Aren/NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
b52ab07dd4
NetBSD/sys/netipsec/ipsec_osdep.h
gdt 0c9c715c58 Fix FAST_IPSEC locking violation. 
Without this change, using ESP tunnels with FAST_IPSEC on a 2-cpu i386
machine results in an mbuf leak.  This change was tested in netbsd-6.

When FAST_IPSEC is enabled and a tunnel is set up, after the
outer packet is stripped off, FAST_IPSEC queues the inner
packet on the appropriate queue (ipinstrq or ip6instrq).
These queues require the KERNEL_LOCK to be held before
using the queue, and the FAST_IPSEC code did not take the
KERNEL_LOCK as required.
KERNEL_LOCK and KERNEL_UNLOCK_ONE calls have been added.

If a struct ifnet instance is passed to the if_handoff
function which does this queuing, the interface's
if_start function may be called.  Some hardware devices
require KERNEL_LOCK to be held; others do not.  Looking
at the body of NetBSD code, other places where an if_start
function is called, KERNEL_LOCK is held.  Thus, the lock is
not released in if_handoff until after the if_start function
is called.  In practice, having the kernel lock when
if_start is called makes no difference - there is not
a single instance in all of the NetBSD code where
if_handoff is passed an instance of struct ifnet.

This commit is the work of Bev Schwartz of BBN.

Approved for Public Release, Distribution Unlimited
This material is based upon work supported by the Defense Advanced Research
Projects Agency and Space and Naval Warfare Systems Center, Pacific, under
Contract No. N66001-09-C-2073.
2013-05-09 19:21:50 +00:00

