +00:00
|
|
|
{
|
|
|
|
struct ifmultiaddr *ifma;
|
2005-08-02 23:23:26 +00:00
|
|
|
struct sockaddr *dupsa;
|
|
|
|
|
2008-10-23 15:53:51 +00:00
|
|
|
ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
|
2005-08-02 23:23:26 +00:00
|
|
|
M_ZERO);
|
|
|
|
if (ifma == NULL)
|
|
|
|
return (NULL);
|
|
|
|
|
2008-10-23 15:53:51 +00:00
|
|
|
dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
|
2005-08-02 23:23:26 +00:00
|
|
|
if (dupsa == NULL) {
|
2008-10-23 15:53:51 +00:00
|
|
|
free(ifma, M_IFMADDR);
|
2005-08-02 23:23:26 +00:00
|
|
|
return (NULL);
|
|
|
|
}
|
|
|
|
bcopy(sa, dupsa, sa->sa_len);
|
|
|
|
ifma->ifma_addr = dupsa;
|
|
|
|
|
|
|
|
ifma->ifma_ifp = ifp;
|
|
|
|
ifma->ifma_refcount = 1;
|
|
|
|
ifma->ifma_protospec = NULL;
|
|
|
|
|
|
|
|
if (llsa == NULL) {
|
|
|
|
ifma->ifma_lladdr = NULL;
|
|
|
|
return (ifma);
|
|
|
|
}
|
|
|
|
|
2008-10-23 15:53:51 +00:00
|
|
|
dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
|
2005-08-02 23:23:26 +00:00
|
|
|
if (dupsa == NULL) {
|
2008-10-23 15:53:51 +00:00
|
|
|
free(ifma->ifma_addr, M_IFMADDR);
|
|
|
|
free(ifma, M_IFMADDR);
|
2005-08-02 23:23:26 +00:00
|
|
|
return (NULL);
|
|
|
|
}
|
|
|
|
bcopy(llsa, dupsa, llsa->sa_len);
|
|
|
|
ifma->ifma_lladdr = dupsa;
|
|
|
|
|
|
|
|
return (ifma);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if_freemulti: free ifmultiaddr structure and possibly attached related
|
|
|
|
* addresses. The caller is responsible for implementing reference
|
|
|
|
* counting, notifying the driver, handling routing messages, and releasing
|
|
|
|
* any dependent link layer state.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
if_freemulti(struct ifmultiaddr *ifma)
|
|
|
|
{
|
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
|
2005-08-02 23:23:26 +00:00
|
|
|
ifma->ifma_refcount));
|
|
|
|
KASSERT(ifma->ifma_protospec == NULL,
|
|
|
|
("if_freemulti: protospec not NULL"));
|
|
|
|
|
|
|
|
if (ifma->ifma_lladdr != NULL)
|
2008-10-23 15:53:51 +00:00
|
|
|
free(ifma->ifma_lladdr, M_IFMADDR);
|
|
|
|
free(ifma->ifma_addr, M_IFMADDR);
|
|
|
|
free(ifma, M_IFMADDR);
|
2005-08-02 23:23:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Register an additional multicast address with a network interface.
|
|
|
|
*
|
|
|
|
* - If the address is already present, bump the reference count on the
|
|
|
|
* address and return.
|
|
|
|
* - If the address is not link-layer, look up a link layer address.
|
|
|
|
* - Allocate address structures for one or both addresses, and attach to the
|
|
|
|
* multicast address list on the interface. If automatically adding a link
|
|
|
|
* layer address, the protocol address will own a reference to the link
|
|
|
|
* layer address, to be freed when it is freed.
|
|
|
|
* - Notify the network device driver of an addition to the multicast address
|
|
|
|
* list.
|
|
|
|
*
|
|
|
|
* 'sa' points to caller-owned memory with the desired multicast address.
|
|
|
|
*
|
|
|
|
* 'retifma' will be used to return a pointer to the resulting multicast
|
|
|
|
* address reference, if desired.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
|
|
|
|
struct ifmultiaddr **retifma)
|
|
|
|
{
|
|
|
|
struct ifmultiaddr *ifma, *ll_ifma;
|
|
|
|
struct sockaddr *llsa;
|
|
|
|
int error;
|
1997-01-07 19:15:32 +00:00
|
|
|
|
1997-07-07 17:36:06 +00:00
|
|
|
/*
|
2005-08-02 23:23:26 +00:00
|
|
|
* If the address is already present, return a new reference to it;
|
|
|
|
* otherwise, allocate storage and set up a new address.
