freebsd-nq/sys/net/if.c

4058 lines
94 KiB
C
Raw Normal View History

/*-
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.
*
* @(#)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
*/
#include "opt_compat.h"
#include "opt_inet6.h"
#include "opt_inet.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/types.h>
#include <sys/conf.h>
#include <sys/malloc.h>
#include <sys/sbuf.h>
#include <sys/bus.h>
1994-05-24 10:09:53 +00:00
#include <sys/mbuf.h>
#include <sys/systm.h>
#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>
#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>
#include <sys/rwlock.h>
#include <sys/sockio.h>
#include <sys/syslog.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
#include <sys/domain.h>
#include <sys/jail.h>
#include <sys/priv.h>
#include <machine/stdarg.h>
#include <vm/uma.h>
1994-05-24 10:09:53 +00:00
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
#include <net/bpf.h>
#include <net/ethernet.h>
1994-05-24 10:09:53 +00:00
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_clone.h>
1994-05-24 10:09:53 +00:00
#include <net/if_dl.h>
#include <net/if_types.h>
#include <net/if_var.h>
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
#include <net/if_media.h>
#include <net/if_vlan_var.h>
#include <net/radix.h>
#include <net/route.h>
#include <net/vnet.h>
1994-05-24 10:09:53 +00:00
#if defined(INET) || defined(INET6)
#include <net/ethernet.h>
#include <netinet/in.h>
#include <netinet/in_var.h>
#include <netinet/ip.h>
#include <netinet/ip_carp.h>
#ifdef INET
#include <netinet/if_ether.h>
#endif /* INET */
#ifdef INET6
#include <netinet6/in6_var.h>
#include <netinet6/in6_ifattach.h>
#endif /* INET6 */
#endif /* INET || INET6 */
#include <security/mac/mac_framework.h>
#ifdef COMPAT_FREEBSD32
#include <sys/mount.h>
#include <compat/freebsd32/freebsd32.h>
#endif
SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
&ifqmaxlen, 0, "max send queue size");
/* 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");
/* Log promiscuous mode change events */
static int log_promisc_mode_change = 1;
SYSCTL_INT(_net_link, OID_AUTO, log_promisc_mode_change, CTLFLAG_RDTUN,
&log_promisc_mode_change, 1,
"log promiscuous mode change events");
/* 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");
static 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");
void (*bridge_linkstate_p)(struct ifnet *ifp);
void (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
void (*lagg_linkstate_p)(struct ifnet *ifp, int state);
/* These are external hooks for CARP. */
void (*carp_linkstate_p)(struct ifnet *ifp);
void (*carp_demote_adj_p)(int, char *);
int (*carp_master_p)(struct ifaddr *);
#if defined(INET) || defined(INET6)
int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
const struct sockaddr *sa);
int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);
int (*carp_attach_p)(struct ifaddr *, int);
void (*carp_detach_p)(struct ifaddr *);
#endif
#ifdef INET
int (*carp_iamatch_p)(struct ifaddr *, uint8_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
struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
/*
* XXX: Style; these should be sorted alphabetically, and unprototyped
* static functions should be prototyped. Currently they are sorted by
* declaration order.
*/
static void if_attachdomain(void *);
static void if_attachdomain1(struct ifnet *);
static int ifconf(u_long, caddr_t);
static void if_freemulti(struct ifmultiaddr *);
static void if_grow(void);
static void if_input_default(struct ifnet *, struct mbuf *);
Implement interface link header precomputation API. Add if_requestencap() interface method which is capable of calculating various link headers for given interface. Right now there is support for INET/INET6/ARP llheader calculation (IFENCAP_LL type request). Other types are planned to support more complex calculation (L2 multipath lagg nexthops, tunnel encap nexthops, etc..). Reshape 'struct route' to be able to pass additional data (with is length) to prepend to mbuf. These two changes permits routing code to pass pre-calculated nexthop data (like L2 header for route w/gateway) down to the stack eliminating the need for other lookups. It also brings us closer to more complex scenarios like transparently handling MPLS nexthops and tunnel interfaces. Last, but not least, it removes layering violation introduced by flowtable code (ro_lle) and simplifies handling of existing if_output consumers. ARP/ND changes: Make arp/ndp stack pre-calculate link header upon installing/updating lle record. Interface link address change are handled by re-calculating headers for all lles based on if_lladdr event. After these changes, arpresolve()/nd6_resolve() returns full pre-calculated header for supported interfaces thus simplifying if_output(). Move these lookups to separate ether_resolve_addr() function which ether returs error or fully-prepared link header. Add <arp|nd6_>resolve_addr() compat versions to return link addresses instead of pre-calculated data. BPF changes: Raw bpf writes occupied _two_ cases: AF_UNSPEC and pseudo_AF_HDRCMPLT. Despite the naming, both of there have ther header "complete". The only difference is that interface source mac has to be filled by OS for AF_UNSPEC (controlled via BIOCGHDRCMPLT). This logic has to stay inside BPF and not pollute if_output() routines. Convert BPF to pass prepend data via new 'struct route' mechanism. Note that it does not change non-optimized if_output(): ro_prepend handling is purely optional. Side note: hackish pseudo_AF_HDRCMPLT is supported for ethernet and FDDI. It is not needed for ethernet anymore. The only remaining FDDI user is dev/pdq mostly untouched since 2007. FDDI support was eliminated from OpenBSD in 2013 (sys/net/if_fddisubr.c rev 1.65). Flowtable changes: Flowtable violates layering by saving (and not correctly managing) rtes/lles. Instead of passing lle pointer, pass pointer to pre-calculated header data from that lle. Differential Revision: https://reviews.freebsd.org/D4102
2015-12-31 05:03:27 +00:00
static int if_requestencap_default(struct ifnet *, struct if_encap_req *);
static void if_route(struct ifnet *, int flag, int fam);
static int if_setflag(struct ifnet *, int, int, int *, int);
static int if_transmit(struct ifnet *ifp, struct mbuf *m);
static void if_unroute(struct ifnet *, int flag, int fam);
static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
static void do_link_state_change(void *, int);
static int if_getgroup(struct ifgroupreq *, struct ifnet *);
static int if_getgroupmembers(struct ifgroupreq *);
static void if_delgroups(struct ifnet *);
static void if_attach_internal(struct ifnet *, int, struct if_clone *);
static int if_detach_internal(struct ifnet *, int, struct if_clone **);
#ifdef VIMAGE
static void if_vmove(struct ifnet *, struct vnet *);
#endif
#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 *);
#endif
/* ipsec helper hooks */
VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]);
VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]);
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);
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
/* Table of ifnet by index. */
VNET_DEFINE(struct ifnet **, 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
#define V_if_indexlim VNET(if_indexlim)
#define V_ifindex_table VNET(ifindex_table)
/*
* 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;
RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE);
struct sx ifnet_sxlock;
SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE);
/*
* 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)
static if_com_alloc_t *if_com_alloc[256];
static if_com_free_t *if_com_free[256];
static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
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 *
ifnet_byindex_locked(u_short idx)
{
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);
if (V_ifindex_table[idx] == IFNET_HOLD)
return (NULL);
return (V_ifindex_table[idx]);
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
}
struct ifnet *
ifnet_byindex(u_short idx)
{
struct ifnet *ifp;
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);
IFNET_RUNLOCK_NOSLEEP();
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)
{
struct ifnet *ifp;
IFNET_RLOCK_NOSLEEP();
ifp = ifnet_byindex_locked(idx);
if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
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);
IFNET_RUNLOCK_NOSLEEP();
return (ifp);
}
/*
* Allocate an ifindex array entry; return 0 on success or an error on
* failure.
*/
static u_short
ifindex_alloc(void)
{
u_short idx;
IFNET_WLOCK_ASSERT();
retry:
/*
* Try to find an empty slot below V_if_index. If we fail, take the
* next slot.
*/
for (idx = 1; idx <= V_if_index; idx++) {
if (V_ifindex_table[idx] == NULL)
break;
}
/* Catch if_index overflow. */
if (idx >= V_if_indexlim) {
if_grow();
goto retry;
}
if (idx > V_if_index)
V_if_index = idx;
return (idx);
}
static void
ifindex_free_locked(u_short idx)
{
IFNET_WLOCK_ASSERT();
V_ifindex_table[idx] = NULL;
while (V_if_index > 0 &&
V_ifindex_table[V_if_index] == NULL)
V_if_index--;
}
static void
ifindex_free(u_short idx)
{
IFNET_WLOCK();
ifindex_free_locked(idx);
IFNET_WUNLOCK();
}
static void
ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
{
IFNET_WLOCK_ASSERT();
V_ifindex_table[idx] = ifp;
}
static void
ifnet_setbyindex(u_short idx, struct ifnet *ifp)
{
IFNET_WLOCK();
ifnet_setbyindex_locked(idx, ifp);
IFNET_WUNLOCK();
}
struct ifaddr *
ifaddr_byindex(u_short idx)
{
struct ifnet *ifp;
struct ifaddr *ifa = NULL;
IFNET_RLOCK_NOSLEEP();
ifp = ifnet_byindex_locked(idx);
if (ifp != NULL && (ifa = ifp->if_addr) != NULL)
ifa_ref(ifa);
IFNET_RUNLOCK_NOSLEEP();
return (ifa);
}
1994-05-24 10:09:53 +00:00
/*
* Network interface utility routines.
*
* Routines with ifa_ifwith* names take sockaddr *'s as
* parameters.
*/
static void
vnet_if_init(const void *unused __unused)
{
TAILQ_INIT(&V_ifnet);
TAILQ_INIT(&V_ifg_head);
IFNET_WLOCK();
if_grow(); /* create initial table */
IFNET_WUNLOCK();
vnet_if_clone_init();
}
VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
NULL);
#ifdef VIMAGE
static void
vnet_if_uninit(const void *unused __unused)
{
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));
free((caddr_t)V_ifindex_table, M_IFNET);
}
VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
vnet_if_uninit, NULL);
static void
vnet_if_return(const void *unused __unused)
{
struct ifnet *ifp, *nifp;
/* Return all inherited interfaces to their parent vnets. */
TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
if (ifp->if_home_vnet != ifp->if_vnet)
if_vmove(ifp, ifp->if_home_vnet);
}
}
VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY,
vnet_if_return, NULL);
#endif
static void
if_grow(void)
{
int oldlim;
u_int n;
struct ifnet **e;
IFNET_WLOCK_ASSERT();
oldlim = V_if_indexlim;
IFNET_WUNLOCK();
n = (oldlim << 1) * sizeof(*e);
e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
IFNET_WLOCK();
if (V_if_indexlim != oldlim) {
free(e, M_IFNET);
return;
}
if (V_ifindex_table != NULL) {
memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
free((caddr_t)V_ifindex_table, M_IFNET);
}
V_if_indexlim <<= 1;
V_ifindex_table = e;
}
/*
* 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.
