view src/privsep-bpf.c @ 5231:a2c342295221 draft

privsep: Enable Capsicum for all processes. Except for the priviledged process. This is quite an in-depth change: * ARP is now one process per address * BPF flags are now returned via privsep * BPF write filters are locked when supported * The root process sends to the network The last step is done by opening RAW sockets and then sending a UDP header (where applicable) to avoid binding to an address which is already in use by the reader sockets. This is slightly wasteful for OS's without sandboxing but does have the very nice side effect of not needing a source address to unicast DHCPs replies from which makes the code smaller.
author Roy Marples <roy@marples.name>
date Tue, 19 May 2020 16:19:05 +0100
parents bb468c1a3b46
children bcd021398c1d
line wrap: on
line source

/* SPDX-License-Identifier: BSD-2-Clause */
/*
 * Privilege Separation BPF Initiator
 * Copyright (c) 2006-2020 Roy Marples <roy@marples.name>
 * 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.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
 */

#include <sys/socket.h>
#include <sys/types.h>

/* Need these headers just for if_ether on some OS. */
#ifndef __NetBSD__
#include <net/if.h>
#include <net/if_arp.h>
#include <netinet/in.h>
#endif
#include <netinet/if_ether.h>

#include <assert.h>
#include <pwd.h>
#include <errno.h>
#include <signal.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

#include "arp.h"
#include "bpf.h"
#include "dhcp.h"
#include "dhcp6.h"
#include "eloop.h"
#include "ipv6nd.h"
#include "logerr.h"
#include "privsep.h"

#ifdef HAVE_CAPSICUM
#include <sys/capsicum.h>
#endif

static void
ps_bpf_recvbpf(void *arg)
{
	struct ps_process *psp = arg;
	struct bpf *bpf = psp->psp_bpf;
	uint8_t buf[sizeof(bpf->bpf_flags) + FRAMELEN_MAX];
	ssize_t len;
	struct ps_msghdr psm = {
		.ps_id = psp->psp_id,
		.ps_cmd = psp->psp_id.psi_cmd,
	};

	bpf->bpf_flags &= ~BPF_EOF;
	/* A BPF read can read more than one filtered packet at time.
	 * This mechanism allows us to read each packet from the buffer. */
	while (!(bpf->bpf_flags & BPF_EOF)) {
		len = bpf_read(bpf,
		    buf + sizeof(bpf->bpf_flags),
		    sizeof(buf) - sizeof(bpf->bpf_flags));
		if (len == -1)
			logerr(__func__);
		if (len == -1 || len == 0)
			break;
		memcpy(buf, &bpf->bpf_flags, sizeof(bpf->bpf_flags));
		len = ps_sendpsmdata(psp->psp_ctx, psp->psp_ctx->ps_data_fd,
		    &psm, buf, (size_t)len + sizeof(bpf->bpf_flags));
		if (len == -1 && errno != ECONNRESET)
			logerr(__func__);
		if (len == -1 || len == 0)
			break;
	}
}

static ssize_t
ps_bpf_recvmsgcb(void *arg, __unused struct ps_msghdr *psm, struct msghdr *msg)
{
	struct ps_process *psp = arg;
	struct iovec *iov = msg->msg_iov;

	return bpf_send(psp->psp_bpf, psp->psp_proto,
	    iov->iov_base, iov->iov_len);
}

static void
ps_bpf_recvmsg(void *arg)
{
	struct ps_process *psp = arg;

	if (ps_recvpsmsg(psp->psp_ctx, psp->psp_fd,
	    ps_bpf_recvmsgcb, arg) == -1)
		logerr(__func__);
}

static int
ps_bpf_start_bpf(void *arg)
{
	struct ps_process *psp = arg;
	struct dhcpcd_ctx *ctx = psp->psp_ctx;
	char *addr;
	struct in_addr *ia = &psp->psp_id.psi_addr.psa_in_addr;
#ifdef HAVE_CAPSICUM
	cap_rights_t rights;