|
1997-07-07 17:36:06 +00:00
|
|
|
*/
|
2005-08-02 23:23:26 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
ifma = if_findmulti(ifp, sa);
|
|
|
|
if (ifma != NULL) {
|
|
|
|
ifma->ifma_refcount++;
|
|
|
|
if (retifma != NULL)
|
|
|
|
*retifma = ifma;
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
return (0);
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2005-08-02 23:23:26 +00:00
|
|
|
* The address isn't already present; resolve the protocol address
|
|
|
|
* into a link layer address, and then look that up, bump its
|
|
|
|
* refcount or allocate an ifma for that also. If 'llsa' was
|
|
|
|
* returned, we will need to free it later.
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
2005-08-02 23:23:26 +00:00
|
|
|
llsa = NULL;
|
|
|
|
ll_ifma = NULL;
|
2005-07-19 10:12:58 +00:00
|
|
|
if (ifp->if_resolvemulti != NULL) {
|
1997-01-07 19:15:32 +00:00
|
|
|
error = ifp->if_resolvemulti(ifp, &llsa, sa);
|
2005-08-02 23:23:26 +00:00
|
|
|
if (error)
|
|
|
|
goto unlock_out;
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
/*
|
|
|
|
* Allocate the new address. Don't hook it up yet, as we may also
|
|
|
|
* need to allocate a link layer multicast address.
|
|
|
|
*/
|
|
|
|
ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
|
|
|
|
if (ifma == NULL) {
|
|
|
|
error = ENOMEM;
|
|
|
|
goto free_llsa_out;
|
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
/*
|
|
|
|
* If a link layer address is found, we'll need to see if it's
|
|
|
|
* already present in the address list, or allocate is as well.
|
|
|
|
* When this block finishes, the link layer address will be on the
|
|
|
|
* list.
|
|
|
|
*/
|
|
|
|
if (llsa != NULL) {
|
|
|
|
ll_ifma = if_findmulti(ifp, llsa);
|
|
|
|
if (ll_ifma == NULL) {
|
|
|
|
ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
|
|
|
|
if (ll_ifma == NULL) {
|
2007-03-20 00:36:10 +00:00
|
|
|
--ifma->ifma_refcount;
|
2005-08-02 23:23:26 +00:00
|
|
|
if_freemulti(ifma);
|
|
|
|
error = ENOMEM;
|
|
|
|
goto free_llsa_out;
|
|
|
|
}
|
|
|
|
TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
|
|
|
|
ifma_link);
|
|
|
|
} else
|
|
|
|
ll_ifma->ifma_refcount++;
|
2007-03-20 00:36:10 +00:00
|
|
|
ifma->ifma_llifma = ll_ifma;
|
2005-08-02 23:23:26 +00:00
|
|
|
}
|
1997-01-08 13:20:25 +00:00
|
|
|
|
1997-01-07 19:15:32 +00:00
|
|
|
/*
|
2005-08-02 23:23:26 +00:00
|
|
|
* We now have a new multicast address, ifma, and possibly a new or
|
|
|
|
* referenced link layer address. Add the primary address to the
|
|
|
|
* ifnet address list.
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
2001-02-06 10:12:15 +00:00
|
|
|
TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
|
2005-08-02 23:23:26 +00:00
|
|
|
|
2002-07-02 08:23:00 +00:00
|
|
|
if (retifma != NULL)
|
|
|
|
*retifma = ifma;
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
/*
|
|
|
|
* Must generate the message while holding the lock so that 'ifma'
|
|
|
|
* pointer is still valid.
|
|
|
|
*/
|
|
|
|
rt_newmaddrmsg(RTM_NEWMADDR, ifma);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
1997-01-07 19:15:32 +00:00
|
|
|
/*
|
|
|
|
* We are certain we have added something, so call down to the
|
|
|
|
* interface to let them know about it.
