[mirror_frr.git] / zebra / zebra_mpls_openbsd.c

/*
 * Copyright (C) 2016 by Open Source Routing.
 *
 * This file is part of GNU Zebra.
 *
 * GNU Zebra is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the
 * Free Software Foundation; either version 2, or (at your option) any
 * later version.
 *
 * GNU Zebra is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; see the file COPYING; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include <zebra.h>

#ifdef OPEN_BSD

#include <netmpls/mpls.h>
#include "zebra/rt.h"
#include "zebra/zebra_mpls.h"
#include "zebra/debug.h"

#include "privs.h"
#include "prefix.h"
#include "interface.h"
#include "log.h"
#include "lib_errors.h"

extern struct zebra_privs_t zserv_privs;

struct {
	uint32_t rtseq;
	int fd;
	int ioctl_fd;
} kr_state;

static int kernel_send_rtmsg_v4(int action, mpls_label_t in_label,
				zebra_nhlfe_t *nhlfe)
{
	struct iovec iov[5];
	struct rt_msghdr hdr;
	struct sockaddr_mpls sa_label_in, sa_label_out;
	struct sockaddr_in nexthop;
	int iovcnt = 0;
	int ret;

	if (IS_ZEBRA_DEBUG_KERNEL)
		zlog_debug("%s: 0x%x, label=%u", __func__, action, in_label);

	/* initialize header */
	memset(&hdr, 0, sizeof(hdr));
	hdr.rtm_version = RTM_VERSION;

	hdr.rtm_type = action;
	hdr.rtm_flags = RTF_UP;
	hdr.rtm_fmask = RTF_MPLS;
	hdr.rtm_seq = kr_state.rtseq++; /* overflow doesn't matter */
	hdr.rtm_msglen = sizeof(hdr);
	hdr.rtm_hdrlen = sizeof(struct rt_msghdr);
	hdr.rtm_priority = 0;
	/* adjust iovec */
	iov[iovcnt].iov_base = &hdr;
	iov[iovcnt++].iov_len = sizeof(hdr);

	/* in label */
	memset(&sa_label_in, 0, sizeof(sa_label_in));
	sa_label_in.smpls_len = sizeof(sa_label_in);
	sa_label_in.smpls_family = AF_MPLS;
	sa_label_in.smpls_label = htonl(in_label << MPLS_LABEL_OFFSET);
	/* adjust header */
	hdr.rtm_flags |= RTF_MPLS | RTF_MPATH;
	hdr.rtm_addrs |= RTA_DST;
	hdr.rtm_msglen += sizeof(sa_label_in);
	/* adjust iovec */
	iov[iovcnt].iov_base = &sa_label_in;
	iov[iovcnt++].iov_len = sizeof(sa_label_in);

	/* nexthop */
	memset(&nexthop, 0, sizeof(nexthop));
	nexthop.sin_len = sizeof(nexthop);
	nexthop.sin_family = AF_INET;
	nexthop.sin_addr = nhlfe->nexthop->gate.ipv4;
	/* adjust header */
	hdr.rtm_flags |= RTF_GATEWAY;
	hdr.rtm_addrs |= RTA_GATEWAY;
	hdr.rtm_msglen += sizeof(nexthop);
	/* adjust iovec */
	iov[iovcnt].iov_base = &nexthop;
	iov[iovcnt++].iov_len = sizeof(nexthop);

	/* If action is RTM_DELETE we have to get rid of MPLS infos */
	if (action != RTM_DELETE) {
		memset(&sa_label_out, 0, sizeof(sa_label_out));
		sa_label_out.smpls_len = sizeof(sa_label_out);
		sa_label_out.smpls_family = AF_MPLS;
		sa_label_out.smpls_label =
			htonl(nhlfe->nexthop->nh_label->label[0]
			      << MPLS_LABEL_OFFSET);
		/* adjust header */
		hdr.rtm_addrs |= RTA_SRC;
		hdr.rtm_flags |= RTF_MPLS;
		hdr.rtm_msglen += sizeof(sa_label_out);
		/* adjust iovec */
		iov[iovcnt].iov_base = &sa_label_out;
		iov[iovcnt++].iov_len = sizeof(sa_label_out);

		if (nhlfe->nexthop->nh_label->label[0] == MPLS_LABEL_IMPLNULL)
			hdr.rtm_mpls = MPLS_OP_POP;
		else
			hdr.rtm_mpls = MPLS_OP_SWAP;
	}

	if (zserv_privs.change(ZPRIVS_RAISE))
		flog_err(LIB_ERR_PRIVILEGES, "Can't raise privileges");
	ret = writev(kr_state.fd, iov, iovcnt);
	if (zserv_privs.change(ZPRIVS_LOWER))
		flog_err(LIB_ERR_PRIVILEGES, "Can't lower privileges");

	if (ret == -1)
		flog_err_sys(LIB_ERR_SOCKET, "%s: %s", __func__,
			     safe_strerror(errno));

	return ret;
}

#if !defined(ROUNDUP)
#define ROUNDUP(a)                                                             \
	(((a) & (sizeof(long) - 1)) ? (1 + ((a) | (sizeof(long) - 1))) : (a))
#endif

static int kernel_send_rtmsg_v6(int action, mpls_label_t in_label,
				zebra_nhlfe_t *nhlfe)
{
	struct iovec iov[5];
	struct rt_msghdr hdr;
	struct sockaddr_mpls sa_label_in, sa_label_out;
	struct pad {
		struct sockaddr_in6 addr;
		char pad[sizeof(long)]; /* thank you IPv6 */
	} nexthop;
	int iovcnt = 0;
	int ret;

	if (IS_ZEBRA_DEBUG_KERNEL)
		zlog_debug("%s: 0x%x, label=%u", __func__, action, in_label);