*/
struct ifnet *
if_alloc(u_char type)
{
struct ifnet *ifp;
u_short idx;
ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
IFNET_WLOCK();
idx = ifindex_alloc();
ifnet_setbyindex_locked(idx, IFNET_HOLD);
IFNET_WUNLOCK();
ifp->if_index = idx;
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;
if (if_com_alloc[type] != NULL) {
ifp->if_l2com = if_com_alloc[type](type, ifp);
if (ifp->if_l2com == NULL) {
free(ifp, M_IFNET);
ifindex_free(idx);
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
IF_ADDR_LOCK_INIT(ifp);
TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
ifp->if_afdata_initialized = 0;
IF_AFDATA_LOCK_INIT(ifp);
TAILQ_INIT(&ifp->if_addrhead);
TAILQ_INIT(&ifp->if_multiaddrs);
TAILQ_INIT(&ifp->if_groups);
#ifdef MAC
mac_ifnet_init(ifp);
#endif
ifq_init(&ifp->if_snd, 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
refcount_init(&ifp->if_refcount, 1); /* Index reference. */
for (int i = 0; i < IFCOUNTERS; i++)
ifp->if_counters[i] = counter_u64_alloc(M_WAITOK);
ifp->if_get_counter = if_get_counter_default;
ifnet_setbyindex(ifp->if_index, ifp);
return (ifp);
}
/*
* Do the actual work of freeing a struct ifnet, and layer 2 common
* structure. This call is made when the last reference to an
* interface is released.
*/
static void
if_free_internal(struct ifnet *ifp)
{
KASSERT((ifp->if_flags & IFF_DYING),
("if_free_internal: interface not dying"));
if (if_com_free[ifp->if_alloctype] != NULL)
if_com_free[ifp->if_alloctype](ifp->if_l2com,
ifp->if_alloctype);
#ifdef MAC
mac_ifnet_destroy(ifp);
#endif /* MAC */
if (ifp->if_description != NULL)
free(ifp->if_description, M_IFDESCR);
IF_AFDATA_DESTROY(ifp);
IF_ADDR_LOCK_DESTROY(ifp);
ifq_delete(&ifp->if_snd);
for (int i = 0; i < IFCOUNTERS; i++)
counter_u64_free(ifp->if_counters[i]);
free(ifp, M_IFNET);
}
/*
* Deregister an interface and free the associated storage.
*/
void
if_free(struct ifnet *ifp)
{
ifp->if_flags |= IFF_DYING; /* XXX: Locking */
CURVNET_SET_QUIET(ifp->if_vnet);
IFNET_WLOCK();
KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
("%s: freeing unallocated ifnet", ifp->if_xname));
ifindex_free_locked(ifp->if_index);
IFNET_WUNLOCK();
if (refcount_release(&ifp->if_refcount))
if_free_internal(ifp);
CURVNET_RESTORE();
}
/*
* 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)
{
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
}
void
ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
{
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
ifq_delete(struct ifaltq *ifq)
{
mtx_destroy(&ifq->ifq_mtx);
}
1994-05-24 10:09:53 +00:00
/*
* Perform generic interface initialization tasks and attach the interface
* 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.
*
* Note that if_detach_internal() removes group membership unconditionally
* even when vmove flag is set, and if_attach_internal() adds only IFG_ALL.
* Thus, when if_vmove() is applied to a cloned interface, group membership
* is lost while a cloned one always joins a group whose name is
* ifc->ifc_name. To recover this after if_detach_internal() and
* if_attach_internal(), the cloner should be specified to
* if_attach_internal() via ifc. If it is non-NULL, if_attach_internal()
* attempts to join a group whose name is ifc->ifc_name.
*
* 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
if_attach(struct ifnet *ifp)
{
if_attach_internal(ifp, 0, NULL);
}
/*
* Compute the least common TSO limit.
*/
void
if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax)
{
/*
* 1) If there is no limit currently, take the limit from
* the network adapter.
*
* 2) If the network adapter has a limit below the current
* limit, apply it.
*/
if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
pmax->tsomaxbytes = ifp->if_hw_tsomax;
}
if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
}
if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
}
}
/*
* Update TSO limit of a network adapter.
*
* Returns zero if no change. Else non-zero.
*/
int
if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax)
{
int retval = 0;
if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
ifp->if_hw_tsomax = pmax->tsomaxbytes;
retval++;
}
if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
retval++;
}
if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
retval++;
}
return (retval);
}
static void
if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc)
1994-05-24 10:09:53 +00:00
{
unsigned socksize, ifasize;
int namelen, masklen;
struct sockaddr_dl *sdl;
struct ifaddr *ifa;
1994-05-24 10:09:53 +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;
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
if_addgroup(ifp, IFG_ALL);
/* Restore group membership for cloned interfaces. */
if (vmove && ifc != NULL)
if_clone_addgroup(ifp, ifc);
getmicrotime(&ifp->if_lastchange);
ifp->if_epoch = time_uptime;
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;
}
if (ifp->if_input == NULL)
ifp->if_input = if_input_default;
Implement interface link header precomputation API. Add if_requestencap() interface method which is capable of calculating various link headers for given interface. Right now there is support for INET/INET6/ARP llheader calculation (IFENCAP_LL type request). Other types are planned to support more complex calculation (L2 multipath lagg nexthops, tunnel encap nexthops, etc..). Reshape 'struct route' to be able to pass additional data (with is length) to prepend to mbuf. These two changes permits routing code to pass pre-calculated nexthop data (like L2 header for route w/gateway) down to the stack eliminating the need for other lookups. It also brings us closer to more complex scenarios like transparently handling MPLS nexthops and tunnel interfaces. Last, but not least, it removes layering violation introduced by flowtable code (ro_lle) and simplifies handling of existing if_output consumers. ARP/ND changes: Make arp/ndp stack pre-calculate link header upon installing/updating lle record. Interface link address change are handled by re-calculating headers for all lles based on if_lladdr event. After these changes, arpresolve()/nd6_resolve() returns full pre-calculated header for supported interfaces thus simplifying if_output(). Move these lookups to separate ether_resolve_addr() function which ether returs error or fully-prepared link header. Add <arp|nd6_>resolve_addr() compat versions to return link addresses instead of pre-calculated data. BPF changes: Raw bpf writes occupied _two_ cases: AF_UNSPEC and pseudo_AF_HDRCMPLT. Despite the naming, both of there have ther header "complete". The only difference is that interface source mac has to be filled by OS for AF_UNSPEC (controlled via BIOCGHDRCMPLT). This logic has to stay inside BPF and not pollute if_output() routines. Convert BPF to pass prepend data via new 'struct route' mechanism. Note that it does not change non-optimized if_output(): ro_prepend handling is purely optional. Side note: hackish pseudo_AF_HDRCMPLT is supported for ethernet and FDDI. It is not needed for ethernet anymore. The only remaining FDDI user is dev/pdq mostly untouched since 2007. FDDI support was eliminated from OpenBSD in 2013 (sys/net/if_fddisubr.c rev 1.65). Flowtable changes: Flowtable violates layering by saving (and not correctly managing) rtes/lles. Instead of passing lle pointer, pass pointer to pre-calculated header data from that lle. Differential Revision: https://reviews.freebsd.org/D4102
2015-12-31 05:03:27 +00:00
if (ifp->if_requestencap == NULL)
ifp->if_requestencap = if_requestencap_default;
if (!vmove) {
#ifdef MAC
mac_ifnet_create(ifp);
#endif
/*
* 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 = ifa_alloc(ifasize, M_WAITOK);
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;
#if defined(INET) || defined(INET6)
/* Use defaults for TSO, if nothing is set */
if (ifp->if_hw_tsomax == 0 &&
ifp->if_hw_tsomaxsegcount == 0 &&
ifp->if_hw_tsomaxsegsize == 0) {
/*
* The TSO defaults needs to be such that an
* NFS mbuf list of 35 mbufs totalling just
* below 64K works and that a chain of mbufs
* can be defragged into at most 32 segments:
*/
ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
(ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
ifp->if_hw_tsomaxsegcount = 35;
ifp->if_hw_tsomaxsegsize = 2048; /* 2K */
/* XXX some drivers set IFCAP_TSO after ethernet attach */
if (ifp->if_capabilities & IFCAP_TSO) {
if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
ifp->if_hw_tsomax,
ifp->if_hw_tsomaxsegcount,
ifp->if_hw_tsomaxsegsize);
}
}
#endif
}
#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
IFNET_WLOCK();
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
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
IFNET_WUNLOCK();
if (domain_init_status >= 2)
if_attachdomain1(ifp);
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);
/* Announce the interface. */
rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1994-05-24 10:09:53 +00:00
}
static void
if_attachdomain(void *dummy)
{
struct ifnet *ifp;
TAILQ_FOREACH(ifp, &V_ifnet, if_link)
if_attachdomain1(ifp);
}
SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
if_attachdomain, NULL);
static void
if_attachdomain1(struct ifnet *ifp)
{
struct domain *dp;
/*
* Since dp->dom_ifattach calls malloc() with M_WAITOK, we
* cannot lock ifp->if_afdata initialization, entirely.
*/
IF_AFDATA_LOCK(ifp);
if (ifp->if_afdata_initialized >= domain_init_status) {
IF_AFDATA_UNLOCK(ifp);
log(LOG_WARNING, "%s called more than once on %s\n",
__func__, ifp->if_xname);
return;
}
ifp->if_afdata_initialized = domain_init_status;
IF_AFDATA_UNLOCK(ifp);
/* 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);
}
}
/*
* Remove any unicast or broadcast network addresses from an interface.
*/
void
if_purgeaddrs(struct ifnet *ifp)
{
struct ifaddr *ifa, *next;
/* XXX cannot hold IF_ADDR_WLOCK over called functions. */
TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
if (ifa->ifa_addr->sa_family == AF_LINK)
continue;
#ifdef INET
/* XXX: Ugly!! ad hoc just for INET */
if (ifa->ifa_addr->sa_family == AF_INET) {
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
if (ifa->ifa_addr->sa_family == AF_INET6) {
in6_purgeaddr(ifa);
/* ifp_addrhead is already updated */
continue;
}
#endif /* INET6 */
IF_ADDR_WLOCK(ifp);
TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
IF_ADDR_WUNLOCK(ifp);
ifa_free(ifa);
}
}
/*
* Remove any multicast network addresses from an interface when an ifnet
* is going away.
*/
static void
if_purgemaddrs(struct ifnet *ifp)
{
struct ifmultiaddr *ifma;
struct ifmultiaddr *next;
IF_ADDR_WLOCK(ifp);
TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
if_delmulti_locked(ifp, ifma, 1);
IF_ADDR_WUNLOCK(ifp);
}
/*
* 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.
*
* XXXRW: There are some significant questions about event ordering, and
* how to prevent things from starting to use the interface during detach.
*/
void
if_detach(struct ifnet *ifp)
{
CURVNET_SET_QUIET(ifp->if_vnet);
if_detach_internal(ifp, 0, NULL);
CURVNET_RESTORE();
}
static int
if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp)
{
struct ifaddr *ifa;
int i;
struct domain *dp;
struct ifnet *iter;
int found = 0;
IFNET_WLOCK();
TAILQ_FOREACH(iter, &V_ifnet, if_link)
if (iter == ifp) {
TAILQ_REMOVE(&V_ifnet, ifp, if_link);
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)
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
IFNET_WUNLOCK();
if (!found) {
/*
* While we would want to panic here, we cannot
* guarantee that the interface is indeed still on
* the list given we don't hold locks all the way.
