summaryrefslogtreecommitdiffstats
path: root/contrib/ipfilter/mlfk_ipl.c
diff options
context:
space:
mode:
Diffstat (limited to 'contrib/ipfilter/mlfk_ipl.c')
-rw-r--r--contrib/ipfilter/mlfk_ipl.c529
1 files changed, 529 insertions, 0 deletions
diff --git a/contrib/ipfilter/mlfk_ipl.c b/contrib/ipfilter/mlfk_ipl.c
new file mode 100644
index 0000000..ba1f44f
--- /dev/null
+++ b/contrib/ipfilter/mlfk_ipl.c
@@ -0,0 +1,529 @@
+/* $FreeBSD$ */
+
+/*
+ * Copyright (C) 2012 by Darren Reed.
+ *
+ * See the IPFILTER.LICENCE file for details on licencing.
+ */
+
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/kernel.h>
+#include <sys/module.h>
+#include <sys/conf.h>
+#include <sys/socket.h>
+#include <sys/sysctl.h>
+#include <sys/select.h>
+#if __FreeBSD_version >= 500000
+# include <sys/selinfo.h>
+#endif
+#include <net/if.h>
+#include <netinet/in_systm.h>
+#include <netinet/in.h>
+
+
+#include "netinet/ipl.h"
+#include "netinet/ip_compat.h"
+#include "netinet/ip_fil.h"
+#include "netinet/ip_state.h"
+#include "netinet/ip_nat.h"
+#include "netinet/ip_auth.h"
+#include "netinet/ip_frag.h"
+#include "netinet/ip_sync.h"
+
+extern ipf_main_softc_t ipfmain;
+
+#if __FreeBSD_version >= 502116
+static struct cdev *ipf_devs[IPL_LOGSIZE];
+#else
+static dev_t ipf_devs[IPL_LOGSIZE];
+#endif
+
+#if 0
+static int sysctl_ipf_int ( SYSCTL_HANDLER_ARGS );
+#endif
+static int ipf_modload(void);
+static int ipf_modunload(void);
+
+#if (__FreeBSD_version >= 500024)
+# if (__FreeBSD_version >= 502116)
+static int ipfopen __P((struct cdev*, int, int, struct thread *));
+static int ipfclose __P((struct cdev*, int, int, struct thread *));
+# else
+static int ipfopen __P((dev_t, int, int, struct thread *));
+static int ipfclose __P((dev_t, int, int, struct thread *));
+# endif /* __FreeBSD_version >= 502116 */
+#else
+static int ipfopen __P((dev_t, int, int, struct proc *));
+static int ipfclose __P((dev_t, int, int, struct proc *));
+#endif
+#if (__FreeBSD_version >= 502116)
+static int ipfread __P((struct cdev*, struct uio *, int));
+static int ipfwrite __P((struct cdev*, struct uio *, int));
+#else
+static int ipfread __P((dev_t, struct uio *, int));
+static int ipfwrite __P((dev_t, struct uio *, int));
+#endif /* __FreeBSD_version >= 502116 */
+
+
+
+SYSCTL_DECL(_net_inet);
+#define SYSCTL_IPF(parent, nbr, name, access, ptr, val, descr) \
+ SYSCTL_OID(parent, nbr, name, CTLTYPE_INT|access, \
+ ptr, val, sysctl_ipf_int, "I", descr);
+#define CTLFLAG_OFF 0x00800000 /* IPFilter must be disabled */
+#define CTLFLAG_RWO (CTLFLAG_RW|CTLFLAG_OFF)
+SYSCTL_NODE(_net_inet, OID_AUTO, ipf, CTLFLAG_RW, 0, "IPF");
+#if 0
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_flags, CTLFLAG_RW, &ipf_flags, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_pass, CTLFLAG_RW, &ipf_pass, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_active, CTLFLAG_RD, &ipf_active, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpidletimeout, CTLFLAG_RWO,
+ &ipf_tcpidletimeout, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcphalfclosed, CTLFLAG_RWO,
+ &ipf_tcphalfclosed, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpclosewait, CTLFLAG_RWO,