	/* initialize header */
	memset(&hdr, 0, sizeof(hdr));
	hdr.rtm_version = RTM_VERSION;

	hdr.rtm_type = action;
	hdr.rtm_flags = RTF_UP;
	hdr.rtm_fmask = RTF_MPLS;
	hdr.rtm_seq = kr_state.rtseq++; /* overflow doesn't matter */
	hdr.rtm_msglen = sizeof(hdr);
	hdr.rtm_hdrlen = sizeof(struct rt_msghdr);
	hdr.rtm_priority = 0;
	/* adjust iovec */
	iov[iovcnt].iov_base = &hdr;
	iov[iovcnt++].iov_len = sizeof(hdr);

	/* in label */
	memset(&sa_label_in, 0, sizeof(sa_label_in));
	sa_label_in.smpls_len = sizeof(sa_label_in);
	sa_label_in.smpls_family = AF_MPLS;
	sa_label_in.smpls_label = htonl(in_label << MPLS_LABEL_OFFSET);
	/* adjust header */
	hdr.rtm_flags |= RTF_MPLS | RTF_MPATH;
	hdr.rtm_addrs |= RTA_DST;
	hdr.rtm_msglen += sizeof(sa_label_in);
	/* adjust iovec */
	iov[iovcnt].iov_base = &sa_label_in;
	iov[iovcnt++].iov_len = sizeof(sa_label_in);

	/* nexthop */
	memset(&nexthop, 0, sizeof(nexthop));
	nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
	nexthop.addr.sin6_family = AF_INET6;
	nexthop.addr.sin6_addr = nhlfe->nexthop->gate.ipv6;
	if (IN6_IS_ADDR_LINKLOCAL(&nexthop.addr.sin6_addr)) {
		uint16_t tmp16;
		struct sockaddr_in6 *sin6 = &nexthop.addr;

		nexthop.addr.sin6_scope_id = nhlfe->nexthop->ifindex;

		memcpy(&tmp16, &sin6->sin6_addr.s6_addr[2], sizeof(tmp16));
		tmp16 = htons(sin6->sin6_scope_id);
		memcpy(&sin6->sin6_addr.s6_addr[2], &tmp16, sizeof(tmp16));
		sin6->sin6_scope_id = 0;
	}

	/* adjust header */
	hdr.rtm_flags |= RTF_GATEWAY;
	hdr.rtm_addrs |= RTA_GATEWAY;
	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
	/* adjust iovec */
	iov[iovcnt].iov_base = &nexthop;
	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));

	/* If action is RTM_DELETE we have to get rid of MPLS infos */
	if (action != RTM_DELETE) {
		memset(&sa_label_out, 0, sizeof(sa_label_out));
		sa_label_out.smpls_len = sizeof(sa_label_out);
		sa_label_out.smpls_family = AF_MPLS;
		sa_label_out.smpls_label =
			htonl(nhlfe->nexthop->nh_label->label[0]
			      << MPLS_LABEL_OFFSET);
		/* adjust header */
		hdr.rtm_addrs |= RTA_SRC;
		hdr.rtm_flags |= RTF_MPLS;
		hdr.rtm_msglen += sizeof(sa_label_out);
		/* adjust iovec */
		iov[iovcnt].iov_base = &sa_label_out;
		iov[iovcnt++].iov_len = sizeof(sa_label_out);

		if (nhlfe->nexthop->nh_label->label[0] == MPLS_LABEL_IMPLNULL)
			hdr.rtm_mpls = MPLS_OP_POP;
		else
			hdr.rtm_mpls = MPLS_OP_SWAP;
	}

	if (zserv_privs.change(ZPRIVS_RAISE))
		flog_err(LIB_ERR_PRIVILEGES, "Can't raise privileges");
	ret = writev(kr_state.fd, iov, iovcnt);
	if (zserv_privs.change(ZPRIVS_LOWER))
		flog_err(LIB_ERR_PRIVILEGES, "Can't lower privileges");

	if (ret == -1)
		flog_err_sys(LIB_ERR_SOCKET, "%s: %s", __func__,
			     safe_strerror(errno));

	return ret;
}

static int kernel_lsp_cmd(int action, zebra_lsp_t *lsp)
{
	zebra_nhlfe_t *nhlfe;
	struct nexthop *nexthop = NULL;
	unsigned int nexthop_num = 0;

	for (nhlfe = lsp->nhlfe_list; nhlfe; nhlfe = nhlfe->next) {
		nexthop = nhlfe->nexthop;
		if (!nexthop)
			continue;

