freebsd-nq/sys/net/net_osdep.h
Hajimu UMEMOTO 4a201dfc18 - update comments to refrect recent BSDs.
- nuke unused macro PSUEDO_SET().
- I believe our if_xname stuff is nothing strange against other BSDs.

Obtained from:	KAME
2003-11-04 14:08:31 +00:00

340 lines
11 KiB
C

/* $FreeBSD$ */
/* $KAME: net_osdep.h,v 1.80 2003/08/09 17:06:39 suz Exp $ */
/*
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the project 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 PROJECT 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 PROJECT 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.
*/
/*
* glue for kernel code programming differences.
*/
/*
* OS dependencies:
* - ioctl
* FreeBSD 3 and later warn when sys/ioctl.h is included in a kernel source
* file. For socket ioctl, we are suggested to use sys/sockio.h.
*
* - RTFREE()
* bsdi does not escape this macro using do-clause, so it is recommended
* to escape the macro explicitly.
* e.g.
* if (rt) {
* RTFREE(rt);
* }
*
* - whether the IPv4 input routine convert the byte order of some fileds
* of the IP header (x: convert to the host byte order, s: strip the header
* length for possible reassembly)
* ip_len ip_id ip_off
* bsdi3: xs x x
* bsdi4: xs x
* freebsd[23]: xs x x
* freebsd4: xs x
* NetBSD: x x
* OpenBSD: xs x x
*
* - ifa_ifwithaf()
* bsdi[34], netbsd, and openbsd define it in sys/net/if.c
* freebsd (all versions) does not have it.
*
* - struct rt_addrinfo
* bsdi4, netbsd 1.5R and beyond: rti_addrs, rti_info[], rti_flags, rti_ifa,
* rti_ifp, and rti_rtm.
* others: rti_addrs and rti_info[] only.
*
* - ifa->ifa_rtrequest
* bsdi4, netbsd 1.5R and beyond: rt_addrinfo *
* others: sockaddr * (note that sys/net/route.c:rtrequest() has an unsafe
* typecast code, from 4.3BSD-reno)
*
* - side effects of rtrequest{,1}(RTM_DELETE)
* BSDI[34]: delete all cloned routes underneath the route.
* FreeBSD[234]: delete all protocol-cloned routes underneath the route.
* note that cloned routes from an interface direct route
* still remain.
* NetBSD: 1.5 have no side effects. KAME/netbsd15, and post-1.5R, have
* the same effects as of BSDI.
* OpenBSD: have no side effects. KAME/openbsd has the same effects as
* of BSDI (the change is not merged - yet).
*
* - privileged process
* NetBSD, FreeBSD 3
* struct proc *p;
* if (p && !suser(p->p_ucred, &p->p_acflag))
* privileged;
* FreeBSD 4
* struct proc *p;
* if (p && !suser(p))
* privileged;
* FreeBSD 5
* struct thread *td;
* if (suser(td))
* privileged;
* OpenBSD, BSDI [34], FreeBSD 2
* struct socket *so;
* if (so->so_state & SS_PRIV)
* privileged;
* - foo_control
* NetBSD, FreeBSD 3
* needs to give struct proc * as argument
* OpenBSD, BSDI [34], FreeBSD 2
* do not need struct proc *
*
* - bpf:
* OpenBSD, NetBSD 1.5, BSDI [34]
* need caddr_t * (= if_bpf **) and struct ifnet *
* FreeBSD 2, FreeBSD 3, NetBSD post-1.5N
* need only struct ifnet * as argument
*
* - bpfattach:
* OpenBSD, NetBSD 1.5, BSDI [34]
* bpfattach(caddr_t *, struct ifnet *, u_int, u_int)
* FreeBSD, NetBSD 1.6:
* bpfattach(struct ifnet *, u_int, u_int)
*
* - bpf_mtap:
* OpenBSD, NetBSD, BSDI [34]
* bpf_mtap(caddr_t, struct mbuf *)
* FreeBSD
* bpf_mtap(struct ifnet *, struct mbuf *)
*
* - struct ifnet
* use queue.h? member names if name
* --- --- ---
* FreeBSD 2 no old standard if_name+unit