362 lines
12 KiB
C
Raw Blame History

/* $NetBSD: ipsec_osdep.h,v 1.24 2013/05/09 19:21:50 gdt Exp $ */
/* $FreeBSD: /repoman/r/ncvs/src/sys/netipsec/ipsec_osdep.h,v 1.1 2003/09/29 22:47:45 sam Exp $ */
/*
* Copyright (c) 2003 Jonathan Stone (jonathan@cs.stanford.edu)
*
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 AUTHOR 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.
*/
#ifndef _NETIPSEC_OSDEP_H_
#define _NETIPSEC_OSDEP_H_
#ifdef _KERNEL
/*
* Hide porting differences across different 4.4BSD-derived platforms.
*
* 1. KASSERT() differences:
* 2. Kernel Random-number API differences.
* 3. Is packet data in an mbuf object writeable?
* 4. Packet-header semantics.
* 5. Fast mbuf-cluster allocation.
* 6. Network packet-output macros.
* 7. Elased time, in seconds.
* 8. Test if a socket object opened by a privileged (super) user.
* 9. Global SLIST of all open raw sockets.
* 10. Global SLIST of known interface addresses.
* 11. Type of initialization functions.
* 12. Byte order of ip_off
*/
/*
* 1. KASSERT and spl differences
*
* FreeBSD takes an expression and parenthesized printf() argument-list.
* NetBSD takes one arg: the expression being asserted.
* FreeBSD's SPLASSERT() takes an SPL level as 1st arg and a
* parenthesized printf-format argument list as the second argument.
*
* This difference is hidden by two 2-argument macros and one 1-arg macro:
* IPSEC_ASSERT(expr, msg)
* IPSEC_SPLASSERT(spl, msg)
* One further difference is the spl names:
* NetBSD splsoftnet equates to FreeBSD splnet;
* NetBSD splnet equates to FreeBSD splimp.
* which is hidden by the macro IPSEC_SPLASSERT_SOFTNET(msg).
*/
#ifdef __FreeBSD__
#define IPSEC_SPLASSERT(x,y) SPLASSERT(x, y)
#define IPSEC_ASSERT(c,m) KASSERT(c, m)
#define IPSEC_SPLASSERT_SOFTNET(m) SPLASSERT(splnet, m)
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
#define IPSEC_SPLASSERT(x,y) (void)0
#define IPSEC_ASSERT(c,m) KASSERT(c)
#define IPSEC_SPLASSERT_SOFTNET(m) IPSEC_SPLASSERT(softnet, m)
#endif /* __NetBSD__ */
/*
* 2. Kernel Randomness API.
* FreeBSD uses:
* u_int read_random(void *outbuf, int nbytes).
*/
#ifdef __FreeBSD__
#include <sys/random.h>
/* do nothing, use native random code. */
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
#include <sys/cprng.h>
static __inline u_int read_random(void *p, u_int len);
static __inline u_int
read_random(void *bufp, u_int len)
{
return cprng_fast(bufp, len);
}
#endif /* __NetBSD__ */
/*
* 3. Test for mbuf mutability
* FreeBSD 4.x uses: M_EXT_WRITABLE
* NetBSD has M_READONLY(). Use !M_READONLY().
* Not an exact match to FreeBSD semantics, but adequate for IPsec purposes.
*
*/
#ifdef __NetBSD__
/* XXX wrong, but close enough for restricted ipsec usage. */
#define M_EXT_WRITABLE(m) (!M_READONLY(m))
#endif /* __NetBSD__ */
/*
* 4. mbuf packet-header/packet-tag semantics.
*/
/*
* nothing.
*/
/*
* 5. Fast mbuf-cluster allocation.
*/
/*
* nothing.
*/
/*
* 6. Network output macros
* FreeBSD uses the IF_HANDOFF(), which raises SPL, enqueues
* a packet, and updates interface counters. NetBSD has IFQ_ENQUE(),
* which leaves SPL changes up to the caller.
* For now, we provide an emulation of IF_HANOOFF() which works
* for protocol input queues.
*/
#ifdef __FreeBSD__
/* nothing to do */
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
#define IF_HANDOFF(ifq, m, f) if_handoff(ifq, m, f, 0)
#include <net/if.h>
static __inline int
if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
{
int need_if_start = 0;
int s = splnet();
KERNEL_LOCK(1, NULL);
if (IF_QFULL(ifq)) {
IF_DROP(ifq);
KERNEL_UNLOCK_ONE(NULL);
splx(s);
m_freem(m);
return (0);
}
if (ifp != NULL) {
ifp->if_obytes += m->m_pkthdr.len + adjust;
if (m->m_flags & M_MCAST)
ifp->if_omcasts++;
need_if_start = !(ifp->if_flags & IFF_OACTIVE);
}
IF_ENQUEUE(ifq, m);
if (need_if_start)
(*ifp->if_start)(ifp);
KERNEL_UNLOCK_ONE(NULL);
splx(s);
return (1);
}
#endif /* __NetBSD__ */
/*
* 7. Elapsed Time: time_second as time in seconds.
* Original FreeBSD fast-ipsec code references a FreeBSD kernel global,
* time_second().
* XXX is this the right time scale - shouldn't we measure timeout/life times
* using a monotonic time scale (time_uptime, mono_time) - why if the FreeBSD
* base code using UTC based time for this ?
*/
/* protosw glue */
#ifdef __NetBSD__
#include <sys/protosw.h>
#define ipprotosw protosw
#endif /* __NetBSD__ */
/*
* 8. Test for "privileged" socket opened by superuser.
* FreeBSD tests ((so)->so_cred && (so)->so_cred.cr_uid == 0),
* NetBSD (1.6N) tests (so)->so_uid == 0).
* This difference is wrapped inside the IPSEC_PRIVILEGED_SO() macro.
*
*/
#ifdef __FreeBSD__
#define IPSEC_PRIVILEGED_SO(so) ((so)->so_cred && (so)->so_cred.cr_uid == 0)
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
/* superuser opened socket? */
#define IPSEC_PRIVILEGED_SO(so) ((so)->so_uidinfo->ui_uid == 0)
#endif /* __NetBSD__ */
/*
* 9. Raw socket list
* FreeBSD uses: listhead = rawcb_list, SLIST()-next field "list".
* NetBSD uses: listhead = rawcb, SLIST()-next field "list"
*
* This version of fast-ipsec source code uses rawcb_list as the head,
* and (to avoid namespace collisions) uses rcb_list as the "next" field.
*/
#ifdef __FreeBSD__
#define rcb_list list
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
#define rawcb_list rawcb
#endif /* __NetBSD__ */
/*
* 10. List of all known network interfaces.
* FreeBSD has listhead in_ifaddrhead, with ia_link as link.