		if (nexthop_num >= multipath_num)
			break;

		if (((action == RTM_ADD || action == RTM_CHANGE)
		     && (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
			 && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)))
		    || (action == RTM_DELETE
			&& (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_INSTALLED)
			    && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)))) {
			if (nhlfe->nexthop->nh_label->num_labels > 1) {
				zlog_warn(
					"%s: can't push %u labels at once "
					"(maximum is 1)",
					__func__,
					nhlfe->nexthop->nh_label->num_labels);
				continue;
			}

			nexthop_num++;

			switch (NHLFE_FAMILY(nhlfe)) {
			case AF_INET:
				kernel_send_rtmsg_v4(action, lsp->ile.in_label,
						     nhlfe);
				break;
			case AF_INET6:
				kernel_send_rtmsg_v6(action, lsp->ile.in_label,
						     nhlfe);
				break;
			default:
				break;
			}
		}
	}

	return (0);
}

enum dp_req_result kernel_add_lsp(zebra_lsp_t *lsp)
{
	int ret;

	if (!lsp || !lsp->best_nhlfe) { // unexpected
		kernel_lsp_pass_fail(lsp, DP_INSTALL_FAILURE);
		return DP_REQUEST_FAILURE;
	}

	ret = kernel_lsp_cmd(RTM_ADD, lsp);

	kernel_lsp_pass_fail(lsp,
			     (!ret) ? DP_INSTALL_SUCCESS
				    : DP_INSTALL_FAILURE);

	return DP_REQUEST_SUCCESS;
}

enum dp_req_result kernel_upd_lsp(zebra_lsp_t *lsp)
{
	int ret;

	if (!lsp || !lsp->best_nhlfe) { // unexpected
		kernel_lsp_pass_fail(lsp, DP_INSTALL_FAILURE);
		return DP_REQUEST_FAILURE;
	}

	ret = kernel_lsp_cmd(RTM_CHANGE, lsp);

	kernel_lsp_pass_fail(lsp,
			     (!ret) ? DP_INSTALL_SUCCESS
				    : DP_INSTALL_FAILURE);
	return DP_REQUEST_SUCCESS;
}

enum dp_req_result kernel_del_lsp(zebra_lsp_t *lsp)
{
	int ret;

	if (!lsp) { // unexpected
		kernel_lsp_pass_fail(lsp, DP_DELETE_FAILURE);
		return DP_REQUEST_FAILURE;
	}

	if (!CHECK_FLAG(lsp->flags, LSP_FLAG_INSTALLED)) {
		kernel_lsp_pass_fail(lsp, DP_DELETE_FAILURE);
		return DP_REQUEST_FAILURE;
	}

	ret = kernel_lsp_cmd(RTM_DELETE, lsp);

	kernel_lsp_pass_fail(lsp,
			     (!ret) ? DP_DELETE_SUCCESS
				    : DP_DELETE_FAILURE);

	return DP_REQUEST_SUCCESS;
}

static int kmpw_install(struct zebra_pw *pw)
{
	struct ifreq ifr;
	struct ifmpwreq imr;
	struct sockaddr_storage ss;
	struct sockaddr_in *sa_in = (struct sockaddr_in *)&ss;
	struct sockaddr_in6 *sa_in6 = (struct sockaddr_in6 *)&ss;

	memset(&imr, 0, sizeof(imr));
	switch (pw->type) {
	case PW_TYPE_ETHERNET:
		imr.imr_type = IMR_TYPE_ETHERNET;
		break;
	case PW_TYPE_ETHERNET_TAGGED:
		imr.imr_type = IMR_TYPE_ETHERNET_TAGGED;
		break;
	default:
		zlog_warn("%s: unhandled pseudowire type (%#X)", __func__,
			  pw->type);
		return -1;
	}

	if (pw->flags & F_PSEUDOWIRE_CWORD)
		imr.imr_flags |= IMR_FLAG_CONTROLWORD;

	/* pseudowire nexthop */
	memset(&ss, 0, sizeof(ss));
	switch (pw->af) {
	case AF_INET:
		sa_in->sin_family = AF_INET;
		sa_in->sin_len = sizeof(struct sockaddr_in);
		sa_in->sin_addr = pw->nexthop.ipv4;
		break;
	case AF_INET6:
		sa_in6->sin6_family = AF_INET6;
		sa_in6->sin6_len = sizeof(struct sockaddr_in6);
		sa_in6->sin6_addr = pw->nexthop.ipv6;
		break;
	default:
		zlog_warn("%s: unhandled pseudowire address-family (%u)",
			  __func__, pw->af);
		return -1;
	}
	memcpy(&imr.imr_nexthop, (struct sockaddr *)&ss,
	       sizeof(imr.imr_nexthop));

	/* pseudowire local/remote labels */
	imr.imr_lshim.shim_label = pw->local_label;
	imr.imr_rshim.shim_label = pw->remote_label;