* FreeBSD 3 yes strange if_name+unit
* FreeBSD 4 yes strange if_name+unit
* FreeBSD 5 yes standard if_xname
* OpenBSD yes standard if_xname
* NetBSD yes standard if_xname
* BSDI [34] no old standard if_name+unit
*
* - usrreq
* NetBSD, OpenBSD, BSDI [34], FreeBSD 2
* single function with PRU_xx, arguments are mbuf
* FreeBSD 3
* separates functions, non-mbuf arguments
*
* - {set,get}sockopt
* NetBSD, OpenBSD, BSDI [34], FreeBSD 2
* manipulation based on mbuf
* FreeBSD 3
* non-mbuf manipulation using sooptcopy{in,out}()
*
* - timeout() and untimeout()
* NetBSD 1.4.x, OpenBSD, BSDI [34], FreeBSD 2
* timeout() is a void function
* FreeBSD 3
* timeout() is non-void, must keep returned value for untimeout()
* callout_xx is also available (sys/callout.h)
* NetBSD 1.5
* timeout() is obsoleted, use callout_xx (sys/callout.h)
* OpenBSD 2.8
* timeout_{add,set,del} is encouraged (sys/timeout.h)
*
* - kernel internal time structure
* FreeBSD 2, NetBSD, OpenBSD, BSD/OS
* mono_time.tv_u?sec, time.tv_u?sec
* FreeBSD [34]
* time_second
* if you need portability, #ifdef out FreeBSD[34], or use microtime(&tv)
* then touch tv.tv_sec (note: microtime is an expensive operation, so
* the use of mono_time is preferred).
*
* - sysctl
* NetBSD, OpenBSD
* foo_sysctl()
* BSDI [34]
* foo_sysctl() but with different style. sysctl_int_arr() takes
* care of most of the cases.
* FreeBSD
* linker hack. however, there are freebsd version differences
* (how wonderful!).
* on FreeBSD[23] function arg #define includes paren.
* int foo SYSCTL_HANDLER_ARGS;
* on FreeBSD4, function arg #define does not include paren.
* int foo(SYSCTL_HANDLER_ARGS);
* on some versions, forward reference to the tree is okay.
* on some versions, you need SYSCTL_DECL(). you need things
* like this.
* #ifdef SYSCTL_DECL
* SYSCTL_DECL(net_inet_ip6);
* #endif
* it is hard to share functions between freebsd and non-freebsd.
*
* - if_ioctl
* NetBSD, FreeBSD 3, BSDI [34]
* 2nd argument is u_long cmd
* FreeBSD 2
* 2nd argument is int cmd
*
* - if attach routines
* NetBSD
* void xxattach(int);
* FreeBSD 2, FreeBSD 3
* void xxattach(void *);
* PSEUDO_SET(xxattach, if_xx);
*
* - ovbcopy()
* in NetBSD 1.4 or later, ovbcopy() is not supplied in the kernel.
* we have updated sys/systm.h to include declaration.
*
* - splnet()
* NetBSD 1.4 or later, and OpenBSD, requires splsoftnet().
* other operating systems use splnet().
*
* - splimp()
* NetBSD 1.6: use splnet() in network, splvm() in vm.
* other operating systems: use splimp().
*
* - dtom()
* NEVER USE IT!
*
* - struct ifnet for loopback interface
* BSDI3: struct ifnet loif;
* BSDI4: struct ifnet *loifp;
* NetBSD, OpenBSD 2.8, FreeBSD2: struct ifnet loif[NLOOP];
* OpenBSD 2.9: struct ifnet *lo0ifp;
*
* odd thing is that many of them refers loif as ifnet *loif,
* not loif[NLOOP], from outside of if_loop.c.
*
* - number of bpf pseudo devices
* others: bpfilter.h, NBPFILTER
* FreeBSD4: bpf.h, NBPF
* solution:
* #if defined(__FreeBSD__) && __FreeBSD__ >= 4
* #include "bpf.h"
* #define NBPFILTER NBPF
* #else
* #include "bpfilter.h"
* #endif
*
* - protosw for IPv4 (sys/netinet)
* FreeBSD4: struct ipprotosw in netinet/ipprotosw.h
* others: struct protosw in sys/protosw.h
*
* - protosw in general.
* NetBSD 1.5 has extra member for ipfilter (netbsd-current dropped
* it so it went away in 1.6).