|
|
|
|
*/
|
2005-07-19 10:12:58 +00:00
|
|
|
if (ifp->if_ioctl != NULL) {
|
2005-07-14 13:56:51 +00:00
|
|
|
(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
|
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
if (llsa != NULL)
|
2008-10-23 15:53:51 +00:00
|
|
|
free(llsa, M_IFMADDR);
|
2005-08-02 23:23:26 +00:00
|
|
|
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
free_llsa_out:
|
|
|
|
if (llsa != NULL)
|
2008-10-23 15:53:51 +00:00
|
|
|
free(llsa, M_IFMADDR);
|
2005-08-02 23:23:26 +00:00
|
|
|
|
|
|
|
unlock_out:
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
return (error);
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2007-03-20 00:36:10 +00:00
|
|
|
* Delete a multicast group membership by network-layer group address.
|
|
|
|
*
|
|
|
|
* Returns ENOENT if the entry could not be found. If ifp no longer
|
|
|
|
* exists, results are undefined. This entry point should only be used
|
|
|
|
* from subsystems which do appropriate locking to hold ifp for the
|
|
|
|
* duration of the call.
|
|
|
|
* Network-layer protocol domains must use if_delmulti_ifma().
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
|
|
|
int
|
2003-10-23 13:49:10 +00:00
|
|
|
if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
|
1997-01-07 19:15:32 +00:00
|
|
|
{
|
2007-03-20 00:36:10 +00:00
|
|
|
struct ifmultiaddr *ifma;
|
|
|
|
int lastref;
|
|
|
|
#ifdef INVARIANTS
|
|
|
|
struct ifnet *oifp;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(oifp, &V_ifnet, if_link)
|
2007-03-20 00:36:10 +00:00
|
|
|
if (ifp == oifp)
|
|
|
|
break;
|
|
|
|
if (ifp != oifp)
|
|
|
|
ifp = NULL;
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2007-03-20 00:36:10 +00:00
|
|
|
|
|
|
|
KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
|
|
|
|
#endif
|
|
|
|
if (ifp == NULL)
|
|
|
|
return (ENOENT);
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2007-03-20 00:36:10 +00:00
|
|
|
lastref = 0;
|
2005-08-02 23:23:26 +00:00
|
|
|
ifma = if_findmulti(ifp, sa);
|
2007-03-20 00:36:10 +00:00
|
|
|
if (ifma != NULL)
|
|
|
|
lastref = if_delmulti_locked(ifp, ifma, 0);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
|
|
|
|
if (ifma == NULL)
|
|
|
|
return (ENOENT);
|
|
|
|
|
|
|
|
if (lastref && ifp->if_ioctl != NULL) {
|
|
|
|
(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
|
2005-08-02 23:23:26 +00:00
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2010-01-24 16:17:58 +00:00
|
|
|
/*
|
|
|
|
* Delete all multicast group membership for an interface.
|
|
|
|
* Should be used to quickly flush all multicast filters.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_delallmulti(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
struct ifmultiaddr *ifma;
|
|
|
|
struct ifmultiaddr *next;
|
|
|
|
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
|
|
|
|
if_delmulti_locked(ifp, ifma, 0);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
|
|
|
}
|
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
/*
|
|
|
|
* Delete a multicast group membership by group membership pointer.
|
|
|
|
* Network-layer protocol domains must use this routine.
|
|
|
|
*
|
2009-03-15 14:21:05 +00:00
|
|
|
* It is safe to call this routine if the ifp disappeared.