	/* ioctl */
	memset(&ifr, 0, sizeof(ifr));
	strlcpy(ifr.ifr_name, pw->ifname, sizeof(ifr.ifr_name));
	ifr.ifr_data = (caddr_t)&imr;
	if (ioctl(kr_state.ioctl_fd, SIOCSETMPWCFG, &ifr) == -1) {
		flog_err_sys(LIB_ERR_SYSTEM_CALL, "ioctl SIOCSETMPWCFG: %s",
			     safe_strerror(errno));
		return -1;
	}

	return 0;
}

static int kmpw_uninstall(struct zebra_pw *pw)
{
	struct ifreq ifr;
	struct ifmpwreq imr;

	memset(&ifr, 0, sizeof(ifr));
	memset(&imr, 0, sizeof(imr));
	strlcpy(ifr.ifr_name, pw->ifname, sizeof(ifr.ifr_name));
	ifr.ifr_data = (caddr_t)&imr;
	if (ioctl(kr_state.ioctl_fd, SIOCSETMPWCFG, &ifr) == -1) {
		flog_err_sys(LIB_ERR_SYSTEM_CALL, "ioctl SIOCSETMPWCFG: %s",
			     safe_strerror(errno));
		return -1;
	}

	return 0;
}

#define MAX_RTSOCK_BUF	128 * 1024
int mpls_kernel_init(void)
{
	int rcvbuf, default_rcvbuf;
	socklen_t optlen;

	if ((kr_state.fd = socket(AF_ROUTE, SOCK_RAW, 0)) == -1) {
		zlog_warn("%s: socket", __func__);
		return -1;
	}

	if ((kr_state.ioctl_fd = socket(AF_INET, SOCK_DGRAM | SOCK_NONBLOCK, 0))
	    == -1) {
		zlog_warn("%s: ioctl socket", __func__);
		return -1;
	}

	/* grow receive buffer, don't wanna miss messages */
	optlen = sizeof(default_rcvbuf);
	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, &default_rcvbuf,
		       &optlen)
	    == -1)
		zlog_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
	else
		for (rcvbuf = MAX_RTSOCK_BUF;
		     rcvbuf > default_rcvbuf
		     && setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf,
				   sizeof(rcvbuf))
				== -1
		     && errno == ENOBUFS;
		     rcvbuf /= 2)
			; /* nothing */

	kr_state.rtseq = 1;

	/* register hook to install/uninstall pseudowires */
	hook_register(pw_install, kmpw_install);
	hook_register(pw_uninstall, kmpw_uninstall);