	/* We need CAP_IOCTL so we can change the BPF filter when we
	 * need to. */
	cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_EVENT, CAP_IOCTL);
#endif

	if (ia->s_addr == INADDR_ANY) {
		ia = NULL;
		addr = NULL;
	} else
		addr = inet_ntoa(*ia);
	setproctitle("[BPF %s] %s%s%s", psp->psp_protostr, psp->psp_ifname,
	    addr != NULL ? " " : "", addr != NULL ? addr : "");
	ps_freeprocesses(ctx, psp);

	psp->psp_bpf = bpf_open(&psp->psp_ifp, psp->psp_filter, ia);
	if (psp->psp_bpf == NULL)
		logerr("%s: bpf_open",__func__);
#ifdef HAVE_CAPSICUM
	else if (cap_rights_limit(psp->psp_bpf->bpf_fd, &rights) == -1 &&
	    errno != ENOSYS)
		logerr("%s: cap_rights_limit", __func__);
#endif
	else if (eloop_event_add(ctx->eloop,
	    psp->psp_bpf->bpf_fd, ps_bpf_recvbpf, psp) == -1)
		logerr("%s: eloop_event_add", __func__);
	else {
		psp->psp_work_fd = psp->psp_bpf->bpf_fd;
		return 0;
	}

	eloop_exit(ctx->eloop, EXIT_FAILURE);
	return -1;
}

static void
ps_bpf_signal_bpfcb(int sig, void *arg)
{
	struct dhcpcd_ctx *ctx = arg;

	eloop_exit(ctx->eloop, sig == SIGTERM ? EXIT_SUCCESS : EXIT_FAILURE);
}

ssize_t
ps_bpf_cmd(struct dhcpcd_ctx *ctx, struct ps_msghdr *psm, struct msghdr *msg)
{
	uint16_t cmd;
	struct ps_process *psp;
	pid_t start;
	struct iovec *iov = msg->msg_iov;
	struct interface *ifp;
	struct ipv4_state *istate;

	cmd = (uint16_t)(psm->ps_cmd & ~(PS_START | PS_STOP));
	psp = ps_findprocess(ctx, &psm->ps_id);

#ifdef PRIVSEP_DEBUG
	logerrx("%s: IN cmd %x, psp %p", __func__, psm->ps_cmd, psp);
#endif

	switch (cmd) {
#ifdef ARP
	case PS_BPF_ARP:	/* FALLTHROUGH */
#endif
	case PS_BPF_BOOTP:
		break;
	default:
		logerrx("%s: unknown command %x", __func__, psm->ps_cmd);
		errno = ENOTSUP;
		return -1;
	}

	if (!(psm->ps_cmd & PS_START)) {
		errno = EINVAL;
		return -1;
	}

	if (psp != NULL)
		return 1;

	psp = ps_newprocess(ctx, &psm->ps_id);
	if (psp == NULL)
		return -1;

	ifp = &psp->psp_ifp;
	assert(msg->msg_iovlen == 1);
	assert(iov->iov_len == sizeof(*ifp));
	memcpy(ifp, iov->iov_base, sizeof(*ifp));
	ifp->ctx = psp->psp_ctx;
	ifp->options = NULL;
	memset(ifp->if_data, 0, sizeof(ifp->if_data));

	if ((istate = ipv4_getstate(ifp)) == NULL) {
		ps_freeprocess(psp);
		return -1;
	}

	memcpy(psp->psp_ifname, ifp->name, sizeof(psp->psp_ifname));

	switch (cmd) {
#ifdef ARP
	case PS_BPF_ARP:
		psp->psp_proto = ETHERTYPE_ARP;
		psp->psp_protostr = "ARP";
		psp->psp_filter = bpf_arp;
		break;
#endif
	case PS_BPF_BOOTP:
		psp->psp_proto = ETHERTYPE_IP;
		psp->psp_protostr = "BOOTP";
		psp->psp_filter = bpf_bootp;
		break;
	}