|
2007-03-20 00:36:10 +00:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_delmulti_ifma(struct ifmultiaddr *ifma)
|
|
|
|
{
|
|
|
|
struct ifnet *ifp;
|
|
|
|
int lastref;
|
|
|
|
|
|
|
|
ifp = ifma->ifma_ifp;
|
|
|
|
#ifdef DIAGNOSTIC
|
|
|
|
if (ifp == NULL) {
|
|
|
|
printf("%s: ifma_ifp seems to be detached\n", __func__);
|
|
|
|
} else {
|
|
|
|
struct ifnet *oifp;
|
|
|
|
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RLOCK_NOSLEEP();
|
Commit step 1 of the vimage project, (network stack)
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
2008-08-17 23:27:27 +00:00
|
|
|
TAILQ_FOREACH(oifp, &V_ifnet, if_link)
|
2007-03-20 00:36:10 +00:00
|
|
|
if (ifp == oifp)
|
|
|
|
break;
|
|
|
|
if (ifp != oifp) {
|
|
|
|
printf("%s: ifnet %p disappeared\n", __func__, ifp);
|
|
|
|
ifp = NULL;
|
|
|
|
}
|
2009-08-23 20:40:19 +00:00
|
|
|
IFNET_RUNLOCK_NOSLEEP();
|
2007-03-20 00:36:10 +00:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
/*
|
|
|
|
* If and only if the ifnet instance exists: Acquire the address lock.
|
|
|
|
*/
|
|
|
|
if (ifp != NULL)
|
|
|
|
IF_ADDR_LOCK(ifp);
|
|
|
|
|
|
|
|
lastref = if_delmulti_locked(ifp, ifma, 0);
|
|
|
|
|
|
|
|
if (ifp != NULL) {
|
|
|
|
/*
|
|
|
|
* If and only if the ifnet instance exists:
|
|
|
|
* Release the address lock.
|
|
|
|
* If the group was left: update the hardware hash filter.
|
|
|
|
*/
|
2005-08-02 23:23:26 +00:00
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2007-03-20 00:36:10 +00:00
|
|
|
if (lastref && ifp->if_ioctl != NULL) {
|
|
|
|
(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
|
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
2007-03-20 00:36:10 +00:00
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
/*
|
|
|
|
* Perform deletion of network-layer and/or link-layer multicast address.
|
|
|
|
*
|
|
|
|
* Return 0 if the reference count was decremented.
|
|
|
|
* Return 1 if the final reference was released, indicating that the
|
|
|
|
* hardware hash filter should be reprogrammed.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
|
|
|
|
{
|
|
|
|
struct ifmultiaddr *ll_ifma;
|
|
|
|
|
|
|
|
if (ifp != NULL && ifma->ifma_ifp != NULL) {
|
|
|
|
KASSERT(ifma->ifma_ifp == ifp,
|
|
|
|
("%s: inconsistent ifp %p", __func__, ifp));
|
|
|
|
IF_ADDR_LOCK_ASSERT(ifp);
|
|
|
|
}
|
|
|
|
|
|
|
|
ifp = ifma->ifma_ifp;
|
1997-01-07 19:15:32 +00:00
|
|
|
|
|
|
|
/*
|
2007-03-20 00:36:10 +00:00
|
|
|
* If the ifnet is detaching, null out references to ifnet,
|
|
|
|
* so that upper protocol layers will notice, and not attempt
|
2007-03-27 16:11:28 +00:00
|
|
|
* to obtain locks for an ifnet which no longer exists. The
|
|
|
|
* routing socket announcement must happen before the ifnet
|
|
|
|
* instance is detached from the system.
|
1997-01-07 19:15:32 +00:00
|
|
|
*/
|
2007-03-20 00:36:10 +00:00
|
|
|
if (detaching) {
|
|
|
|
#ifdef DIAGNOSTIC
|
|
|
|
printf("%s: detaching ifnet instance %p\n", __func__, ifp);
|
|
|
|
#endif
|
2007-03-27 16:11:28 +00:00
|
|
|
/*
|
|
|
|
* ifp may already be nulled out if we are being reentered
|
|
|
|
* to delete the ll_ifma.
|
|
|
|
*/
|
|
|
|
if (ifp != NULL) {
|
|
|
|
rt_newmaddrmsg(RTM_DELMADDR, ifma);
|
|
|
|
ifma->ifma_ifp = NULL;
|
|
|
|
}
|
2007-03-20 00:36:10 +00:00
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2007-03-20 00:36:10 +00:00
|
|
|
if (--ifma->ifma_refcount > 0)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this ifma is a network-layer ifma, a link-layer ifma may
|
|
|
|
* have been associated with it. Release it first if so.