	return 0;
}

#endif /* OPEN_BSD */
Commit	Line	Data
7fc02572 RW	1	/*
	2	* Copyright (C) 2016 by Open Source Routing.
	3	*
	4	* This file is part of GNU Zebra.
	5	*
	6	* GNU Zebra is free software; you can redistribute it and/or modify it
	7	* under the terms of the GNU General Public License as published by the
	8	* Free Software Foundation; either version 2, or (at your option) any
	9	* later version.
	10	*
	11	* GNU Zebra is distributed in the hope that it will be useful, but
	12	* WITHOUT ANY WARRANTY; without even the implied warranty of
	13	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	14	* General Public License for more details.
	15	*
896014f4 DL	16	* You should have received a copy of the GNU General Public License along
	17	* with this program; see the file COPYING; if not, write to the Free Software
	18	* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
7fc02572 RW	19	*/
7fc02572 RW	20
d3e2c74a	21	#include <zebra.h>
ddfeb486 DL	22
	23	#ifdef OPEN_BSD
	24
d3e2c74a RW	25	#include <netmpls/mpls.h>
	26	#include "zebra/rt.h"
	27	#include "zebra/zebra_mpls.h"
	28	#include "zebra/debug.h"
	29
	30	#include "privs.h"
	31	#include "prefix.h"
	32	#include "interface.h"
	33	#include "log.h"
174482ef	34	#include "lib_errors.h"
d3e2c74a RW	35
	36	extern struct zebra_privs_t zserv_privs;
	37
	38	struct {
d7c0a89a	39	uint32_t rtseq;
d62a17ae	40	int fd;
d42127da	41	int ioctl_fd;
d3e2c74a RW	42	} kr_state;
d3e2c74a RW	43
d62a17ae	44	static int kernel_send_rtmsg_v4(int action, mpls_label_t in_label,
d62a17ae	45	zebra_nhlfe_t *nhlfe)
d3e2c74a	46	{
d62a17ae	47	struct iovec iov[5];
	48	struct rt_msghdr hdr;
	49	struct sockaddr_mpls sa_label_in, sa_label_out;
	50	struct sockaddr_in nexthop;
	51	int iovcnt = 0;
	52	int ret;
	53
	54	if (IS_ZEBRA_DEBUG_KERNEL)
	55	zlog_debug("%s: 0x%x, label=%u", __func__, action, in_label);
	56
	57	/* initialize header */
	58	memset(&hdr, 0, sizeof(hdr));
	59	hdr.rtm_version = RTM_VERSION;
	60
	61	hdr.rtm_type = action;
	62	hdr.rtm_flags = RTF_UP;
	63	hdr.rtm_fmask = RTF_MPLS;
	64	hdr.rtm_seq = kr_state.rtseq++; /* overflow doesn't matter */
	65	hdr.rtm_msglen = sizeof(hdr);
	66	hdr.rtm_hdrlen = sizeof(struct rt_msghdr);
	67	hdr.rtm_priority = 0;
	68	/* adjust iovec */
	69	iov[iovcnt].iov_base = &hdr;
	70	iov[iovcnt++].iov_len = sizeof(hdr);
	71
	72	/* in label */
	73	memset(&sa_label_in, 0, sizeof(sa_label_in));
	74	sa_label_in.smpls_len = sizeof(sa_label_in);
	75	sa_label_in.smpls_family = AF_MPLS;
	76	sa_label_in.smpls_label = htonl(in_label << MPLS_LABEL_OFFSET);
	77	/* adjust header */
	78	hdr.rtm_flags \|= RTF_MPLS \| RTF_MPATH;
	79	hdr.rtm_addrs \|= RTA_DST;
	80	hdr.rtm_msglen += sizeof(sa_label_in);
	81	/* adjust iovec */
	82	iov[iovcnt].iov_base = &sa_label_in;
	83	iov[iovcnt++].iov_len = sizeof(sa_label_in);
	84
	85	/* nexthop */
	86	memset(&nexthop, 0, sizeof(nexthop));
	87	nexthop.sin_len = sizeof(nexthop);
	88	nexthop.sin_family = AF_INET;
	89	nexthop.sin_addr = nhlfe->nexthop->gate.ipv4;
	90	/* adjust header */
	91	hdr.rtm_flags \|= RTF_GATEWAY;
	92	hdr.rtm_addrs \|= RTA_GATEWAY;
	93	hdr.rtm_msglen += sizeof(nexthop);
	94	/* adjust iovec */
	95	iov[iovcnt].iov_base = &nexthop;
	96	iov[iovcnt++].iov_len = sizeof(nexthop);
	97
	98	/* If action is RTM_DELETE we have to get rid of MPLS infos */
	99	if (action != RTM_DELETE) {
	100	memset(&sa_label_out, 0, sizeof(sa_label_out));
	101	sa_label_out.smpls_len = sizeof(sa_label_out);
	102	sa_label_out.smpls_family = AF_MPLS;
	103	sa_label_out.smpls_label =
	104	htonl(nhlfe->nexthop->nh_label->label[0]
	105	<< MPLS_LABEL_OFFSET);
	106	/* adjust header */
	107	hdr.rtm_addrs \|= RTA_SRC;
	108	hdr.rtm_flags \|= RTF_MPLS;
	109	hdr.rtm_msglen += sizeof(sa_label_out);
	110	/* adjust iovec */
111	iov[iovcnt].iov_base = &sa_label_out;
112	iov[iovcnt++].iov_len = sizeof(sa_label_out);