* NetBSD has listhead in_ifaddr, with ia_list as link.
* No name-clahses, so just #define the appropriate names on NetBSD.
* NB: Is it worth introducing iterator (find-first-list/find-next-list)
* functions or macros to encapsulate these?
*/
#ifdef __FreeBSD__
/* nothing to do for raw interface list */
#endif /* FreeBSD */
#ifdef __NetBSD__
#define ia_link ia_list
#endif /* __NetBSD__ */
/*
* 11. Type of initialization functions.
*/
#ifdef __FreeBSD__
#define INITFN static
#endif
#ifdef __NetBSD__
#define INITFN extern
#endif
/* 12. On FreeBSD, ip_off assumed in host endian;
* it is converted (if necessary) by ip_input().
* On NetBSD, ip_off is in network byte order.
* We hide the difference with the macro IP_OFF_CONVERT
*/
#ifdef __FreeBSD__
#define IP_OFF_CONVERT(x) (x)
#endif
#ifdef __NetBSD__
#define IP_OFF_CONVERT(x) (htons(x))
#endif
/*
* 13. IPv6 support, and "generic" inpcb vs. IPv4 pcb vs. IPv6 pcb.
* To IPv6 V4-mapped addresses (and the KAME-derived implementation
* of IPv6 v4-mapped addresses) we must support limited polymorphism:
* partway down the stack we detect an IPv6 protocol address is really
* a mapped V4 address, and then start dispatching that address to
* native IPv4 PCB lookup. In KAME-derived IPsec (including fas-ipsec)
* some functions must handle arguments which (dynamically) may be either
* a IPv4 pcb (struct inpcb *) or an IPv6 pcb (struct in6pcb *).
*
* In FreeBSD 4.x, sgtrucr in6pcb is syntactic sugar for struct inpcb,
* so punning between struct inpcb* and struct in6pcb* is trivial.
* NetBSD until recently used completely different structs for IPv4
* and IPv6 PCBs. To simplify fast-ipsec coexisting with IPv6,
* NetBSD's struct inpcb and struct in6pcb were changed to both have
* common struct, struct inpcb_hdr, as their first member. NetBSD can
* thus pass arguments as struct inpcb_hdr*, and dispatch on a v4/v6
* flag in the inpcb_hdr at runtime.
*
* We hide the NetBSD-vs-FreeBSD differences inside the following abstraction:
*
* PCB_T: a macro name for a struct type which is used as a "generic"
* argument for actual arguments an in4pcb or an in6pcb.
*
* PCB_FAMILY(p): given a "generic" pcb_t p, returns the protocol
* family (AF_INET, AF_INET6) of the unperlying inpcb/in6pcb.
*
* PCB_SOCKET(p): given a "generic" pcb_t p, returns the associated
* socket pointer
*
* PCB_TO_IN4PCB(p): given generic pcb_t *p, returns a struct inpcb *
* PCB_TO_IN6PCB(p): given generic pcb_t *p, returns a struct in6pcb *
*
* IN4PCB_TO_PCB(inp): given a struct inpcb *inp, returns a pcb_t *
* IN6PCB_TO_PCB(in6p): given a struct in6pcb *in6p, returns a pcb_t *
*/
#ifdef __FreeBSD__
#define PCB_T struct inpcb
#define PCB_FAMILY(p) ((p)->inp_socket->so_proto->pr_domain->dom_family)
#define PCB_SOCKET(p) ((p)->inp_socket)
/* Convert generic pcb to IPv4/IPv6 pcb */
#define PCB_TO_IN4PCB(p) (p)
#define PCB_TO_IN6PCB(p) (p)
/* Convert IPv4/IPv6 pcb to generic pcb, for callers of fast-ipsec */
#define IN4PCB_TO_PCB(p) (p)
#define IN6PCB_TO_PCB(p) (p)
#endif /* __FreeBSD__ */
#ifdef __NetBSD__
#define PCB_T struct inpcb_hdr
#define PCB_FAMILY(p) ((p)->inph_af)
#define PCB_SOCKET(p) ((p)->inph_socket)
#define PCB_TO_IN4PCB(p) ((struct inpcb *)(p))
#define PCB_TO_IN6PCB(p) ((struct in6pcb *)(p))
#define IN4PCB_TO_PCB(p) ((PCB_T *)(&(p)->inp_head))
#define IN6PCB_TO_PCB(p) ((PCB_T *)(&(p)->in6p_head))
#endif /* __NetBSD__ */
/*
* Differences that we don't attempt to hide:
*
* A. Initialization code. This is the largest difference of all.
*
* FreeBSD uses compile/link-time perl hackery to generate special
* .o files with linker sections that give the moral equivalent of
* C++ file-level-object constructors. NetBSD has no such facility.
*
* Either we implement it (ideally, in a way that can emulate
* FreeBSD's SYSINIT() macros), or we must take other means
* to have the per-file init functions called at some appropriate time.
*
* In the absence of SYSINIT(), all the file-level init functions
* now have "extern" linkage. There is a new fast-ipsec init()
* function which calls each of the per-file in an appropriate order.
* init_main will arrange to call the fast-ipsec init function
* after the crypto framework has registered its transforms (including
* any autoconfigured hardware crypto accelerators) but before
* initializing the network stack to send or receive packet.
*
* B. Protosw() differences.
* CSRG-style BSD TCP/IP uses a generic protocol-dispatch-function
* where the specific request is identified by an enum argument.
* FreeBSD replaced that with an array of request-specific
* function pointers.
*
* These differences affect the handlers for key-protocol user requests
* so pervasively that I gave up on the fast-ipsec code, and re-worked the
* NetBSD KAME code to match the (relative few) API differences
* between NetBSD and FreeBSD's KAME netkey, and Fast-IPsec netkey.
*
* C. Timeout() versus callout(9):
* The FreeBSD 4.x netipsec/ code still uses timeout().
* FreeBSD 4.7 has callout(9), so I just replaced
* timeout_*() with the nearest callout_*() equivalents,
* and added a callout handle to the ipsec context.
*
* D. SPL name differences.
* FreeBSD splnet() equates directly to NetBSD's splsoftnet();
* FreeBSD uses splimp() where (for networking) NetBSD would use splnet().
*/
#endif /* _KERNEL */
#endif /* !_NETIPSEC_OSDEP_H_ */
Reference in New Issue View Git Blame Copy Permalink