|
|
|
|
*/
|
|
|
|
ll_ifma = ifma->ifma_llifma;
|
2005-08-02 23:23:26 +00:00
|
|
|
if (ll_ifma != NULL) {
|
2007-03-20 00:36:10 +00:00
|
|
|
KASSERT(ifma->ifma_lladdr != NULL,
|
|
|
|
("%s: llifma w/o lladdr", __func__));
|
|
|
|
if (detaching)
|
|
|
|
ll_ifma->ifma_ifp = NULL; /* XXX */
|
|
|
|
if (--ll_ifma->ifma_refcount == 0) {
|
|
|
|
if (ifp != NULL) {
|
|
|
|
TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
|
|
|
|
ifma_link);
|
|
|
|
}
|
2005-08-02 23:23:26 +00:00
|
|
|
if_freemulti(ll_ifma);
|
2007-03-20 00:36:10 +00:00
|
|
|
}
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
2007-03-20 00:36:10 +00:00
|
|
|
|
|
|
|
if (ifp != NULL)
|
|
|
|
TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
|
|
|
|
|
|
|
|
if_freemulti(ifma);
|
1997-01-07 19:15:32 +00:00
|
|
|
|
2005-08-02 23:23:26 +00:00
|
|
|
/*
|
2007-03-20 00:36:10 +00:00
|
|
|
* The last reference to this instance of struct ifmultiaddr
|
|
|
|
* was released; the hardware should be notified of this change.
|
2005-08-02 23:23:26 +00:00
|
|
|
*/
|
2007-03-20 00:36:10 +00:00
|
|
|
return 1;
|
1997-01-07 19:15:32 +00:00
|
|
|
}
|
|
|
|
|
2000-08-15 00:48:38 +00:00
|
|
|
/*
|
|
|
|
* Set the link layer address on an interface.
|
|
|
|
*
|
|
|
|
* At this time we only support certain types of interfaces,
|
|
|
|
* and we don't allow the length of the address to change.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
|
|
|
|
{
|
|
|
|
struct sockaddr_dl *sdl;
|
|
|
|
struct ifaddr *ifa;
|
2002-04-10 06:07:16 +00:00
|
|
|
struct ifreq ifr;
|
2000-08-15 00:48:38 +00:00
|
|
|
|
2009-06-24 10:36:48 +00:00
|
|
|
IF_ADDR_LOCK(ifp);
|
2005-11-11 16:04:59 +00:00
|
|
|
ifa = ifp->if_addr;
|
2009-06-24 10:36:48 +00:00
|
|
|
if (ifa == NULL) {
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2000-08-15 00:48:38 +00:00
|
|
|
return (EINVAL);
|
2009-06-24 10:36:48 +00:00
|
|
|
}
|
|
|
|
ifa_ref(ifa);
|
|
|
|
IF_ADDR_UNLOCK(ifp);
|
2000-08-15 00:48:38 +00:00
|
|
|
sdl = (struct sockaddr_dl *)ifa->ifa_addr;
|
2009-06-24 10:36:48 +00:00
|
|
|
if (sdl == NULL) {
|
|
|
|
ifa_free(ifa);
|
2000-08-15 00:48:38 +00:00
|
|
|
return (EINVAL);
|
2009-06-24 10:36:48 +00:00
|
|
|
}
|
|
|
|
if (len != sdl->sdl_alen) { /* don't allow length to change */
|
|
|
|
ifa_free(ifa);
|
2000-08-15 00:48:38 +00:00
|
|
|
return (EINVAL);
|
2009-06-24 10:36:48 +00:00
|
|
|
}
|
2000-08-15 00:48:38 +00:00
|
|
|
switch (ifp->if_type) {
|
2005-11-11 07:36:14 +00:00
|
|
|
case IFT_ETHER:
|
2000-08-15 00:48:38 +00:00
|
|
|
case IFT_FDDI:
|
|
|
|
case IFT_XETHER:
|
|
|
|
case IFT_ISO88025:
|
2001-04-04 15:10:58 +00:00
|
|
|
case IFT_L2VLAN:
|
2005-06-05 03:13:13 +00:00
|
|
|
case IFT_BRIDGE:
|
2003-01-24 01:32:20 +00:00
|
|
|
case IFT_ARCNET:
|
2007-04-10 00:27:25 +00:00
|
|
|
case IFT_IEEE8023ADLAG:
|
2009-03-28 17:36:56 +00:00
|
|
|
case IFT_IEEE80211:
|
2000-08-15 00:48:38 +00:00
|
|
|
bcopy(lladdr, LLADDR(sdl), len);
|
2009-06-24 10:36:48 +00:00
|
|
|
ifa_free(ifa);
|
2000-08-15 00:48:38 +00:00
|
|
|
break;
|
|
|
|
default:
|
2009-06-24 10:36:48 +00:00
|
|
|
ifa_free(ifa);
|
2000-08-15 00:48:38 +00:00
|
|
|
return (ENODEV);
|
|
|
|
}
|
2009-06-24 10:36:48 +00:00
|
|
|
|
2000-08-15 00:48:38 +00:00
|
|
|
/*
|
|
|
|
* If the interface is already up, we need
|
|
|
|
* to re-init it in order to reprogram its
|
|
|
|
* address filter.