* NetBSD 1.5 requires PR_LISTEN flag bit with protocols that permit
* listen/accept (like tcp).
*
* - header files with defopt (opt_xx.h)
* FreeBSD3: opt_{inet,ipsec,ip6fw,altq}.h
* FreeBSD4: opt_{inet,inet6,ipsec,ip6fw,altq}.h
* NetBSD: opt_{inet,ipsec,altq}.h
* others: does not use defopt
*
* - IN_MULTICAST/IN_CLASS[A-D] macro.
* OpenBSD and NetBSD: net endian (kernel) or host endian (userland)
* others: always host endian
*
* - (m->m_flags & M_EXT) != 0 does *not* mean that the max data length of
* the mbuf == MCLBYTES.
*
* - sys/kern/uipc_mbuf.c:m_dup()
* freebsd[34]: copies the whole mbuf chain.
* netbsd: similar arg with m_copym().
* others: no m_dup().
*
* - ifa_refcnt (struct ifaddr) management (IFAREF/IFAFREE).
* NetBSD 1.5: always use IFAREF whenever reference gets added.
* always use IFAFREE whenever reference gets freed.
* IFAFREE frees ifaddr when ifa_refcnt reaches 0.
* others: do not increase refcnt for ifp->if_addrlist and in_ifaddr.
* use IFAFREE once when ifaddr is disconnected from
* ifp->if_addrlist and in_ifaddr. IFAFREE frees ifaddr when
* ifa_refcnt goes negative. in KAME environment, IFAREF is
* provided as a compatibility wrapper (use it instead of
* ifa_refcnt++ to reduce #ifdef).
*
* - ifnet.if_lastchange
* freebsd, bsdi, netbsd-current (jun 14 2001-),
* openbsd-current (jun 15 2001-): updated only when IFF_UP changes.
* (RFC1573 ifLastChange interpretation)
* netbsd151, openbsd29: updated whenever packets go through the interface.
* (4.4BSD interpretation)
*
* - kernel compilation options ("options HOGE" in kernel config file)
* freebsd4: sys/conf/options has to have mapping between option
* and a header file (opt_hoge.h).
* netbsd: by default, -DHOGE will go into
* sys/arch/foo/compile/BAR/Makefile.
* if you define mapping in sys/conf/files, you can create
* a header file like opt_hoge.h to help make dependencies.
* bsdi/openbsd: always use -DHOGE in Makefile. there's no need/way
* to have opt_hoge.h.
*
* therefore, opt_hoge.h is mandatory on freebsd4 only.
*
* - MALLOC() macro
* Use it only if the size of the allocation is constant.
* When we do NOT collect statistics about kernel memory usage, the result
* of macro expansion contains a large set of condition branches. If the
* size is not constant, compilation optimization cannot be applied, and
* a bunch of the large branch will be embedded in the kernel code.
*
* - M_COPY_PKTHDR
* openbsd30, freebsd4(after 4.8): M_COPY_PKTHDR is deprecated.
* use M_MOVE_PKTHDR or M_DUP_PKTHDR, depending on how you want
* to handle m_tag.
* others: M_COPY_PKTHDR is available as usual.
*
* - M_READONLY() macro
* OpenBSD 3.0 and NetBSD 1.6 has it.
* FreeBSD 4.x uses M_WRITABLE() macro, which is opposite (NIH!).
* KAME tree has it for all platforms except FreeBSD 4.x.
*
* - TAILQ_EMPTY
* BSD/OS 4.x does not have this macro.
*/
#ifndef __NET_NET_OSDEP_H_DEFINED_
#define __NET_NET_OSDEP_H_DEFINED_
#ifdef _KERNEL
struct ifnet;
extern const char *if_name(struct ifnet *);
#define HAVE_OLD_BPF
#define HAVE_NEW_BPFATTACH
#define ifa_list ifa_link
#define if_addrlist if_addrhead
#define if_list if_link
#define HAVE_PPSRATECHECK
/* sys/net/if.h */
#define WITH_CONVERT_AND_STRIP_IP_LEN
#define WITH_CONVERT_IP_OFF
#endif /*_KERNEL*/
#endif /*__NET_NET_OSDEP_H_DEFINED_ */