+ &ipf_tcpclosewait, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcplastack, CTLFLAG_RWO,
+ &ipf_tcplastack, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcptimeout, CTLFLAG_RWO,
+ &ipf_tcptimeout, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpclosed, CTLFLAG_RWO,
+ &ipf_tcpclosed, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_udptimeout, CTLFLAG_RWO,
+ &ipf_udptimeout, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_udpacktimeout, CTLFLAG_RWO,
+ &ipf_udpacktimeout, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_icmptimeout, CTLFLAG_RWO,
+ &ipf_icmptimeout, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_defnatage, CTLFLAG_RWO,
+ &ipf_nat_defage, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_ipfrttl, CTLFLAG_RW,
+ &ipf_ipfrttl, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_running, CTLFLAG_RD,
+ &ipf_running, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_statesize, CTLFLAG_RWO,
+ &ipf_state_size, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_statemax, CTLFLAG_RWO,
+ &ipf_state_max, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_nattable_sz, CTLFLAG_RWO,
+ &ipf_nat_table_sz, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_natrules_sz, CTLFLAG_RWO,
+ &ipf_nat_maprules_sz, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_rdrrules_sz, CTLFLAG_RWO,
+ &ipf_nat_rdrrules_sz, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_hostmap_sz, CTLFLAG_RWO,
+ &ipf_nat_hostmap_sz, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_authsize, CTLFLAG_RWO,
+ &ipf_auth_size, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_authused, CTLFLAG_RD,
+ &ipf_auth_used, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_defaultauthage, CTLFLAG_RW,
+ &ipf_auth_defaultage, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_chksrc, CTLFLAG_RW, &ipf_chksrc, 0, "");
+SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_minttl, CTLFLAG_RW, &ipf_minttl, 0, "");
+#endif
+
+#define CDEV_MAJOR 79
+#include <sys/poll.h>
+#if __FreeBSD_version >= 500043
+# include <sys/select.h>
+static int ipfpoll(struct cdev *dev, int events, struct thread *td);
+
+static struct cdevsw ipf_cdevsw = {
+#if __FreeBSD_version >= 502103
+ .d_version = D_VERSION,
+ .d_flags = 0, /* D_NEEDGIANT - Should be SMP safe */
+#endif
+ .d_open = ipfopen,
+ .d_close = ipfclose,
+ .d_read = ipfread,
+ .d_write = ipfwrite,
+ .d_ioctl = ipfioctl,
+ .d_poll = ipfpoll,
+ .d_name = "ipf",
+#if __FreeBSD_version < 600000
+ .d_maj = CDEV_MAJOR,
+#endif
+};
+#else
+static int ipfpoll(dev_t dev, int events, struct proc *td);
+
+static struct cdevsw ipf_cdevsw = {
+ /* open */ ipfopen,
+ /* close */ ipfclose,
+ /* read */ ipfread,
+ /* write */ ipfwrite,
+ /* ioctl */ ipfioctl,
+ /* poll */ ipfpoll,
+ /* mmap */ nommap,
+ /* strategy */ nostrategy,
+ /* name */ "ipf",
+ /* maj */ CDEV_MAJOR,
+ /* dump */ nodump,
+ /* psize */ nopsize,
+ /* flags */ 0,
+# if (__FreeBSD_version < 500043)
+ /* bmaj */ -1,
+# endif
+# if (__FreeBSD_version >= 430000)
+ /* kqfilter */ NULL
+# endif
+};
+#endif
+
+static char *ipf_devfiles[] = { IPL_NAME, IPNAT_NAME, IPSTATE_NAME, IPAUTH_NAME,