113
114	if (nhlfe->nexthop->nh_label->label[0] == MPLS_LABEL_IMPLNULL)
115	hdr.rtm_mpls = MPLS_OP_POP;
116	else
117	hdr.rtm_mpls = MPLS_OP_SWAP;
118	}
119
120	if (zserv_privs.change(ZPRIVS_RAISE))
af4c2728	121	flog_err(LIB_ERR_PRIVILEGES, "Can't raise privileges");
d62a17ae	122	ret = writev(kr_state.fd, iov, iovcnt);
d62a17ae	123	if (zserv_privs.change(ZPRIVS_LOWER))
af4c2728	124	flog_err(LIB_ERR_PRIVILEGES, "Can't lower privileges");
d62a17ae	125
d62a17ae	126	if (ret == -1)
09c866e3 QY	127	flog_err_sys(LIB_ERR_SOCKET, "%s: %s", __func__,
09c866e3 QY	128	safe_strerror(errno));
d62a17ae	129
d62a17ae	130	return ret;
e07486ce RW	131	}
	132
	133	#if !defined(ROUNDUP)
d62a17ae	134	#define ROUNDUP(a) \
d62a17ae	135	(((a) & (sizeof(long) - 1)) ? (1 + ((a) \| (sizeof(long) - 1))) : (a))
e07486ce RW	136	#endif
e07486ce RW	137
d62a17ae	138	static int kernel_send_rtmsg_v6(int action, mpls_label_t in_label,
d62a17ae	139	zebra_nhlfe_t *nhlfe)
e07486ce	140	{
d62a17ae	141	struct iovec iov[5];
	142	struct rt_msghdr hdr;
	143	struct sockaddr_mpls sa_label_in, sa_label_out;
	144	struct pad {
	145	struct sockaddr_in6 addr;
	146	char pad[sizeof(long)]; /* thank you IPv6 */
	147	} nexthop;
	148	int iovcnt = 0;
	149	int ret;
	150
	151	if (IS_ZEBRA_DEBUG_KERNEL)
	152	zlog_debug("%s: 0x%x, label=%u", __func__, action, in_label);
	153
	154	/* initialize header */
	155	memset(&hdr, 0, sizeof(hdr));
	156	hdr.rtm_version = RTM_VERSION;
	157
	158	hdr.rtm_type = action;
	159	hdr.rtm_flags = RTF_UP;
	160	hdr.rtm_fmask = RTF_MPLS;
	161	hdr.rtm_seq = kr_state.rtseq++; /* overflow doesn't matter */
	162	hdr.rtm_msglen = sizeof(hdr);
	163	hdr.rtm_hdrlen = sizeof(struct rt_msghdr);
	164	hdr.rtm_priority = 0;
	165	/* adjust iovec */
	166	iov[iovcnt].iov_base = &hdr;
	167	iov[iovcnt++].iov_len = sizeof(hdr);
	168
	169	/* in label */
	170	memset(&sa_label_in, 0, sizeof(sa_label_in));
	171	sa_label_in.smpls_len = sizeof(sa_label_in);
	172	sa_label_in.smpls_family = AF_MPLS;
	173	sa_label_in.smpls_label = htonl(in_label << MPLS_LABEL_OFFSET);
	174	/* adjust header */
	175	hdr.rtm_flags \|= RTF_MPLS \| RTF_MPATH;
	176	hdr.rtm_addrs \|= RTA_DST;
	177	hdr.rtm_msglen += sizeof(sa_label_in);
	178	/* adjust iovec */
	179	iov[iovcnt].iov_base = &sa_label_in;
	180	iov[iovcnt++].iov_len = sizeof(sa_label_in);
	181
	182	/* nexthop */
	183	memset(&nexthop, 0, sizeof(nexthop));
	184	nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
	185	nexthop.addr.sin6_family = AF_INET6;
	186	nexthop.addr.sin6_addr = nhlfe->nexthop->gate.ipv6;
	187	if (IN6_IS_ADDR_LINKLOCAL(&nexthop.addr.sin6_addr)) {
	188	uint16_t tmp16;
	189	struct sockaddr_in6 *sin6 = &nexthop.addr;
	190
	191	nexthop.addr.sin6_scope_id = nhlfe->nexthop->ifindex;
	192
	193	memcpy(&tmp16, &sin6->sin6_addr.s6_addr[2], sizeof(tmp16));
	194	tmp16 = htons(sin6->sin6_scope_id);
	195	memcpy(&sin6->sin6_addr.s6_addr[2], &tmp16, sizeof(tmp16));
	196	sin6->sin6_scope_id = 0;
	197	}
	198
	199	/* adjust header */
	200	hdr.rtm_flags \|= RTF_GATEWAY;
	201	hdr.rtm_addrs \|= RTA_GATEWAY;
	202	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
	203	/* adjust iovec */
	204	iov[iovcnt].iov_base = &nexthop;
205	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
206
207	/* If action is RTM_DELETE we have to get rid of MPLS infos */
208	if (action != RTM_DELETE) {
209	memset(&sa_label_out, 0, sizeof(sa_label_out));
210	sa_label_out.smpls_len = sizeof(sa_label_out);
211	sa_label_out.smpls_family = AF_MPLS;
212	sa_label_out.smpls_label =
213	htonl(nhlfe->nexthop->nh_label->label[0]
214	<< MPLS_LABEL_OFFSET);
215	/* adjust header */
216	hdr.rtm_addrs \|= RTA_SRC;
217	hdr.rtm_flags \|= RTF_MPLS;
218	hdr.rtm_msglen += sizeof(sa_label_out);
219	/* adjust iovec */
220	iov[iovcnt].iov_base = &sa_label_out;
221	iov[iovcnt++].iov_len = sizeof(sa_label_out);
222