*/
return (ENOENT);
#if 0
if (vmove)
panic("%s: ifp=%p not on the ifnet tailq %p",
__func__, ifp, &V_ifnet);
else
return; /* XXX this should panic as well? */
#endif
}
/* Check if this is a cloned interface or not. */
if (vmove && ifcp != NULL)
*ifcp = if_clone_findifc(ifp);
/*
* Remove/wait for pending events.
*/
taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
/*
* Remove routes and flush queues.
*/
if_down(ifp);
#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
if_purgeaddrs(ifp);
#ifdef INET
in_ifdetach(ifp);
#endif
#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
if_purgemaddrs(ifp);
/* Announce that the interface is gone. */
rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
if (IS_DEFAULT_VNET(curvnet))
devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
if (!vmove) {
/*
* Prevent further calls into the device driver via ifnet.
*/
if_dead(ifp);
/*
* Remove link ifaddr pointer and maybe decrement if_index.
* Clean up all addresses.
*/
ifp->if_addr = NULL;
/* We can now free link ifaddr. */
IF_ADDR_WLOCK(ifp);
if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
ifa = TAILQ_FIRST(&ifp->if_addrhead);
TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
IF_ADDR_WUNLOCK(ifp);
ifa_free(ifa);
} else
IF_ADDR_WUNLOCK(ifp);
}
rt_flushifroutes(ifp);
if_delgroups(ifp);
/*
* 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.
*/
IF_AFDATA_LOCK(ifp);
i = ifp->if_afdata_initialized;
ifp->if_afdata_initialized = 0;
IF_AFDATA_UNLOCK(ifp);
for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
(*dp->dom_ifdetach)(ifp,
ifp->if_afdata[dp->dom_family]);
}
return (0);
}
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
/*
* 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)
{
struct if_clone *ifc;
u_int bif_dlt, bif_hdrlen;
int rc;
/*
* if_detach_internal() will call the eventhandler to notify
* interface departure. That will detach if_bpf. We need to
* safe the dlt and hdrlen so we can re-attach it later.
*/
bpf_get_bp_params(ifp->if_bpf, &bif_dlt, &bif_hdrlen);
/*
* Detach from current vnet, but preserve LLADDR info, do not
* mark as dead etc. so that the ifnet can be reattached later.
* If we cannot find it, we lost the race to someone else.
*/
rc = if_detach_internal(ifp, 1, &ifc);
if (rc != 0)
return;
/*
* 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.
*/
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();
ifindex_free_locked(ifp->if_index);
IFNET_WUNLOCK();
/*
* Perform interface-specific reassignment tasks, if provided by
* the driver.
*/
if (ifp->if_reassign != NULL)
ifp->if_reassign(ifp, new_vnet, NULL);
/*
* Switch to the context of the target vnet.
*/
CURVNET_SET_QUIET(new_vnet);
IFNET_WLOCK();
ifp->if_index = ifindex_alloc();
ifnet_setbyindex_locked(ifp->if_index, ifp);
IFNET_WUNLOCK();
if_attach_internal(ifp, 1, ifc);
if (ifp->if_bpf == NULL)
bpfattach(ifp, bif_dlt, bif_hdrlen);
CURVNET_RESTORE();
}
/*
* 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. */
CURVNET_SET_QUIET(pr->pr_vnet);
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);
}
#endif /* VIMAGE */
/*
* 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;
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
int new = 0;
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);
}
TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
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);
TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
new = 1;
}
ifg->ifg_refcnt++;
ifgl->ifgl_group = ifg;
ifgm->ifgm_ifp = ifp;
IF_ADDR_WLOCK(ifp);
TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
IF_ADDR_WUNLOCK(ifp);
IFNET_WUNLOCK();
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
if (new)
EVENTHANDLER_INVOKE(group_attach_event, ifg);
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_WLOCK(ifp);
TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
IF_ADDR_WUNLOCK(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);
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
IFNET_WUNLOCK();
EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
free(ifgl->ifgl_group, M_TEMP);
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
} else
IFNET_WUNLOCK();
free(ifgl, M_TEMP);
EVENTHANDLER_INVOKE(group_change_event, groupname);
return (0);
}
/*
* 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_WLOCK(ifp);
TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
IF_ADDR_WUNLOCK(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);
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
IFNET_WUNLOCK();
EVENTHANDLER_INVOKE(group_detach_event,
ifgl->ifgl_group);
free(ifgl->ifgl_group, M_TEMP);
Merge the projects/pf/head branch, that was worked on for last six months, into head. The most significant achievements in the new code: o Fine grained locking, thus much better performance. o Fixes to many problems in pf, that were specific to FreeBSD port. New code doesn't have that many ifdefs and much less OpenBSDisms, thus is more attractive to our developers. Those interested in details, can browse through SVN log of the projects/pf/head branch. And for reference, here is exact list of revisions merged: r232043, r232044, r232062, r232148, r232149, r232150, r232298, r232330, r232332, r232340, r232386, r232390, r232391, r232605, r232655, r232656, r232661, r232662, r232663, r232664, r232673, r232691, r233309, r233782, r233829, r233830, r233834, r233835, r233836, r233865, r233866, r233868, r233873, r234056, r234096, r234100, r234108, r234175, r234187, r234223, r234271, r234272, r234282, r234307, r234309, r234382, r234384, r234456, r234486, r234606, r234640, r234641, r234642, r234644, r234651, r235505, r235506, r235535, r235605, r235606, r235826, r235991, r235993, r236168, r236173, r236179, r236180, r236181, r236186, r236223, r236227, r236230, r236252, r236254, r236298, r236299, r236300, r236301, r236397, r236398, r236399, r236499, r236512, r236513, r236525, r236526, r236545, r236548, r236553, r236554, r236556, r236557, r236561, r236570, r236630, r236672, r236673, r236679, r236706, r236710, r236718, r237154, r237155, r237169, r237314, r237363, r237364, r237368, r237369, r237376, r237440, r237442, r237751, r237783, r237784, r237785, r237788, r237791, r238421, r238522, r238523, r238524, r238525, r239173, r239186, r239644, r239652, r239661, r239773, r240125, r240130, r240131, r240136, r240186, r240196, r240212. I'd like to thank people who participated in early testing: Tested by: Florian Smeets <flo freebsd.org> Tested by: Chekaluk Vitaly <artemrts ukr.net> Tested by: Ben Wilber <ben desync.com> Tested by: Ian FREISLICH <ianf cloudseed.co.za>
2012-09-08 06:41:54 +00:00
} else
IFNET_WUNLOCK();
free(ifgl, M_TEMP);
EVENTHANDLER_INVOKE(group_change_event, groupname);
IFNET_WLOCK();
}
IFNET_WUNLOCK();
}
/*
* 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_RLOCK(ifp);
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
ifgr->ifgr_len += sizeof(struct ifg_req);
IF_ADDR_RUNLOCK(ifp);
return (0);
}
len = ifgr->ifgr_len;
ifgp = ifgr->ifgr_groups;
/* XXX: wire */
IF_ADDR_RLOCK(ifp);
TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
if (len < sizeof(ifgrq)) {
IF_ADDR_RUNLOCK(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_RUNLOCK(ifp);
return (error);
}
len -= sizeof(ifgrq);
ifgp++;
}
IF_ADDR_RUNLOCK(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();
TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
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);
}
/*
* Return counter values from counter(9)s stored in ifnet.
*/
uint64_t
if_get_counter_default(struct ifnet *ifp, ift_counter cnt)
{
KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
return (counter_u64_fetch(ifp->if_counters[cnt]));
}
/*
* Increase an ifnet counter. Usually used for counters shared
* between the stack and a driver, but function supports them all.
*/
void
if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc)
{
KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
counter_u64_add(ifp->if_counters[cnt], inc);
}
/*
* Copy data from ifnet to userland API structure if_data.
*/
void
if_data_copy(struct ifnet *ifp, struct if_data *ifd)
{
ifd->ifi_type = ifp->if_type;
ifd->ifi_physical = 0;
ifd->ifi_addrlen = ifp->if_addrlen;
ifd->ifi_hdrlen = ifp->if_hdrlen;
ifd->ifi_link_state = ifp->if_link_state;
ifd->ifi_vhid = 0;
ifd->ifi_datalen = sizeof(struct if_data);
ifd->ifi_mtu = ifp->if_mtu;
ifd->ifi_metric = ifp->if_metric;
ifd->ifi_baudrate = ifp->if_baudrate;
ifd->ifi_hwassist = ifp->if_hwassist;
ifd->ifi_epoch = ifp->if_epoch;
ifd->ifi_lastchange = ifp->if_lastchange;
ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS);
ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS);
ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS);
ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS);
ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS);
ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES);
ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES);
ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS);
ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS);
ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS);
ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS);
ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO);
}
/*
* 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_RLOCK(ifp);
}
void
if_addr_runlock(struct ifnet *ifp)
{
IF_ADDR_RUNLOCK(ifp);
}
void
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
if_maddr_rlock(if_t ifp)
{
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
IF_ADDR_RLOCK((struct ifnet *)ifp);
}
void
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
if_maddr_runlock(if_t ifp)
{
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
IF_ADDR_RUNLOCK((struct ifnet *)ifp);
}
/*
* Initialization, destruction and refcounting functions for ifaddrs.
*/
struct ifaddr *
ifa_alloc(size_t size, int flags)
{
struct ifaddr *ifa;
KASSERT(size >= sizeof(struct ifaddr),
("%s: invalid size %zu", __func__, size));
ifa = malloc(size, M_IFADDR, M_ZERO | flags);
if (ifa == NULL)
return (NULL);
if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL)
goto fail;
if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL)
goto fail;
if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL)
goto fail;
if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL)
goto fail;
refcount_init(&ifa->ifa_refcnt, 1);
return (ifa);
fail:
/* free(NULL) is okay */
counter_u64_free(ifa->ifa_opackets);
counter_u64_free(ifa->ifa_ipackets);
counter_u64_free(ifa->ifa_obytes);
counter_u64_free(ifa->ifa_ibytes);
free(ifa, M_IFADDR);
return (NULL);
}
void
ifa_ref(struct ifaddr *ifa)
{
refcount_acquire(&ifa->ifa_refcnt);
}
void
ifa_free(struct ifaddr *ifa)
{
if (refcount_release(&ifa->ifa_refcnt)) {
counter_u64_free(ifa->ifa_opackets);
counter_u64_free(ifa->ifa_ipackets);
counter_u64_free(ifa->ifa_obytes);
counter_u64_free(ifa->ifa_ibytes);
free(ifa, M_IFADDR);
}
}
static int
ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa,
struct sockaddr *ia)
{
int error;
struct rt_addrinfo info;
struct sockaddr_dl null_sdl;
struct ifnet *ifp;
ifp = ifa->ifa_ifp;
bzero(&info, sizeof(info));
if (cmd != RTM_DELETE)
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;
link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type);
error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib);
if (error != 0)
log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n",
__func__, otype, if_name(ifp), error);
return (error);
}
int
ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
{
return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia));
}
int
ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
{
return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia));
}
int
ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
{
return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia));
}
/*
* 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_dl_equal(a1, a2) \
((((const struct sockaddr_dl *)(a1))->sdl_len == \
((const struct sockaddr_dl *)(a2))->sdl_len) && \
(bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \
CLLADDR((const struct sockaddr_dl *)(a2)), \
((const 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*/
static struct ifaddr *
ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref)
1994-05-24 10:09:53 +00:00
{
struct ifnet *ifp;
struct ifaddr *ifa;
1994-05-24 10:09:53 +00:00
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
IF_ADDR_RLOCK(ifp);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
if (ifa->ifa_addr->sa_family != addr->sa_family)
continue;
if (sa_equal(addr, ifa->ifa_addr)) {
if (getref)
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
goto done;
}
/* IP6 doesn't have broadcast */
if ((ifp->if_flags & IFF_BROADCAST) &&
ifa->ifa_broadaddr &&
ifa->ifa_broadaddr->sa_len != 0 &&
sa_equal(ifa->ifa_broadaddr, addr)) {
if (getref)
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
goto done;
}
}
IF_ADDR_RUNLOCK(ifp);
}
ifa = NULL;
done:
IFNET_RUNLOCK_NOSLEEP();
return (ifa);
1994-05-24 10:09:53 +00:00
}
struct ifaddr *
ifa_ifwithaddr(const struct sockaddr *addr)
{
return (ifa_ifwithaddr_internal(addr, 1));
}
int
ifa_ifwithaddr_check(const struct sockaddr *addr)
{
return (ifa_ifwithaddr_internal(addr, 0) != NULL);
}
/*
* Locate an interface based on the broadcast address.