|
|
|
|
*/
|
|
|
|
if ((ifp->if_flags & IFF_UP) != 0) {
|
2005-07-14 13:56:51 +00:00
|
|
|
if (ifp->if_ioctl) {
|
|
|
|
ifp->if_flags &= ~IFF_UP;
|
|
|
|
ifr.ifr_flags = ifp->if_flags & 0xffff;
|
|
|
|
ifr.ifr_flagshigh = ifp->if_flags >> 16;
|
|
|
|
(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
|
|
|
|
ifp->if_flags |= IFF_UP;
|
|
|
|
ifr.ifr_flags = ifp->if_flags & 0xffff;
|
|
|
|
ifr.ifr_flagshigh = ifp->if_flags >> 16;
|
|
|
|
(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
|
|
|
|
}
|
2002-02-18 22:50:13 +00:00
|
|
|
#ifdef INET
|
|
|
|
/*
|
|
|
|
* Also send gratuitous ARPs to notify other nodes about
|
|
|
|
* the address change.
|
|
|
|
*/
|
|
|
|
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
|
2006-06-29 19:22:05 +00:00
|
|
|
if (ifa->ifa_addr->sa_family == AF_INET)
|
2002-02-26 01:11:08 +00:00
|
|
|
arp_ifinit(ifp, ifa);
|
2002-02-18 22:50:13 +00:00
|
|
|
}
|
|
|
|
#endif
|
2000-08-15 00:48:38 +00:00
|
|
|
}
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2003-10-31 18:32:15 +00:00
|
|
|
/*
|
|
|
|
* The name argument must be a pointer to storage which will last as
|
|
|
|
* long as the interface does. For physical devices, the result of
|
|
|
|
* device_get_name(dev) is a good choice and for pseudo-devices a
|
|
|
|
* static string works well.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
if_initname(struct ifnet *ifp, const char *name, int unit)
|
|
|
|
{
|
|
|
|
ifp->if_dname = name;
|
|
|
|
ifp->if_dunit = unit;
|
|
|
|
if (unit != IF_DUNIT_NONE)
|
|
|
|
snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
|
|
|
|
else
|
|
|
|
strlcpy(ifp->if_xname, name, IFNAMSIZ);
|
|
|
|
}
|
|
|
|
|
2002-09-24 17:35:08 +00:00
|
|
|
int
|
|
|
|
if_printf(struct ifnet *ifp, const char * fmt, ...)