223	if (nhlfe->nexthop->nh_label->label[0] == MPLS_LABEL_IMPLNULL)
224	hdr.rtm_mpls = MPLS_OP_POP;
225	else
226	hdr.rtm_mpls = MPLS_OP_SWAP;
227	}
228
229	if (zserv_privs.change(ZPRIVS_RAISE))
af4c2728	230	flog_err(LIB_ERR_PRIVILEGES, "Can't raise privileges");
d62a17ae	231	ret = writev(kr_state.fd, iov, iovcnt);
d62a17ae	232	if (zserv_privs.change(ZPRIVS_LOWER))
af4c2728	233	flog_err(LIB_ERR_PRIVILEGES, "Can't lower privileges");
d62a17ae	234
d62a17ae	235	if (ret == -1)
09c866e3 QY	236	flog_err_sys(LIB_ERR_SOCKET, "%s: %s", __func__,
09c866e3 QY	237	safe_strerror(errno));
d62a17ae	238
d62a17ae	239	return ret;
d3e2c74a RW	240	}
d3e2c74a RW	241
d62a17ae	242	static int kernel_lsp_cmd(int action, zebra_lsp_t *lsp)
d3e2c74a	243	{
d62a17ae	244	zebra_nhlfe_t *nhlfe;
	245	struct nexthop *nexthop = NULL;
	246	unsigned int nexthop_num = 0;
	247
	248	for (nhlfe = lsp->nhlfe_list; nhlfe; nhlfe = nhlfe->next) {
	249	nexthop = nhlfe->nexthop;
	250	if (!nexthop)
	251	continue;
	252
	253	if (nexthop_num >= multipath_num)
	254	break;
	255
	256	if (((action == RTM_ADD \|\| action == RTM_CHANGE)
	257	&& (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
	258	&& CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE)))
	259	\|\| (action == RTM_DELETE
	260	&& (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_INSTALLED)
	261	&& CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB)))) {
5e8c8947 RW	262	if (nhlfe->nexthop->nh_label->num_labels > 1) {
	263	zlog_warn(
	264	"%s: can't push %u labels at once "
	265	"(maximum is 1)",
	266	__func__,
	267	nhlfe->nexthop->nh_label->num_labels);
	268	continue;
	269	}
	270
d62a17ae	271	nexthop_num++;
	272
	273	switch (NHLFE_FAMILY(nhlfe)) {
	274	case AF_INET:
	275	kernel_send_rtmsg_v4(action, lsp->ile.in_label,
	276	nhlfe);
	277	break;
	278	case AF_INET6:
	279	kernel_send_rtmsg_v6(action, lsp->ile.in_label,
	280	nhlfe);
	281	break;
	282	default:
	283	break;
	284	}
d62a17ae	285	}
	286	}
	287
	288	return (0);
d3e2c74a RW	289	}
d3e2c74a RW	290
7c5d0e18	291	enum dp_req_result kernel_add_lsp(zebra_lsp_t *lsp)
d3e2c74a	292	{
d62a17ae	293	int ret;
c4c8dec0	294
4a83e7a0	295	if (!lsp \|\| !lsp->best_nhlfe) { // unexpected
215181cb	296	kernel_lsp_pass_fail(lsp, DP_INSTALL_FAILURE);
7c5d0e18	297	return DP_REQUEST_FAILURE;
4a83e7a0	298	}
d3e2c74a	299
d62a17ae	300	ret = kernel_lsp_cmd(RTM_ADD, lsp);
c4c8dec0	301
4a83e7a0	302	kernel_lsp_pass_fail(lsp,
215181cb DS	303	(!ret) ? DP_INSTALL_SUCCESS
215181cb DS	304	: DP_INSTALL_FAILURE);
7c5d0e18 DS	305
7c5d0e18 DS	306	return DP_REQUEST_SUCCESS;
d3e2c74a RW	307	}
d3e2c74a RW	308
7c5d0e18	309	enum dp_req_result kernel_upd_lsp(zebra_lsp_t *lsp)
d3e2c74a	310	{
d62a17ae	311	int ret;
c4c8dec0	312
4a83e7a0	313	if (!lsp \|\| !lsp->best_nhlfe) { // unexpected
215181cb	314	kernel_lsp_pass_fail(lsp, DP_INSTALL_FAILURE);
7c5d0e18	315	return DP_REQUEST_FAILURE;
4a83e7a0	316	}
d3e2c74a	317
d62a17ae	318	ret = kernel_lsp_cmd(RTM_CHANGE, lsp);
c4c8dec0	319
4a83e7a0	320	kernel_lsp_pass_fail(lsp,
215181cb DS	321	(!ret) ? DP_INSTALL_SUCCESS
215181cb DS	322	: DP_INSTALL_FAILURE);
7c5d0e18	323	return DP_REQUEST_SUCCESS;
d3e2c74a RW	324	}
d3e2c74a RW	325
7c5d0e18	326	enum dp_req_result kernel_del_lsp(zebra_lsp_t *lsp)
d3e2c74a	327	{
d62a17ae	328	int ret;
c4c8dec0	329
4a83e7a0	330	if (!lsp) { // unexpected
215181cb	331	kernel_lsp_pass_fail(lsp, DP_DELETE_FAILURE);
7c5d0e18	332	return DP_REQUEST_FAILURE;
4a83e7a0	333	}
d3e2c74a	334
4a83e7a0	335	if (!CHECK_FLAG(lsp->flags, LSP_FLAG_INSTALLED)) {
215181cb	336	kernel_lsp_pass_fail(lsp, DP_DELETE_FAILURE);
7c5d0e18	337	return DP_REQUEST_FAILURE;
4a83e7a0	338	}
c4c8dec0	339
d62a17ae	340	ret = kernel_lsp_cmd(RTM_DELETE, lsp);
c4c8dec0	341
4a83e7a0	342	kernel_lsp_pass_fail(lsp,
215181cb DS	343	(!ret) ? DP_DELETE_SUCCESS
215181cb DS	344	: DP_DELETE_FAILURE);
7c5d0e18 DS	345