*/
/* ARGSUSED */
struct ifaddr *
ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum)
{
struct ifnet *ifp;
struct ifaddr *ifa;
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
continue;
IF_ADDR_RLOCK(ifp);
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 &&
sa_equal(ifa->ifa_broadaddr, addr)) {
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
goto done;
}
}
IF_ADDR_RUNLOCK(ifp);
}
ifa = NULL;
done:
IFNET_RUNLOCK_NOSLEEP();
return (ifa);
}
1994-05-24 10:09:53 +00:00
/*
* Locate the point to point interface with a given destination address.
*/
/*ARGSUSED*/
struct ifaddr *
ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum)
1994-05-24 10:09:53 +00:00
{
struct ifnet *ifp;
struct ifaddr *ifa;
1994-05-24 10:09:53 +00:00
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
continue;
if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
Fix subnet and default routes on different FIBs on the same subnet. These two bugs are closely related. The root cause is that ifa_ifwithnet does not consider FIBs when searching for an interface address. sys/net/if_var.h sys/net/if.c Add a fib argument to ifa_ifwithnet and ifa_ifwithdstadddr. Those functions will only return an address whose interface fib equals the argument. sys/net/route.c Update calls to ifa_ifwithnet and ifa_ifwithdstaddr with fib arguments. sys/netinet/in.c Update in_addprefix to consider the interface fib when adding prefixes. This will prevent it from not adding a subnet route when one already exists on a different fib. sys/net/rtsock.c sys/netinet/in_pcb.c sys/netinet/ip_output.c sys/netinet/ip_options.c sys/netinet6/nd6.c Add RT_DEFAULT_FIB arguments to ifa_ifwithdstaddr and ifa_ifwithnet. In some cases it there wasn't a clear specific fib number to use. In others, I was unable to test those functions so I chose RT_DEFAULT_FIB to minimize divergence from current behavior. I will fix some of the latter changes along with PR kern/187553. tests/sys/netinet/fibs_test.sh tests/sys/netinet/udp_dontroute.c tests/sys/netinet/Makefile Revert r263738. The udp_dontroute test was right all along. However, bugs kern/187550 and kern/187553 cancelled each other out when it came to this test. Because of kern/187553, ifa_ifwithnet searched the default fib instead of the requested one, but because of kern/187550, there was an applicable subnet route on the default fib. The new test added in r263738 doesn't work right, however. I can verify with dtrace that ifa_ifwithnet returned the wrong address before I applied this commit, but route(8) miraculously found the correct interface to use anyway. I don't know how. Clear expected failure messages for kern/187550 and kern/187552. PR: kern/187550 PR: kern/187552 Reviewed by: melifaro MFC after: 3 weeks Sponsored by: Spectra Logic
2014-04-24 23:56:56 +00:00
continue;
IF_ADDR_RLOCK(ifp);
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;
if (ifa->ifa_dstaddr != NULL &&
sa_equal(addr, ifa->ifa_dstaddr)) {
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
goto done;
}
}
IF_ADDR_RUNLOCK(ifp);
1994-05-24 10:09:53 +00:00
}
ifa = NULL;
done:
IFNET_RUNLOCK_NOSLEEP();
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 *
ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum)
1994-05-24 10:09:53 +00:00
{
struct ifnet *ifp;
struct ifaddr *ifa;
struct ifaddr *ifa_maybe = NULL;
1994-05-24 10:09:53 +00:00
u_int af = addr->sa_family;
const char *addr_data = addr->sa_data, *cplim;
1994-05-24 10:09:53 +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) {
const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr;
if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
return (ifaddr_byindex(sdl->sdl_index));
1994-05-24 10:09:53 +00:00
}
/*
* Scan though each interface, looking for ones that have addresses
Fix subnet and default routes on different FIBs on the same subnet. These two bugs are closely related. The root cause is that ifa_ifwithnet does not consider FIBs when searching for an interface address. sys/net/if_var.h sys/net/if.c Add a fib argument to ifa_ifwithnet and ifa_ifwithdstadddr. Those functions will only return an address whose interface fib equals the argument. sys/net/route.c Update calls to ifa_ifwithnet and ifa_ifwithdstaddr with fib arguments. sys/netinet/in.c Update in_addprefix to consider the interface fib when adding prefixes. This will prevent it from not adding a subnet route when one already exists on a different fib. sys/net/rtsock.c sys/netinet/in_pcb.c sys/netinet/ip_output.c sys/netinet/ip_options.c sys/netinet6/nd6.c Add RT_DEFAULT_FIB arguments to ifa_ifwithdstaddr and ifa_ifwithnet. In some cases it there wasn't a clear specific fib number to use. In others, I was unable to test those functions so I chose RT_DEFAULT_FIB to minimize divergence from current behavior. I will fix some of the latter changes along with PR kern/187553. tests/sys/netinet/fibs_test.sh tests/sys/netinet/udp_dontroute.c tests/sys/netinet/Makefile Revert r263738. The udp_dontroute test was right all along. However, bugs kern/187550 and kern/187553 cancelled each other out when it came to this test. Because of kern/187553, ifa_ifwithnet searched the default fib instead of the requested one, but because of kern/187550, there was an applicable subnet route on the default fib. The new test added in r263738 doesn't work right, however. I can verify with dtrace that ifa_ifwithnet returned the wrong address before I applied this commit, but route(8) miraculously found the correct interface to use anyway. I don't know how. Clear expected failure messages for kern/187550 and kern/187552. PR: kern/187550 PR: kern/187552 Reviewed by: melifaro MFC after: 3 weeks Sponsored by: Spectra Logic
2014-04-24 23:56:56 +00:00
* in this address family and the requested fib. Maintain a reference
* on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
* kept it stable when we move onto the next interface.
*/
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
Fix subnet and default routes on different FIBs on the same subnet. These two bugs are closely related. The root cause is that ifa_ifwithnet does not consider FIBs when searching for an interface address. sys/net/if_var.h sys/net/if.c Add a fib argument to ifa_ifwithnet and ifa_ifwithdstadddr. Those functions will only return an address whose interface fib equals the argument. sys/net/route.c Update calls to ifa_ifwithnet and ifa_ifwithdstaddr with fib arguments. sys/netinet/in.c Update in_addprefix to consider the interface fib when adding prefixes. This will prevent it from not adding a subnet route when one already exists on a different fib. sys/net/rtsock.c sys/netinet/in_pcb.c sys/netinet/ip_output.c sys/netinet/ip_options.c sys/netinet6/nd6.c Add RT_DEFAULT_FIB arguments to ifa_ifwithdstaddr and ifa_ifwithnet. In some cases it there wasn't a clear specific fib number to use. In others, I was unable to test those functions so I chose RT_DEFAULT_FIB to minimize divergence from current behavior. I will fix some of the latter changes along with PR kern/187553. tests/sys/netinet/fibs_test.sh tests/sys/netinet/udp_dontroute.c tests/sys/netinet/Makefile Revert r263738. The udp_dontroute test was right all along. However, bugs kern/187550 and kern/187553 cancelled each other out when it came to this test. Because of kern/187553, ifa_ifwithnet searched the default fib instead of the requested one, but because of kern/187550, there was an applicable subnet route on the default fib. The new test added in r263738 doesn't work right, however. I can verify with dtrace that ifa_ifwithnet returned the wrong address before I applied this commit, but route(8) miraculously found the correct interface to use anyway. I don't know how. Clear expected failure messages for kern/187550 and kern/187552. PR: kern/187550 PR: kern/187552 Reviewed by: melifaro MFC after: 3 weeks Sponsored by: Spectra Logic
2014-04-24 23:56:56 +00:00
continue;
IF_ADDR_RLOCK(ifp);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
const char *cp, *cp2, *cp3;
1994-05-24 10:09:53 +00:00
if (ifa->ifa_addr->sa_family != af)
next: continue;
if (af == AF_INET &&
ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
/*
* This is a bit broken as it doesn't
* take into account that the remote end may
* be a single node in the network we are
* looking for.
* The trouble is that we don't know the
* netmask for the remote end.
*/
if (ifa->ifa_dstaddr != NULL &&
sa_equal(addr, ifa->ifa_dstaddr)) {
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
goto done;
}
} else {
/*
* 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)
*/
if (ifa->ifa_netmask == 0)
continue;
cp = addr_data;
cp2 = ifa->ifa_addr->sa_data;
cp3 = ifa->ifa_netmask->sa_data;
cplim = ifa->ifa_netmask->sa_len
+ (char *)ifa->ifa_netmask;
while (cp3 < cplim)
if ((*cp++ ^ *cp2++) & *cp3++)
goto next; /* next address! */
/*
* If the netmask of what we just found
* is more specific than what we had before
* (if we had one), or if the virtual status
* of new prefix is better than of the old one,
* then remember the new one before continuing
* to search for an even better one.
*/
if (ifa_maybe == NULL ||
ifa_preferred(ifa_maybe, ifa) ||
rn_refines((caddr_t)ifa->ifa_netmask,
(caddr_t)ifa_maybe->ifa_netmask)) {
if (ifa_maybe != NULL)
ifa_free(ifa_maybe);
ifa_maybe = ifa;
ifa_ref(ifa_maybe);
}
}
}
IF_ADDR_RUNLOCK(ifp);
}
ifa = ifa_maybe;
ifa_maybe = NULL;
done:
IFNET_RUNLOCK_NOSLEEP();
if (ifa_maybe != NULL)
ifa_free(ifa_maybe);
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 *
ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1994-05-24 10:09:53 +00:00
{
struct ifaddr *ifa;
const char *cp, *cp2, *cp3;
char *cplim;
struct ifaddr *ifa_maybe = NULL;
1994-05-24 10:09:53 +00:00
u_int af = addr->sa_family;
if (af >= AF_MAX)
return (NULL);
IF_ADDR_RLOCK(ifp);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1994-05-24 10:09:53 +00:00
if (ifa->ifa_addr->sa_family != af)
continue;
if (ifa_maybe == NULL)
ifa_maybe = ifa;
1994-05-24 10:09:53 +00:00
if (ifa->ifa_netmask == 0) {
if (sa_equal(addr, ifa->ifa_addr) ||
(ifa->ifa_dstaddr &&
sa_equal(addr, ifa->ifa_dstaddr)))
goto done;
1994-05-24 10:09:53 +00:00
continue;
}
if (ifp->if_flags & IFF_POINTOPOINT) {
if (sa_equal(addr, ifa->ifa_dstaddr))
goto done;
} else {
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)
goto done;
}
1994-05-24 10:09:53 +00:00
}
ifa = ifa_maybe;
done:
if (ifa != NULL)
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
return (ifa);
1994-05-24 10:09:53 +00:00
}
/*
* See whether new ifa is better than current one:
* 1) A non-virtual one is preferred over virtual.