	start = ps_dostart(ctx,
	    &psp->psp_pid, &psp->psp_fd,
	    ps_bpf_recvmsg, NULL, psp,
	    ps_bpf_start_bpf, ps_bpf_signal_bpfcb,
	    PSF_DROPPRIVS);
	switch (start) {
	case -1:
		ps_freeprocess(psp);
		return -1;
	case 0:
#ifdef HAVE_CAPSICUM
		if (cap_enter() == -1 && errno != ENOSYS)
			logerr("%s: cap_enter", __func__);
#endif
#ifdef HAVE_PLEDGE
		if (pledge("stdio", NULL) == -1)
			logerr("%s: pledge", __func__);
#endif
		break;
	default:
#ifdef PRIVSEP_DEBUG
		logdebugx("%s: spawned BPF %s on PID %d",
		    psp->psp_ifname, psp->psp_protostr, start);
#endif
		break;
	}
	return start;
}

ssize_t
ps_bpf_dispatch(struct dhcpcd_ctx *ctx,
    struct ps_msghdr *psm, struct msghdr *msg)
{
	struct iovec *iov = msg->msg_iov;
	struct interface *ifp;
	uint8_t bpf_flags, *bpf;
	size_t bpf_len;

	ifp = if_findindex(ctx->ifaces, psm->ps_id.psi_ifindex);
	bpf = iov->iov_base;
	bpf_len = iov->iov_len;
	memcpy(&bpf_flags, bpf, sizeof(bpf_flags));
	bpf += sizeof(bpf_flags);
	bpf_len -= sizeof(bpf_flags);

	switch (psm->ps_cmd) {
#ifdef ARP
	case PS_BPF_ARP:
		arp_packet(ifp, bpf, bpf_len, bpf_flags);
		break;
#endif
	case PS_BPF_BOOTP:
		dhcp_packet(ifp, bpf, bpf_len, bpf_flags);
		break;
	default:
		errno = ENOTSUP;
		return -1;
	}

	return 1;
}

static ssize_t
ps_bpf_send(const struct interface *ifp, const struct in_addr *ia,
    uint16_t cmd, const void *data, size_t len)
{
	struct dhcpcd_ctx *ctx = ifp->ctx;
	struct ps_msghdr psm = {
		.ps_cmd = cmd,
		.ps_id = {
			.psi_ifindex = ifp->index,
			.psi_cmd = (uint8_t)(cmd & ~(PS_START | PS_STOP)),
		},
	};

	if (ia != NULL)
		psm.ps_id.psi_addr.psa_in_addr = *ia;

	return ps_sendpsmdata(ctx, ctx->ps_root_fd, &psm, data, len);
}

#ifdef ARP
ssize_t
ps_bpf_openarp(const struct interface *ifp, const struct in_addr *ia)
{

	assert(ia != NULL);
	return ps_bpf_send(ifp, ia, PS_BPF_ARP | PS_START,
	    ifp, sizeof(*ifp));
}

ssize_t
ps_bpf_closearp(const struct interface *ifp, const struct in_addr *ia)
{

	return ps_bpf_send(ifp, ia, PS_BPF_ARP | PS_STOP, NULL, 0);
}

ssize_t
ps_bpf_sendarp(const struct interface *ifp, const struct in_addr *ia,
    const void *data, size_t len)
{

	assert(ia != NULL);
	return ps_bpf_send(ifp, ia, PS_BPF_ARP, data, len);
}
#endif

ssize_t
ps_bpf_openbootp(const struct interface *ifp)
{

	return ps_bpf_send(ifp, NULL, PS_BPF_BOOTP | PS_START,
	    ifp, sizeof(*ifp));
}

ssize_t
ps_bpf_closebootp(const struct interface *ifp)
{

	return ps_bpf_send(ifp, NULL, PS_BPF_BOOTP | PS_STOP, NULL, 0);
}

ssize_t
ps_bpf_sendbootp(const struct interface *ifp, const void *data, size_t len)
{

	return ps_bpf_send(ifp, NULL, PS_BPF_BOOTP, data, len);
}