+ IPSYNC_NAME, IPSCAN_NAME, IPLOOKUP_NAME, NULL };
+
+
+static int
+ipfilter_modevent(module_t mod, int type, void *unused)
+{
+ int error = 0;
+
+ switch (type)
+ {
+ case MOD_LOAD :
+ error = ipf_modload();
+ break;
+
+ case MOD_UNLOAD :
+ error = ipf_modunload();
+ break;
+ default:
+ error = EINVAL;
+ break;
+ }
+ return error;
+}
+
+
+static int
+ipf_modload()
+{
+ char *defpass, *c, *str;
+ int i, j, error;
+
+ if (ipf_load_all() != 0)
+ return EIO;
+
+ if (ipf_create_all(&ipfmain) == NULL)
+ return EIO;
+
+ error = ipfattach(&ipfmain);
+ if (error)
+ return error;
+
+ for (i = 0; i < IPL_LOGSIZE; i++)
+ ipf_devs[i] = NULL;
+
+ for (i = 0; (str = ipf_devfiles[i]); i++) {
+ c = NULL;
+ for(j = strlen(str); j > 0; j--)
+ if (str[j] == '/') {
+ c = str + j + 1;
+ break;
+ }
+ if (!c)
+ c = str;
+ ipf_devs[i] = make_dev(&ipf_cdevsw, i, 0, 0, 0600, c);
+ }
+
+ error = ipf_pfil_hook();
+ if (error != 0)
+ return error;
+ ipf_event_reg();
+
+ if (FR_ISPASS(ipfmain.ipf_pass))
+ defpass = "pass";
+ else if (FR_ISBLOCK(ipfmain.ipf_pass))
+ defpass = "block";
+ else
+ defpass = "no-match -> block";
+
+ printf("%s initialized. Default = %s all, Logging = %s%s\n",
+ ipfilter_version, defpass,
+#ifdef IPFILTER_LOG
+ "enabled",
+#else
+ "disabled",
+#endif
+#ifdef IPFILTER_COMPILED
+ " (COMPILED)"
+#else
+ ""
+#endif
+ );
+ return 0;
+}
+
+
+static int
+ipf_modunload()
+{
+ int error, i;
+
+ if (ipfmain.ipf_refcnt)
+ return EBUSY;
+
+ error = ipf_pfil_unhook();
+ if (error != 0)
+ return error;
+
+ if (ipfmain.ipf_running >= 0) {
+ error = ipfdetach(&ipfmain);
+ if (error != 0)
+ return error;
+
+ ipf_destroy_all(&ipfmain);
+ ipf_unload_all();
+ } else
+ error = 0;
+
+ ipfmain.ipf_running = -2;
+
+ for (i = 0; ipf_devfiles[i]; i++) {
+ if (ipf_devs[i] != NULL)
+ destroy_dev(ipf_devs[i]);
+ }
+
+ printf("%s unloaded\n", ipfilter_version);
+
+ return error;
+}
+
+
+static moduledata_t ipfiltermod = {
+ "ipfilter",
+ ipfilter_modevent,
+ 0
+};
+
+
+DECLARE_MODULE(ipfilter, ipfiltermod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY);
+#ifdef MODULE_VERSION
+MODULE_VERSION(ipfilter, 1);
+#endif
+
+
+#if 0
+#ifdef SYSCTL_IPF
+int
+sysctl_ipf_int ( SYSCTL_HANDLER_ARGS )
+{
+ int error = 0;
+
+ if (arg1)
+ error = SYSCTL_OUT(req, arg1, sizeof(int));
+ else
+ error = SYSCTL_OUT(req, &arg2, sizeof(int));
+
+ if (error || !req->newptr)
+ return (error);
+
+ if (!arg1)
+ error = EPERM;
+ else {
+ if ((oidp->oid_kind & CTLFLAG_OFF) && (ipfmain.ipf_running > 0))
+ error = EBUSY;
+ else
+ error = SYSCTL_IN(req, arg1, sizeof(int));
+ }
+ return (error);
+}
+#endif
+#endif
+
+
+static int
+#if __FreeBSD_version >= 500043
+ipfpoll(struct cdev *dev, int events, struct thread *td)
+#else
+ipfpoll(dev_t dev, int events, struct proc *td)
+#endif
+{
+ u_int unit = GET_MINOR(dev);
+ int revents;
+
+ if (unit < 0 || unit > IPL_LOGMAX)
+ return 0;
+
+ revents = 0;
+
+ switch (unit)
+ {
+ case IPL_LOGIPF :
+ case IPL_LOGNAT :
+ case IPL_LOGSTATE :
+#ifdef IPFILTER_LOG
+ if ((events & (POLLIN | POLLRDNORM)) && ipf_log_canread(&ipfmain, unit))
+ revents |= events & (POLLIN | POLLRDNORM);
+#endif
+ break;
+ case IPL_LOGAUTH :