7c5d0e18 DS	346	return DP_REQUEST_SUCCESS;
d3e2c74a RW	347	}
d3e2c74a RW	348
d42127da RW	349	static int kmpw_install(struct zebra_pw *pw)
	350	{
	351	struct ifreq ifr;
	352	struct ifmpwreq imr;
	353	struct sockaddr_storage ss;
	354	struct sockaddr_in sa_in = (struct sockaddr_in )&ss;
	355	struct sockaddr_in6 sa_in6 = (struct sockaddr_in6 )&ss;
	356
	357	memset(&imr, 0, sizeof(imr));
	358	switch (pw->type) {
	359	case PW_TYPE_ETHERNET:
	360	imr.imr_type = IMR_TYPE_ETHERNET;
	361	break;
	362	case PW_TYPE_ETHERNET_TAGGED:
	363	imr.imr_type = IMR_TYPE_ETHERNET_TAGGED;
	364	break;
	365	default:
43e52561 QY	366	zlog_warn("%s: unhandled pseudowire type (%#X)", __func__,
43e52561 QY	367	pw->type);
d42127da RW	368	return -1;
	369	}
	370
	371	if (pw->flags & F_PSEUDOWIRE_CWORD)
	372	imr.imr_flags \|= IMR_FLAG_CONTROLWORD;
	373
	374	/* pseudowire nexthop */
	375	memset(&ss, 0, sizeof(ss));
	376	switch (pw->af) {
	377	case AF_INET:
	378	sa_in->sin_family = AF_INET;
	379	sa_in->sin_len = sizeof(struct sockaddr_in);
	380	sa_in->sin_addr = pw->nexthop.ipv4;
	381	break;
	382	case AF_INET6:
	383	sa_in6->sin6_family = AF_INET6;
	384	sa_in6->sin6_len = sizeof(struct sockaddr_in6);
	385	sa_in6->sin6_addr = pw->nexthop.ipv6;
	386	break;
	387	default:
43e52561 QY	388	zlog_warn("%s: unhandled pseudowire address-family (%u)",
43e52561 QY	389	__func__, pw->af);
d42127da RW	390	return -1;
	391	}
	392	memcpy(&imr.imr_nexthop, (struct sockaddr *)&ss,
	393	sizeof(imr.imr_nexthop));
	394
	395	/* pseudowire local/remote labels */
	396	imr.imr_lshim.shim_label = pw->local_label;
	397	imr.imr_rshim.shim_label = pw->remote_label;
	398
	399	/* ioctl */
	400	memset(&ifr, 0, sizeof(ifr));
	401	strlcpy(ifr.ifr_name, pw->ifname, sizeof(ifr.ifr_name));
	402	ifr.ifr_data = (caddr_t)&imr;
	403	if (ioctl(kr_state.ioctl_fd, SIOCSETMPWCFG, &ifr) == -1) {
09c866e3 QY	404	flog_err_sys(LIB_ERR_SYSTEM_CALL, "ioctl SIOCSETMPWCFG: %s",
09c866e3 QY	405	safe_strerror(errno));
d42127da RW	406	return -1;
	407	}
	408
	409	return 0;
	410	}
	411
	412	static int kmpw_uninstall(struct zebra_pw *pw)
	413	{
	414	struct ifreq ifr;
	415	struct ifmpwreq imr;
	416
	417	memset(&ifr, 0, sizeof(ifr));
	418	memset(&imr, 0, sizeof(imr));
	419	strlcpy(ifr.ifr_name, pw->ifname, sizeof(ifr.ifr_name));
	420	ifr.ifr_data = (caddr_t)&imr;
	421	if (ioctl(kr_state.ioctl_fd, SIOCSETMPWCFG, &ifr) == -1) {
09c866e3 QY	422	flog_err_sys(LIB_ERR_SYSTEM_CALL, "ioctl SIOCSETMPWCFG: %s",
09c866e3 QY	423	safe_strerror(errno));
d42127da RW	424	return -1;
	425	}
	426
	427	return 0;
	428	}
	429
d3e2c74a	430	#define MAX_RTSOCK_BUF 128 * 1024
d62a17ae	431	int mpls_kernel_init(void)
d3e2c74a	432	{
d62a17ae	433	int rcvbuf, default_rcvbuf;
	434	socklen_t optlen;
	435
	436	if ((kr_state.fd = socket(AF_ROUTE, SOCK_RAW, 0)) == -1) {
	437	zlog_warn("%s: socket", __func__);
	438	return -1;
	439	}
	440
d42127da RW	441	if ((kr_state.ioctl_fd = socket(AF_INET, SOCK_DGRAM \| SOCK_NONBLOCK, 0))
	442	== -1) {
	443	zlog_warn("%s: ioctl socket", __func__);
	444	return -1;
	445	}
	446
d62a17ae	447	/* grow receive buffer, don't wanna miss messages */
	448	optlen = sizeof(default_rcvbuf);
	449	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, &default_rcvbuf,
	450	&optlen)
	451	== -1)
	452	zlog_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
	453	else
	454	for (rcvbuf = MAX_RTSOCK_BUF;
	455	rcvbuf > default_rcvbuf
	456	&& setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf,
	457	sizeof(rcvbuf))
	458	== -1
	459	&& errno == ENOBUFS;
	460	rcvbuf /= 2)
	461	; /* nothing */
	462
	463	kr_state.rtseq = 1;
	464
d42127da RW	465	/* register hook to install/uninstall pseudowires */
	466	hook_register(pw_install, kmpw_install);
	467	hook_register(pw_uninstall, kmpw_uninstall);
	468
d62a17ae	469	return 0;
d3e2c74a	470	}
ddfeb486 DL	471
ddfeb486 DL	472	#endif /* OPEN_BSD */