* 2) A virtual in master state preferred over any other state.
*
* Used in several address selecting functions.
*/
int
ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
{
return (cur->ifa_carp && (!next->ifa_carp ||
((*carp_master_p)(next) && !(*carp_master_p)(cur))));
}
#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
link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1994-05-24 10:09:53 +00:00
{
struct ifaddr *ifa, *oifa;
1994-05-24 10:09:53 +00:00
struct sockaddr *dst;
struct ifnet *ifp;
if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == NULL) ||
((ifp = ifa->ifa_ifp) == NULL) || ((dst = rt_key(rt)) == NULL))
1994-05-24 10:09:53 +00:00
return;
ifa = ifaof_ifpforaddr(dst, ifp);
if (ifa) {
oifa = rt->rt_ifa;
1994-05-24 10:09:53 +00:00
rt->rt_ifa = ifa;
ifa_free(oifa);
1994-05-24 10:09:53 +00:00
if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
ifa->ifa_rtrequest(cmd, rt, info);
1994-05-24 10:09:53 +00:00
}
}
struct sockaddr_dl *
link_alloc_sdl(size_t size, int flags)
{
return (malloc(size, M_TEMP, flags));
}
void
link_free_sdl(struct sockaddr *sa)
{
free(sa, M_TEMP);
}
/*
* Fills in given sdl with interface basic info.
* Returns pointer to filled sdl.
*/
struct sockaddr_dl *
link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype)
{
struct sockaddr_dl *sdl;
sdl = (struct sockaddr_dl *)paddr;
memset(sdl, 0, sizeof(struct sockaddr_dl));
sdl->sdl_len = sizeof(struct sockaddr_dl);
sdl->sdl_family = AF_LINK;
sdl->sdl_index = ifp->if_index;
sdl->sdl_type = iftype;
return (sdl);
}
1994-05-24 10:09:53 +00:00
/*
* Mark an interface down and notify protocols of
* the transition.
*/
static void
if_unroute(struct ifnet *ifp, int flag, int fam)
1994-05-24 10:09:53 +00:00
{
struct ifaddr *ifa;
1994-05-24 10:09:53 +00:00
KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
ifp->if_flags &= ~flag;
getmicrotime(&ifp->if_lastchange);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
ifp->if_qflush(ifp);
if (ifp->if_carp)
(*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.
*/
static void
if_route(struct ifnet *ifp, int flag, int fam)
1994-05-24 10:09:53 +00:00
{
struct ifaddr *ifa;
1994-05-24 10:09:53 +00:00
KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
ifp->if_flags |= flag;
getmicrotime(&ifp->if_lastchange);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
pfctlinput(PRC_IFUP, ifa->ifa_addr);
if (ifp->if_carp)
(*carp_linkstate_p)(ifp);
1994-05-24 10:09:53 +00:00
rt_ifmsg(ifp);
#ifdef INET6
in6_if_up(ifp);
#endif
1994-05-24 10:09:53 +00:00
}
void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */
2006-01-30 13:45:15 +00:00
void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */
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 *);
/*
* 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().
*/
void
if_link_state_change(struct ifnet *ifp, int link_state)
{
/* Return if state hasn't changed. */
if (ifp->if_link_state == link_state)
return;
ifp->if_link_state = link_state;
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;
CURVNET_SET(ifp->if_vnet);
/* Notify that the link state has changed. */
rt_ifmsg(ifp);
2006-01-30 13:45:15 +00:00
if (ifp->if_vlantrunk != NULL)
(*vlan_link_state_p)(ifp);
if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
ifp->if_l2com != NULL)
(*ng_ether_link_state_p)(ifp, link_state);
if (ifp->if_carp)
(*carp_linkstate_p)(ifp);
if (ifp->if_bridge)
(*bridge_linkstate_p)(ifp);
if (ifp->if_lagg)
(*lagg_linkstate_p)(ifp, link_state);
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);
if (pending > 1)
if_printf(ifp, "%d link states coalesced\n", pending);
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" );
EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
CURVNET_RESTORE();
}
/*
* Mark an interface down and notify protocols of
* the transition.
*/
void
if_down(struct ifnet *ifp)
{
if_unroute(ifp, IFF_UP, AF_UNSPEC);
}
/*
* Mark an interface up and notify protocols of
* the transition.
*/
void
if_up(struct ifnet *ifp)
{
if_route(ifp, IFF_UP, AF_UNSPEC);
}
1994-05-24 10:09:53 +00:00
/*
* Flush an interface queue.
*/
void
if_qflush(struct ifnet *ifp)
1994-05-24 10:09:53 +00:00
{
struct mbuf *m, *n;
struct ifaltq *ifq;
ifq = &ifp->if_snd;
IFQ_LOCK(ifq);
#ifdef ALTQ
if (ALTQ_IS_ENABLED(ifq))
ALTQ_PURGE(ifq);
#endif
1994-05-24 10:09:53 +00:00
n = ifq->ifq_head;
while ((m = n) != NULL) {
2014-07-17 05:21:16 +00:00
n = m->m_nextpkt;
1994-05-24 10:09:53 +00:00
m_freem(m);
}
ifq->ifq_head = 0;
ifq->ifq_tail = 0;
ifq->ifq_len = 0;
IFQ_UNLOCK(ifq);
1994-05-24 10:09:53 +00:00
}
/*
* Map interface name to interface structure pointer, with or without
* returning a reference.
1994-05-24 10:09:53 +00:00
*/
struct ifnet *
ifunit_ref(const char *name)
{
struct ifnet *ifp;
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
!(ifp->if_flags & IFF_DYING))
break;
}
if (ifp != NULL)
if_ref(ifp);
IFNET_RUNLOCK_NOSLEEP();
return (ifp);
}
1994-05-24 10:09:53 +00:00
struct ifnet *
ifunit(const char *name)
1994-05-24 10:09:53 +00:00
{
struct ifnet *ifp;
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1994-05-24 10:09:53 +00:00
break;
}
IFNET_RUNLOCK_NOSLEEP();
1994-05-24 10:09:53 +00:00
return (ifp);
}
/*
* Hardware specific interface ioctls.
1994-05-24 10:09:53 +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
{
struct ifreq *ifr;
int error = 0;
int new_flags, temp_flags;
size_t namelen, onamelen;
size_t descrlen;
char *descrbuf, *odescrbuf;
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) {
case SIOCGIFINDEX:
ifr->ifr_index = ifp->if_index;
break;
1994-05-24 10:09:53 +00:00
case SIOCGIFFLAGS:
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;
case SIOCGIFCAP:
ifr->ifr_reqcap = ifp->if_capabilities;
ifr->ifr_curcap = ifp->if_capenable;
break;
#ifdef MAC
case SIOCGIFMAC:
error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
break;
#endif
1994-05-24 10:09:53 +00:00
case SIOCGIFMETRIC:
ifr->ifr_metric = ifp->if_metric;
break;
case SIOCGIFMTU:
ifr->ifr_mtu = ifp->if_mtu;
break;
case SIOCGIFPHYS:
/* XXXGL: did this ever worked? */
ifr->ifr_phys = 0;
break;
case SIOCGIFDESCR:
error = 0;
sx_slock(&ifdescr_sx);
if (ifp->if_description == NULL)
error = ENOMSG;
else {
/* space for terminating nul */
descrlen = strlen(ifp->if_description) + 1;
if (ifr->ifr_buffer.length < descrlen)
ifr->ifr_buffer.buffer = NULL;
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;
case SIOCGIFFIB:
ifr->ifr_fib = ifp->if_fib;
break;
case SIOCSIFFIB:
error = priv_check(td, PRIV_NET_SETIFFIB);
if (error)
return (error);
if (ifr->ifr_fib >= rt_numfibs)
return (EINVAL);
ifp->if_fib = ifr->ifr_fib;
break;
1994-05-24 10:09:53 +00:00
case SIOCSIFFLAGS:
error = priv_check(td, PRIV_NET_SETIFFLAGS);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
/*
* Currently, no driver owned flags pass the IFF_CANTCHANGE
* check, so we don't need special handling here yet.
*/
new_flags = (ifr->ifr_flags & 0xffff) |
(ifr->ifr_flagshigh << 16);
if (ifp->if_flags & IFF_UP &&
(new_flags & IFF_UP) == 0) {
1994-05-24 10:09:53 +00:00
if_down(ifp);
} else if (new_flags & IFF_UP &&
(ifp->if_flags & IFF_UP) == 0) {
1994-05-24 10:09:53 +00:00
if_up(ifp);
}
/* 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;
if (log_promisc_mode_change)
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) |
(new_flags &~ IFF_CANTCHANGE);
if (ifp->if_ioctl) {
1994-05-24 10:09:53 +00:00
(void) (*ifp->if_ioctl)(ifp, cmd, data);
}
getmicrotime(&ifp->if_lastchange);
1994-05-24 10:09:53 +00:00
break;
case SIOCSIFCAP:
error = priv_check(td, PRIV_NET_SETIFCAP);
if (error)
return (error);
if (ifp->if_ioctl == NULL)
return (EOPNOTSUPP);
if (ifr->ifr_reqcap & ~ifp->if_capabilities)
return (EINVAL);
error = (*ifp->if_ioctl)(ifp, cmd, data);
if (error == 0)
getmicrotime(&ifp->if_lastchange);
break;
#ifdef MAC
case SIOCSIFMAC:
error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
break;
#endif
case SIOCSIFNAME:
error = priv_check(td, PRIV_NET_SETIFNAME);
if (error)
return (error);
error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
if (error != 0)
return (error);
if (new_name[0] == '\0')
return (EINVAL);
if (new_name[IFNAMSIZ-1] != '\0') {
new_name[IFNAMSIZ-1] = '\0';
if (strlen(new_name) == IFNAMSIZ-1)
return (EINVAL);
}
if (ifunit(new_name) != NULL)
return (EEXIST);
/*
* 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;
/* Announce the departure of the interface. */
rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
log(LOG_INFO, "%s: changing name to '%s'\n",
ifp->if_xname, new_name);
IF_ADDR_WLOCK(ifp);
strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
ifa = ifp->if_addr;
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;
IF_ADDR_WUNLOCK(ifp);
EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
/* Announce the return of the interface. */
rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
ifp->if_flags &= ~IFF_RENAMING;
break;
#ifdef VIMAGE
case SIOCSIFVNET:
error = priv_check(td, PRIV_NET_SETIFVNET);
if (error)
return (error);
error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
break;
#endif
1994-05-24 10:09:53 +00:00
case SIOCSIFMETRIC:
error = priv_check(td, PRIV_NET_SETIFMETRIC);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
ifp->if_metric = ifr->ifr_metric;
getmicrotime(&ifp->if_lastchange);
1994-05-24 10:09:53 +00:00
break;
case SIOCSIFPHYS:
error = priv_check(td, PRIV_NET_SETIFPHYS);
if (error)
return (error);
if (ifp->if_ioctl == NULL)
return (EOPNOTSUPP);
error = (*ifp->if_ioctl)(ifp, cmd, data);
if (error == 0)
getmicrotime(&ifp->if_lastchange);
break;
case SIOCSIFMTU:
{
u_long oldmtu = ifp->if_mtu;
error = priv_check(td, PRIV_NET_SETIFMTU);
if (error)
return (error);
if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
return (EINVAL);
if (ifp->if_ioctl == NULL)
return (EOPNOTSUPP);
error = (*ifp->if_ioctl)(ifp, cmd, data);
if (error == 0) {
getmicrotime(&ifp->if_lastchange);
rt_ifmsg(ifp);
}
/*
* If the link MTU changed, do network layer specific procedure.