+ if ((events & (POLLIN | POLLRDNORM)) && ipf_auth_waiting(&ipfmain))
+ revents |= events & (POLLIN | POLLRDNORM);
+ break;
+ case IPL_LOGSYNC :
+ if ((events & (POLLIN | POLLRDNORM)) && ipf_sync_canread(&ipfmain))
+ revents |= events & (POLLIN | POLLRDNORM);
+ if ((events & (POLLOUT | POLLWRNORM)) && ipf_sync_canwrite(&ipfmain))
+ revents |= events & (POLLOUT | POLLWRNORM);
+ break;
+ case IPL_LOGSCAN :
+ case IPL_LOGLOOKUP :
+ default :
+ break;
+ }
+
+ if ((revents == 0) && ((events & (POLLIN|POLLRDNORM)) != 0))
+ selrecord(td, &ipfmain.ipf_selwait[unit]);
+
+ return revents;
+}
+
+
+/*
+ * routines below for saving IP headers to buffer
+ */
+static int ipfopen(dev, flags
+#if ((BSD >= 199506) || (__FreeBSD_version >= 220000))
+, devtype, p)
+ int devtype;
+# if (__FreeBSD_version >= 500024)
+ struct thread *p;
+# else
+ struct proc *p;
+# endif /* __FreeBSD_version >= 500024 */
+#else
+)
+#endif
+#if (__FreeBSD_version >= 502116)
+ struct cdev *dev;
+#else
+ dev_t dev;
+#endif
+ int flags;
+{
+ u_int unit = GET_MINOR(dev);
+ int error;
+
+ if (IPL_LOGMAX < unit)
+ error = ENXIO;
+ else {
+ switch (unit)
+ {
+ case IPL_LOGIPF :
+ case IPL_LOGNAT :
+ case IPL_LOGSTATE :
+ case IPL_LOGAUTH :
+ case IPL_LOGLOOKUP :
+ case IPL_LOGSYNC :
+#ifdef IPFILTER_SCAN
+ case IPL_LOGSCAN :
+#endif
+ error = 0;
+ break;
+ default :
+ error = ENXIO;
+ break;
+ }
+ }
+ return error;
+}
+
+
+static int ipfclose(dev, flags
+#if ((BSD >= 199506) || (__FreeBSD_version >= 220000))
+, devtype, p)
+ int devtype;
+# if (__FreeBSD_version >= 500024)
+ struct thread *p;
+# else
+ struct proc *p;
+# endif /* __FreeBSD_version >= 500024 */
+#else
+)
+#endif
+#if (__FreeBSD_version >= 502116)
+ struct cdev *dev;
+#else
+ dev_t dev;
+#endif
+ int flags;
+{
+ u_int unit = GET_MINOR(dev);
+
+ if (IPL_LOGMAX < unit)
+ unit = ENXIO;
+ else
+ unit = 0;
+ return unit;
+}
+
+/*
+ * ipfread/ipflog
+ * both of these must operate with at least splnet() lest they be
+ * called during packet processing and cause an inconsistancy to appear in
+ * the filter lists.
+ */
+#if (BSD >= 199306)
+static int ipfread(dev, uio, ioflag)
+ int ioflag;
+#else
+static int ipfread(dev, uio)
+#endif
+#if (__FreeBSD_version >= 502116)
+ struct cdev *dev;
+#else
+ dev_t dev;
+#endif
+ struct uio *uio;
+{
+ u_int unit = GET_MINOR(dev);
+
+ if (unit < 0)
+ return ENXIO;
+
+ if (ipfmain.ipf_running < 1)
+ return EIO;
+
+ if (unit == IPL_LOGSYNC)
+ return ipf_sync_read(&ipfmain, uio);
+
+#ifdef IPFILTER_LOG
+ return ipf_log_read(&ipfmain, unit, uio);
+#else
+ return ENXIO;
+#endif
+}
+
+
+/*
+ * ipfwrite
+ * both of these must operate with at least splnet() lest they be
+ * called during packet processing and cause an inconsistancy to appear in
+ * the filter lists.
+ */
+#if (BSD >= 199306)
+static int ipfwrite(dev, uio, ioflag)
+ int ioflag;
+#else
+static int ipfwrite(dev, uio)
+#endif
+#if (__FreeBSD_version >= 502116)
+ struct cdev *dev;
+#else
+ dev_t dev;
+#endif
+ struct uio *uio;
+{
+
+ if (ipfmain.ipf_running < 1)
+ return EIO;
+
+ if (GET_MINOR(dev) == IPL_LOGSYNC)
+ return ipf_sync_write(&ipfmain, uio);
+ return ENXIO;
+}
OpenPOWER on IntegriCloud