|
|
|
|
{
|
|
|
|
va_list ap;
|
|
|
|
int retval;
|
|
|
|
|
2003-10-31 18:32:15 +00:00
|
|
|
retval = printf("%s: ", ifp->if_xname);
|
2002-09-24 17:35:08 +00:00
|
|
|
va_start(ap, fmt);
|
|
|
|
retval += vprintf(fmt, ap);
|
|
|
|
va_end(ap);
|
|
|
|
return (retval);
|
|
|
|
}
|
|
|
|
|
2004-07-27 23:20:45 +00:00
|
|
|
void
|
|
|
|
if_start(struct ifnet *ifp)
|
|
|
|
{
|
|
|
|
|
2009-03-15 14:21:05 +00:00
|
|
|
(*(ifp)->if_start)(ifp);
|
2004-07-27 23:20:45 +00:00
|
|
|
}
|
|
|
|
|
2008-11-22 05:55:56 +00:00
|
|
|
/*
|
|
|
|
* Backwards compatibility interface for drivers
|
|
|
|
* that have not implemented it
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
if_transmit(struct ifnet *ifp, struct mbuf *m)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
IFQ_HANDOFF(ifp, m, error);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
2004-10-30 09:39:13 +00:00
|
|
|
int
|
|
|
|
if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
|
|
|
|
{
|
|
|
|
int active = 0;
|
|
|
|
|
|
|
|
IF_LOCK(ifq);
|
|
|
|
if (_IF_QFULL(ifq)) {
|
|
|
|
_IF_DROP(ifq);
|
|
|
|
IF_UNLOCK(ifq);
|
|
|
|
m_freem(m);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
if (ifp != NULL) {
|
|
|
|
ifp->if_obytes += m->m_pkthdr.len + adjust;
|
|
|
|
if (m->m_flags & (M_BCAST|M_MCAST))
|
|
|
|
ifp->if_omcasts++;
|
Rename IFF_RUNNING to IFF_DRV_RUNNING, IFF_OACTIVE to IFF_DRV_OACTIVE,
and move both flags from ifnet.if_flags to ifnet.if_drv_flags, making
and documenting the locking of these flags the responsibility of the
device driver, not the network stack. The flags for these two fields
will be mutually exclusive so that they can be exposed to user space as
though they were stored in the same variable.
Provide #defines to provide the old names #ifndef _KERNEL, so that user
applications (such as ifconfig) can use the old flag names. Using the
old names in a device driver will result in a compile error in order to
help device driver writers adopt the new model.
When exposing the interface flags to user space, via interface ioctls
or routing sockets, or the two fields together. Since the driver flags
cannot currently be set for user space, no new logic is currently
required to handle this case.
Add some assertions that general purpose network stack routines, such
as if_setflags(), are not improperly used on driver-owned flags.
With this change, a large number of very minor network stack races are
closed, subject to correct device driver locking. Most were likely
never triggered.
Driver sweep to follow; many thanks to pjd and bz for the line-by-line
review they gave this patch.
Reviewed by: pjd, bz
MFC after: 7 days
2005-08-09 10:16:17 +00:00
|
|
|
active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
|
2004-10-30 09:39:13 +00:00
|
|
|
}
|
|
|
|
_IF_ENQUEUE(ifq, m);
|
|
|
|
IF_UNLOCK(ifq);
|
|
|
|
if (ifp != NULL && !active)
|
2009-03-15 14:21:05 +00:00
|
|
|
(*(ifp)->if_start)(ifp);
|
2004-10-30 09:39:13 +00:00
|
|
|
return (1);
|
|
|
|
}
|
2005-06-10 16:49:24 +00:00
|
|
|
|
|
|
|
void
|
|
|
|
if_register_com_alloc(u_char type,
|
|
|
|
if_com_alloc_t *a, if_com_free_t *f)
|
|
|
|
{
|
|
|
|
|
|
|
|
KASSERT(if_com_alloc[type] == NULL,
|
|
|
|
("if_register_com_alloc: %d already registered", type));
|
|
|
|
KASSERT(if_com_free[type] == NULL,
|
|
|
|
("if_register_com_alloc: %d free already registered", type));
|
|
|
|
|
|
|
|
if_com_alloc[type] = a;
|
|
|
|
if_com_free[type] = f;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
if_deregister_com_alloc(u_char type)
|
|
|
|
{
|
|
|
|
|
2006-06-11 22:09:28 +00:00
|
|
|
KASSERT(if_com_alloc[type] != NULL,
|
2005-06-10 16:49:24 +00:00
|
|
|
("if_deregister_com_alloc: %d not registered", type));
|
2006-06-11 22:09:28 +00:00
|
|
|
KASSERT(if_com_free[type] != NULL,
|
2005-06-10 16:49:24 +00:00
|
|
|
("if_deregister_com_alloc: %d free not registered", type));
|
|
|
|
if_com_alloc[type] = NULL;
|
|
|
|
if_com_free[type] = NULL;
|
|
|
|
}
|