*/
if (ifp->if_mtu != oldmtu) {
#ifdef INET6
nd6_setmtu(ifp);
#endif
rt_updatemtu(ifp);
}
break;
}
1994-05-24 10:09:53 +00:00
case SIOCADDMULTI:
case SIOCDELMULTI:
if (cmd == SIOCADDMULTI)
error = priv_check(td, PRIV_NET_ADDMULTI);
else
error = priv_check(td, PRIV_NET_DELMULTI);
if (error)
1994-05-24 10:09:53 +00:00
return (error);
/* Don't allow group membership on non-multicast interfaces. */
if ((ifp->if_flags & IFF_MULTICAST) == 0)
return (EOPNOTSUPP);
/* Don't let users screw up protocols' entries. */
if (ifr->ifr_addr.sa_family != AF_LINK)
return (EINVAL);
if (cmd == SIOCADDMULTI) {
struct ifmultiaddr *ifma;
/*
* 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_RLOCK(ifp);
ifma = if_findmulti(ifp, &ifr->ifr_addr);
IF_ADDR_RUNLOCK(ifp);
if (ifma != NULL)
error = EADDRINUSE;
else
error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
} else {
error = if_delmulti(ifp, &ifr->ifr_addr);
}
if (error == 0)
getmicrotime(&ifp->if_lastchange);
break;
1994-05-24 10:09:53 +00:00
case SIOCSIFPHYADDR:
case SIOCDIFPHYADDR:
#ifdef INET6
case SIOCSIFPHYADDR_IN6:
#endif
case SIOCSIFMEDIA:
case SIOCSIFGENERIC:
error = priv_check(td, PRIV_NET_HWIOCTL);
1997-05-03 21:07:13 +00:00
if (error)
return (error);
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)
getmicrotime(&ifp->if_lastchange);
break;
1997-05-03 21:07:13 +00:00
case SIOCGIFSTATUS:
case SIOCGIFPSRCADDR:
case SIOCGIFPDSTADDR:
1997-05-03 21:07:13 +00:00
case SIOCGIFMEDIA:
case SIOCGIFXMEDIA:
case SIOCGIFGENERIC:
if (ifp->if_ioctl == NULL)
1997-05-03 21:07:13 +00:00
return (EOPNOTSUPP);
error = (*ifp->if_ioctl)(ifp, cmd, data);
break;
1997-05-03 21:07:13 +00:00
case SIOCSIFLLADDR:
error = priv_check(td, PRIV_NET_SETLLADDR);
if (error)
return (error);
error = if_setlladdr(ifp,
ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
break;
case SIOCAIFGROUP:
{
struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
error = priv_check(td, PRIV_NET_ADDIFGROUP);
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;
error = priv_check(td, PRIV_NET_DELIFGROUP);
if (error)
return (error);
if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
return (error);
break;
}
1994-05-24 10:09:53 +00:00
default:
error = ENOIOCTL;
break;
}
return (error);
}
#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
/*
* Interface ioctls.
*/
int
ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
{
struct ifnet *ifp;
struct ifreq *ifr;
int error;
int oif_flags;
CURVNET_SET(so->so_vnet);
switch (cmd) {
case SIOCGIFCONF:
error = ifconf(cmd, data);
CURVNET_RESTORE();
return (error);
#ifdef COMPAT_FREEBSD32
case SIOCGIFCONF32:
{
struct ifconf32 *ifc32;
struct ifconf ifc;
ifc32 = (struct ifconf32 *)data;
ifc.ifc_len = ifc32->ifc_len;
ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
error = ifconf(SIOCGIFCONF, (void *)&ifc);
CURVNET_RESTORE();
if (error == 0)
ifc32->ifc_len = ifc.ifc_len;
return (error);
}
#endif
}
ifr = (struct ifreq *)data;
switch (cmd) {
#ifdef VIMAGE
case SIOCSIFRVNET:
error = priv_check(td, PRIV_NET_SETIFVNET);
if (error == 0)
error = if_vmove_reclaim(td, ifr->ifr_name,
ifr->ifr_jid);
CURVNET_RESTORE();
return (error);
#endif
case SIOCIFCREATE:
case SIOCIFCREATE2:
error = priv_check(td, PRIV_NET_IFCREATE);
if (error == 0)
error = if_clone_create(ifr->ifr_name,
sizeof(ifr->ifr_name),
cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
CURVNET_RESTORE();
return (error);
case SIOCIFDESTROY:
error = priv_check(td, PRIV_NET_IFDESTROY);
if (error == 0)
error = if_clone_destroy(ifr->ifr_name);
CURVNET_RESTORE();
return (error);
case SIOCIFGCLONERS:
error = if_clone_list((struct if_clonereq *)data);
CURVNET_RESTORE();
return (error);
case SIOCGIFGMEMB:
error = if_getgroupmembers((struct ifgroupreq *)data);
CURVNET_RESTORE();
return (error);
#if defined(INET) || defined(INET6)
case SIOCSVH:
case SIOCGVH:
if (carp_ioctl_p == NULL)
error = EPROTONOSUPPORT;
else
error = (*carp_ioctl_p)(ifr, cmd, td);
CURVNET_RESTORE();
return (error);
#endif
}
ifp = ifunit_ref(ifr->ifr_name);
if (ifp == NULL) {
CURVNET_RESTORE();
return (ENXIO);
}
error = ifhwioctl(cmd, ifp, data, td);
if (error != ENOIOCTL) {
if_rele(ifp);
CURVNET_RESTORE();
return (error);
}
oif_flags = ifp->if_flags;
if (so->so_proto == NULL) {
if_rele(ifp);
CURVNET_RESTORE();
return (EOPNOTSUPP);
}
/*
* Pass the request on to the socket control method, and if the
* latter returns EOPNOTSUPP, directly to the interface.
*
* Make an exception for the legacy SIOCSIF* requests. Drivers
* trust SIOCSIFADDR et al to come from an already privileged
* layer, and do not perform any credentials checks or input
* validation.
*/
error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data,
ifp, td));
if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
error = (*ifp->if_ioctl)(ifp, cmd, data);
if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
#ifdef INET6
if (ifp->if_flags & IFF_UP)
in6_if_up(ifp);
#endif
1994-05-24 10:09:53 +00:00
}
if_rele(ifp);
CURVNET_RESTORE();
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* The code common to handling reference counted flags,
* e.g., in ifpromisc() and if_allmulti().
* The "pflag" argument can specify a permanent mode flag to check,
* such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
*
* Only to be used on stack-owned flags, not driver-owned flags.
*/
static int
if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
{
struct ifreq ifr;
int error;
int oldflags, oldcount;
/* Sanity checks to catch programming errors */
KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
("%s: setting driver-owned flag %d", __func__, flag));
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));
/* In case this mode is permanent, just touch refcount */
if (ifp->if_flags & pflag) {
*refcount += onswitch ? 1 : -1;
return (0);
}
/* 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)++)
return (0);
ifp->if_flags |= flag;
} else {
if (--(*refcount))
return (0);
ifp->if_flags &= ~flag;
}
/* Call down the driver since we've changed interface flags */
if (ifp->if_ioctl == NULL) {
error = EOPNOTSUPP;
goto recover;
}
ifr.ifr_flags = ifp->if_flags & 0xffff;
ifr.ifr_flagshigh = ifp->if_flags >> 16;
error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
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) &&
log_promisc_mode_change)
log(LOG_INFO, "%s: promiscuous mode %s\n",
ifp->if_xname,
(ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
return (error);
}
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
ifconf(u_long cmd, caddr_t data)
1994-05-24 10:09:53 +00:00
{
struct ifconf *ifc = (struct ifconf *)data;
struct ifnet *ifp;
struct ifaddr *ifa;
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;
/* Prevent hostile input from being able to crash the system */
if (ifc->ifc_len <= 0)
return (EINVAL);
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
IFNET_RLOCK();
TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
int addrs;
1995-05-30 08:16:23 +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));
if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
>= sizeof(ifr.ifr_name)) {
sbuf_delete(sb);
IFNET_RUNLOCK();
return (ENAMETOOLONG);
}
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;
IF_ADDR_RLOCK(ifp);
TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
struct sockaddr *sa = ifa->ifa_addr;
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
if (sa->sa_len <= sizeof(*sa)) {
ifr.ifr_addr = *sa;
sbuf_bcat(sb, &ifr, sizeof(ifr));
max_len += sizeof(ifr);
1994-05-24 10:09:53 +00:00
} else {
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
}
if (sbuf_error(sb) == 0)
valid_len = sbuf_len(sb);
1994-05-24 10:09:53 +00:00
}
IF_ADDR_RUNLOCK(ifp);
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));
sbuf_bcat(sb, &ifr, sizeof(ifr));
max_len += sizeof(ifr);
if (sbuf_error(sb) == 0)
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();
/*
* 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;
sbuf_finish(sb);
error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
sbuf_delete(sb);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
2005-02-22 15:29:29 +00:00
* Just like ifpromisc(), but for all-multicast-reception mode.
*/
int
if_allmulti(struct ifnet *ifp, int onswitch)
{
return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
}
struct ifmultiaddr *
if_findmulti(struct ifnet *ifp, const struct sockaddr *sa)
{
struct ifmultiaddr *ifma;
IF_ADDR_LOCK_ASSERT(ifp);
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (sa->sa_family == AF_LINK) {
if (sa_dl_equal(ifma->ifma_addr, sa))
break;
} else {
if (sa_equal(ifma->ifma_addr, sa))
break;
}
}
return ifma;
}
/*
* 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.
*/
static struct ifmultiaddr *
if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
int mflags)
{
struct ifmultiaddr *ifma;
struct sockaddr *dupsa;
ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
M_ZERO);
if (ifma == NULL)
return (NULL);
dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
if (dupsa == NULL) {
free(ifma, M_IFMADDR);
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);
}
dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
if (dupsa == NULL) {
free(ifma->ifma_addr, M_IFMADDR);
free(ifma, M_IFMADDR);
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)
{
KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
ifma->ifma_refcount));
if (ifma->ifma_lladdr != NULL)
free(ifma->ifma_lladdr, M_IFMADDR);
free(ifma->ifma_addr, M_IFMADDR);
free(ifma, M_IFMADDR);
}
/*
* 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;
struct sockaddr_dl sdl;
int error;
/*
* If the address is already present, return a new reference to it;
* otherwise, allocate storage and set up a new address.
*/
IF_ADDR_WLOCK(ifp);
ifma = if_findmulti(ifp, sa);
if (ifma != NULL) {
ifma->ifma_refcount++;
if (retifma != NULL)
*retifma = ifma;
IF_ADDR_WUNLOCK(ifp);
return (0);
}
/*
* 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.
* Most link layer resolving functions returns address data which
* fits inside default sockaddr_dl structure. However callback
* can allocate another sockaddr structure, in that case we need to
* free it later.
*/
llsa = NULL;
ll_ifma = NULL;
if (ifp->if_resolvemulti != NULL) {
/* Provide called function with buffer size information */
sdl.sdl_len = sizeof(sdl);
llsa = (struct sockaddr *)&sdl;
error = ifp->if_resolvemulti(ifp, &llsa, sa);
if (error)
goto unlock_out;
}
/*
* 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;
}
/*
* 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) {
--ifma->ifma_refcount;
if_freemulti(ifma);
error = ENOMEM;
goto free_llsa_out;
}
TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
ifma_link);
} else
ll_ifma->ifma_refcount++;
ifma->ifma_llifma = ll_ifma;
}
/*
* 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.
*/
TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
if (retifma != NULL)
*retifma = ifma;
/*
* Must generate the message while holding the lock so that 'ifma'
* pointer is still valid.
*/
rt_newmaddrmsg(RTM_NEWMADDR, ifma);
IF_ADDR_WUNLOCK(ifp);
/*
* We are certain we have added something, so call down to the
* interface to let them know about it.
*/
if (ifp->if_ioctl != NULL) {
(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
}
if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
link_free_sdl(llsa);
return (0);
free_llsa_out:
if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl))
link_free_sdl(llsa);
unlock_out:
IF_ADDR_WUNLOCK(ifp);
return (error);
}
/*
* 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().
*/
int
if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
{
struct ifmultiaddr *ifma;
int lastref;
#ifdef INVARIANTS
struct ifnet *oifp;
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(oifp, &V_ifnet, if_link)
if (ifp == oifp)
break;
if (ifp != oifp)
ifp = NULL;
IFNET_RUNLOCK_NOSLEEP();
KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
#endif
if (ifp == NULL)
return (ENOENT);
IF_ADDR_WLOCK(ifp);
lastref = 0;
ifma = if_findmulti(ifp, sa);
if (ifma != NULL)
lastref = if_delmulti_locked(ifp, ifma, 0);
IF_ADDR_WUNLOCK(ifp);
if (ifma == NULL)
return (ENOENT);
if (lastref && ifp->if_ioctl != NULL) {
(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
}
return (0);
}
/*
* 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_WLOCK(ifp);
TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
if_delmulti_locked(ifp, ifma, 0);
IF_ADDR_WUNLOCK(ifp);
}
/*
* Delete a multicast group membership by group membership pointer.
* Network-layer protocol domains must use this routine.
*
* It is safe to call this routine if the ifp disappeared.
*/
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;
IFNET_RLOCK_NOSLEEP();
TAILQ_FOREACH(oifp, &V_ifnet, if_link)
if (ifp == oifp)
break;
if (ifp != oifp) {
printf("%s: ifnet %p disappeared\n", __func__, ifp);
ifp = NULL;
}
IFNET_RUNLOCK_NOSLEEP();
}
#endif
/*
* If and only if the ifnet instance exists: Acquire the address lock.
*/
if (ifp != NULL)
IF_ADDR_WLOCK(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.
*/
IF_ADDR_WUNLOCK(ifp);
if (lastref && ifp->if_ioctl != NULL) {
(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
}
}
}
/*
* 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_WLOCK_ASSERT(ifp);
}
ifp = ifma->ifma_ifp;
/*
* If the ifnet is detaching, null out references to ifnet,
* so that upper protocol layers will notice, and not attempt
* 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.
*/
if (detaching) {
#ifdef DIAGNOSTIC
printf("%s: detaching ifnet instance %p\n", __func__, ifp);
#endif
/*
* 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;
}
}
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;
if (ll_ifma != NULL) {
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);
}
if_freemulti(ll_ifma);
}
}
if (ifp != NULL)
TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
if_freemulti(ifma);
/*
* The last reference to this instance of struct ifmultiaddr
* was released; the hardware should be notified of this change.
*/
return 1;
}
/*
* 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.
*
* Set noinline to be dtrace-friendly
*/
__noinline int
if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
{
struct sockaddr_dl *sdl;
struct ifaddr *ifa;
struct ifreq ifr;
IF_ADDR_RLOCK(ifp);
ifa = ifp->if_addr;
if (ifa == NULL) {
IF_ADDR_RUNLOCK(ifp);
return (EINVAL);
}
ifa_ref(ifa);
IF_ADDR_RUNLOCK(ifp);
sdl = (struct sockaddr_dl *)ifa->ifa_addr;
if (sdl == NULL) {
ifa_free(ifa);
return (EINVAL);
}
if (len != sdl->sdl_alen) { /* don't allow length to change */
ifa_free(ifa);
return (EINVAL);
}
switch (ifp->if_type) {
case IFT_ETHER:
case IFT_FDDI:
case IFT_XETHER:
case IFT_ISO88025:
case IFT_L2VLAN:
case IFT_BRIDGE:
case IFT_ARCNET:
case IFT_IEEE8023ADLAG:
case IFT_IEEE80211:
bcopy(lladdr, LLADDR(sdl), len);
ifa_free(ifa);
break;
default:
ifa_free(ifa);
return (ENODEV);
}
/*
* 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) {
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);
}
}
EVENTHANDLER_INVOKE(iflladdr_event, ifp);
return (0);
}
Implement interface link header precomputation API. Add if_requestencap() interface method which is capable of calculating various link headers for given interface. Right now there is support for INET/INET6/ARP llheader calculation (IFENCAP_LL type request). Other types are planned to support more complex calculation (L2 multipath lagg nexthops, tunnel encap nexthops, etc..). Reshape 'struct route' to be able to pass additional data (with is length) to prepend to mbuf. These two changes permits routing code to pass pre-calculated nexthop data (like L2 header for route w/gateway) down to the stack eliminating the need for other lookups. It also brings us closer to more complex scenarios like transparently handling MPLS nexthops and tunnel interfaces. Last, but not least, it removes layering violation introduced by flowtable code (ro_lle) and simplifies handling of existing if_output consumers. ARP/ND changes: Make arp/ndp stack pre-calculate link header upon installing/updating lle record. Interface link address change are handled by re-calculating headers for all lles based on if_lladdr event. After these changes, arpresolve()/nd6_resolve() returns full pre-calculated header for supported interfaces thus simplifying if_output(). Move these lookups to separate ether_resolve_addr() function which ether returs error or fully-prepared link header. Add <arp|nd6_>resolve_addr() compat versions to return link addresses instead of pre-calculated data. BPF changes: Raw bpf writes occupied _two_ cases: AF_UNSPEC and pseudo_AF_HDRCMPLT. Despite the naming, both of there have ther header "complete". The only difference is that interface source mac has to be filled by OS for AF_UNSPEC (controlled via BIOCGHDRCMPLT). This logic has to stay inside BPF and not pollute if_output() routines. Convert BPF to pass prepend data via new 'struct route' mechanism. Note that it does not change non-optimized if_output(): ro_prepend handling is purely optional. Side note: hackish pseudo_AF_HDRCMPLT is supported for ethernet and FDDI. It is not needed for ethernet anymore. The only remaining FDDI user is dev/pdq mostly untouched since 2007. FDDI support was eliminated from OpenBSD in 2013 (sys/net/if_fddisubr.c rev 1.65). Flowtable changes: Flowtable violates layering by saving (and not correctly managing) rtes/lles. Instead of passing lle pointer, pass pointer to pre-calculated header data from that lle. Differential Revision: https://reviews.freebsd.org/D4102
2015-12-31 05:03:27 +00:00
/*
* Compat function for handling basic encapsulation requests.
* Not converted stacks (FDDI, IB, ..) supports traditional
* output model: ARP (and other similar L2 protocols) are handled
* inside output routine, arpresolve/nd6_resolve() returns MAC
* address instead of full prepend.
*
* This function creates calculated header==MAC for IPv4/IPv6 and
* returns EAFNOSUPPORT (which is then handled in ARP code) for other
* address families.
*/
static int
if_requestencap_default(struct ifnet *ifp, struct if_encap_req *req)
{
if (req->rtype != IFENCAP_LL)
return (EOPNOTSUPP);
if (req->bufsize < req->lladdr_len)
return (ENOMEM);
switch (req->family) {
case AF_INET:
case AF_INET6:
break;
default:
return (EAFNOSUPPORT);
}
/* Copy lladdr to storage as is */
memmove(req->buf, req->lladdr, req->lladdr_len);
req->bufsize = req->lladdr_len;
req->lladdr_off = 0;
return (0);
}
/*
* 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);
}
int
if_printf(struct ifnet *ifp, const char * fmt, ...)
{
va_list ap;
int retval;
retval = printf("%s: ", ifp->if_xname);
va_start(ap, fmt);
retval += vprintf(fmt, ap);
va_end(ap);
return (retval);
}
void
if_start(struct ifnet *ifp)
{
(*(ifp)->if_start)(ifp);
}
/*
* 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);
}
static void
if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
{
m_freem(m);
}
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_UNLOCK(ifq);
if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
m_freem(m);
return (0);
}
if (ifp != NULL) {
if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust);
if (m->m_flags & (M_BCAST|M_MCAST))
if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1);
active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
}
_IF_ENQUEUE(ifq, m);
IF_UNLOCK(ifq);
if (ifp != NULL && !active)
(*(ifp)->if_start)(ifp);
return (1);
}
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)
{
KASSERT(if_com_alloc[type] != NULL,
("if_deregister_com_alloc: %d not registered", type));
KASSERT(if_com_free[type] != NULL,
("if_deregister_com_alloc: %d free not registered", type));
if_com_alloc[type] = NULL;
if_com_free[type] = NULL;
}
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
/* API for driver access to network stack owned ifnet.*/
uint64_t
if_setbaudrate(struct ifnet *ifp, uint64_t baudrate)
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
{
uint64_t oldbrate;
oldbrate = ifp->if_baudrate;
ifp->if_baudrate = baudrate;
return (oldbrate);
}
uint64_t
if_getbaudrate(if_t ifp)
{
return (((struct ifnet *)ifp)->if_baudrate);
}
int
if_setcapabilities(if_t ifp, int capabilities)
{
((struct ifnet *)ifp)->if_capabilities = capabilities;
return (0);
}
int
if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit)
{
((struct ifnet *)ifp)->if_capabilities |= setbit;
((struct ifnet *)ifp)->if_capabilities &= ~clearbit;
return (0);
}
int
if_getcapabilities(if_t ifp)
{
return ((struct ifnet *)ifp)->if_capabilities;
}
int
if_setcapenable(if_t ifp, int capabilities)
{
((struct ifnet *)ifp)->if_capenable = capabilities;
return (0);
}
int
if_setcapenablebit(if_t ifp, int setcap, int clearcap)
{
if(setcap)
((struct ifnet *)ifp)->if_capenable |= setcap;
if(clearcap)
((struct ifnet *)ifp)->if_capenable &= ~clearcap;
return (0);
}
const char *
if_getdname(if_t ifp)
{
return ((struct ifnet *)ifp)->if_dname;
}
int
if_togglecapenable(if_t ifp, int togglecap)
{
((struct ifnet *)ifp)->if_capenable ^= togglecap;
return (0);
}
int
if_getcapenable(if_t ifp)
{
return ((struct ifnet *)ifp)->if_capenable;
}
/*
* This is largely undesirable because it ties ifnet to a device, but does
* provide flexiblity for an embedded product vendor. Should be used with
* the understanding that it violates the interface boundaries, and should be
* a last resort only.
*/
int
if_setdev(if_t ifp, void *dev)
{
return (0);
}
int
if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags)
{
((struct ifnet *)ifp)->if_drv_flags |= set_flags;
((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags;
return (0);
}
int
if_getdrvflags(if_t ifp)
{
return ((struct ifnet *)ifp)->if_drv_flags;
}
int
if_setdrvflags(if_t ifp, int flags)
{
((struct ifnet *)ifp)->if_drv_flags = flags;
return (0);
}
int
if_setflags(if_t ifp, int flags)
{
((struct ifnet *)ifp)->if_flags = flags;
return (0);
}
int
if_setflagbits(if_t ifp, int set, int clear)
{
((struct ifnet *)ifp)->if_flags |= set;
((struct ifnet *)ifp)->if_flags &= ~clear;
return (0);
}
int
if_getflags(if_t ifp)
{
return ((struct ifnet *)ifp)->if_flags;
}
int
if_clearhwassist(if_t ifp)
{
((struct ifnet *)ifp)->if_hwassist = 0;
return (0);
}
int
if_sethwassistbits(if_t ifp, int toset, int toclear)
{
((struct ifnet *)ifp)->if_hwassist |= toset;
((struct ifnet *)ifp)->if_hwassist &= ~toclear;
return (0);
}
int
if_sethwassist(if_t ifp, int hwassist_bit)
{
((struct ifnet *)ifp)->if_hwassist = hwassist_bit;
return (0);
}
int
if_gethwassist(if_t ifp)
{
return ((struct ifnet *)ifp)->if_hwassist;
}
int
if_setmtu(if_t ifp, int mtu)
{
((struct ifnet *)ifp)->if_mtu = mtu;
return (0);
}
int
if_getmtu(if_t ifp)
{
return ((struct ifnet *)ifp)->if_mtu;
}
Make checks for rt_mtu generic: Some virtual if drivers has (ab)used ifa ifa_rtrequest hook to enforce route MTU to be not bigger that interface MTU. While ifa_rtrequest hooking might be an option in some situation, it is not feasible to do MTU checks there: generic (or per-domain) routing code is perfectly capable of doing this. We currrently have 3 places where MTU is altered: 1) route addition. In this case domain overrides radix _addroute callback (in[6]_addroute) and all necessary checks/fixes are/can be done there. 2) route change (especially, GW change). In this case, there are no explicit per-domain calls, but one can override rte by setting ifa_rtrequest hook to domain handler (inet6 does this). 3) ifconfig ifaceX mtu YYYY In this case, we have no callbacks, but ip[6]_output performes runtime checks and decreases rt_mtu if necessary. Generally, the goals are to be able to handle all MTU changes in control plane, not in runtime part, and properly deal with increased interface MTU. This commit changes the following: * removes hooks setting MTU from drivers side * adds proper per-doman MTU checks for case 1) * adds generic MTU check for case 2) * The latter is done by using new dom_ifmtu callback since if_mtu denotes L3 interface MTU, e.g. maximum trasmitted _packet_ size. However, IPv6 mtu might be different from if_mtu one (e.g. default 1280) for some cases, so we need an abstract way to know maximum MTU size for given interface and domain. * moves rt_setmetrics() before MTU/ifa_rtrequest hooks since it copies user-supplied data which must be checked. * removes RT_LOCK_ASSERT() from other ifa_rtrequest hooks to be able to use this functions on new non-inserted rte. More changes will follow soon. MFC after: 1 month Sponsored by: Yandex LLC
2014-11-06 13:13:09 +00:00
int
if_getmtu_family(if_t ifp, int family)
{
struct domain *dp;
for (dp = domains; dp; dp = dp->dom_next) {
if (dp->dom_family == family && dp->dom_ifmtu != NULL)
return (dp->dom_ifmtu((struct ifnet *)ifp));
}
return (((struct ifnet *)ifp)->if_mtu);
}
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
int
if_setsoftc(if_t ifp, void *softc)
{
((struct ifnet *)ifp)->if_softc = softc;
return (0);
}
void *
if_getsoftc(if_t ifp)
{
return ((struct ifnet *)ifp)->if_softc;
}
void
if_setrcvif(struct mbuf *m, if_t ifp)
{
m->m_pkthdr.rcvif = (struct ifnet *)ifp;
}
void
if_setvtag(struct mbuf *m, uint16_t tag)
{
m->m_pkthdr.ether_vtag = tag;
}
uint16_t
if_getvtag(struct mbuf *m)
{
return (m->m_pkthdr.ether_vtag);
}
int
if_sendq_empty(if_t ifp)
{
return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd);
}
struct ifaddr *
if_getifaddr(if_t ifp)
{
return ((struct ifnet *)ifp)->if_addr;
}
int
if_getamcount(if_t ifp)
{
return ((struct ifnet *)ifp)->if_amcount;
}
int
if_setsendqready(if_t ifp)
{
IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd);
return (0);
}
int
if_setsendqlen(if_t ifp, int tx_desc_count)
{
IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count);
((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count;
return (0);
}
int
if_vlantrunkinuse(if_t ifp)
{
return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0;
}
int
if_input(if_t ifp, struct mbuf* sendmp)
{
(*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp);
return (0);
}
/* XXX */
#ifndef ETH_ADDR_LEN
#define ETH_ADDR_LEN 6
#endif
int
if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max)
{
struct ifmultiaddr *ifma;
uint8_t *lmta = (uint8_t *)mta;
int mcnt = 0;
TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
if (mcnt == max)
break;
bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
&lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
mcnt++;
}
*cnt = mcnt;
return (0);
}
int
if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max)
{
int error;
if_maddr_rlock(ifp);
error = if_setupmultiaddr(ifp, mta, cnt, max);
if_maddr_runlock(ifp);
return (error);
}
int
if_multiaddr_count(if_t ifp, int max)
{
struct ifmultiaddr *ifma;
int count;
count = 0;
if_maddr_rlock(ifp);
TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
count++;
if (count == max)
break;
}
if_maddr_runlock(ifp);
return (count);
}
int
if_multi_apply(struct ifnet *ifp, int (*filter)(void *, struct ifmultiaddr *, int), void *arg)
{
struct ifmultiaddr *ifma;
int cnt = 0;
if_maddr_rlock(ifp);
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
cnt += filter(arg, ifma, cnt);
if_maddr_runlock(ifp);
return (cnt);
}
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
struct mbuf *
if_dequeue(if_t ifp)
{
struct mbuf *m;
IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m);
return (m);
}
int
if_sendq_prepend(if_t ifp, struct mbuf *m)
{
IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m);
return (0);
}
int
if_setifheaderlen(if_t ifp, int len)
{
((struct ifnet *)ifp)->if_hdrlen = len;
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
return (0);
}
caddr_t
if_getlladdr(if_t ifp)
{
return (IF_LLADDR((struct ifnet *)ifp));
}
void *
if_gethandle(u_char type)
{
return (if_alloc(type));
}
void
if_bpfmtap(if_t ifh, struct mbuf *m)
{
struct ifnet *ifp = (struct ifnet *)ifh;
BPF_MTAP(ifp, m);
}
void
if_etherbpfmtap(if_t ifh, struct mbuf *m)
{
struct ifnet *ifp = (struct ifnet *)ifh;
ETHER_BPF_MTAP(ifp, m);
}
void
if_vlancap(if_t ifh)
{
struct ifnet *ifp = (struct ifnet *)ifh;
VLAN_CAPABILITIES(ifp);
}
void
if_setinitfn(if_t ifp, void (*init_fn)(void *))
{
((struct ifnet *)ifp)->if_init = init_fn;
}
void
if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t))
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
{
((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn;
}
void
if_setstartfn(if_t ifp, void (*start_fn)(if_t))
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
{
((struct ifnet *)ifp)->if_start = (void *)start_fn;
}
void
if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn)
{
((struct ifnet *)ifp)->if_transmit = start_fn;
}
void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn)
{
((struct ifnet *)ifp)->if_qflush = flush_fn;
}
void
if_setgetcounterfn(if_t ifp, if_get_counter_t fn)
{
ifp->if_get_counter = fn;
}
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
/* Revisit these - These are inline functions originally. */
int
drbr_inuse_drv(if_t ifh, struct buf_ring *br)
{
return drbr_inuse(ifh, br);
Introduce a procedural interface to the ifnet structure. The new interface allows the ifnet structure to be defined as an opaque type in NIC drivers. This then allows the ifnet structure to be changed without a need to change or recompile NIC drivers. Put differently, NIC drivers can be written and compiled once and be used with different network stack implementations, provided of course that those network stack implementations have an API and ABI compatible interface. This commit introduces the 'if_t' type to replace 'struct ifnet *' as the type of a network interface. The 'if_t' type is defined as 'void *' to enable the compiler to perform type conversion to 'struct ifnet *' and vice versa where needed and without warnings. The functions that implement the API are the only functions that need to have an explicit cast. The MII code has been converted to use the driver API to avoid unnecessary code churn. Code churn comes from having to work with both converted and unconverted drivers in correlation with having callback functions that take an interface. By converting the MII code first, the callback functions can be defined so that the compiler will perform the typecasts automatically. As soon as all drivers have been converted, the if_t type can be redefined as needed and the API functions can be fix to not need an explicit cast. The immediate benefactors of this change are: 1. Juniper Networks - The network stack implementation in Junos is entirely different from FreeBSD's one and this change allows Juniper to build "stock" NIC drivers that can be used in combination with both the FreeBSD and Junos stacks. 2. FreeBSD - This change opens the door towards changing ifnet and implementing new features and optimizations in the network stack without it requiring a change in the many NIC drivers FreeBSD has. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Reviewed by: glebius@ Obtained from: Juniper Networks, Inc.
2014-06-02 17:54:39 +00:00
}
struct mbuf*
drbr_dequeue_drv(if_t ifh, struct buf_ring *br)
{
return drbr_dequeue(ifh, br);
}
int
drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br)
{
return drbr_needs_enqueue(ifh, br);
}
int
drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m)
{
return drbr_enqueue(ifh, br, m);
}