strace/socketutils.c

/*
 * Copyright (c) 2014 Zubin Mithra <zubin.mithra@gmail.com>
 * Copyright (c) 2014-2016 Dmitry V. Levin <ldv@altlinux.org>
 * Copyright (c) 2014-2018 The strace developers.
 * 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. The name of the author may not be used to endorse or promote products
 *    derived from this software without specific prior written permission.
 *
 * 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.
 */

#include "defs.h"
#include <netinet/in.h>
#include <sys/socket.h>
#include <arpa/inet.h>
#include "netlink.h"
#include <linux/sock_diag.h>
#include <linux/inet_diag.h>
#include <linux/unix_diag.h>
#include <linux/netlink_diag.h>
#include <linux/packet_diag.h>
#include <linux/rtnetlink.h>
#include <linux/genetlink.h>

#include <sys/un.h>
#ifndef UNIX_PATH_MAX
# define UNIX_PATH_MAX sizeof(((struct sockaddr_un *) 0)->sun_path)
#endif

#include "xstring.h"

#define XLAT_MACROS_ONLY
# include "xlat/ethernet_protocols.h"
# include "xlat/inet_protocols.h"
# include "xlat/socktypes.h"
#undef XLAT_MACROS_ONLY

typedef struct {
	unsigned long inode;
	char *details;
	bool has_data;
} cache_entry;

#define CACHE_SIZE 1024U
static cache_entry cache[CACHE_SIZE];
#define CACHE_MASK (CACHE_SIZE - 1)

static int
cache_inode_details(const unsigned long inode, char *const details, bool data)
{
	cache_entry *e = &cache[inode & CACHE_MASK];
	free(e->details);
	e->inode = inode;
	e->details = details;
	e->has_data = data;

	return 1;
}

static const char *
get_sockaddr_by_inode_cached(const unsigned long inode)
{
	const cache_entry *const e = &cache[inode & CACHE_MASK];
	return (e && inode == e->inode) ? e->details : NULL;
}

static const char *
get_sockdata_by_inode_cached(const unsigned long inode)
{
	const cache_entry *const e = &cache[inode & CACHE_MASK];
	return (e && inode == e->inode && e->has_data)
		? e->details + strlen(e->details) + 1 : NULL;
}

static bool
print_sockaddr_by_inode_cached(const unsigned long inode)
{
	const char *const details = get_sockaddr_by_inode_cached(inode);
	if (details) {
		tprints(details);
		return true;
	}
	return false;
}

static bool
send_query(struct tcb *tcp, const int fd, void *req, size_t req_size)
{
	struct sockaddr_nl nladdr = {
		.nl_family = AF_NETLINK
	};
	struct iovec iov = {
		.iov_base = req,
		.iov_len = req_size
	};
	const struct msghdr msg = {
		.msg_name = &nladdr,
		.msg_namelen = sizeof(nladdr),
		.msg_iov = &iov,
		.msg_iovlen = 1
	};

	for (;;) {
		if (sendmsg(fd, &msg, 0) < 0) {
			if (errno == EINTR)
				continue;
			return false;
		}
		return true;
	}
}

static bool
inet_send_query(struct tcb *tcp, const int fd, const int family,
		const int proto)
{
	struct {
		const struct nlmsghdr nlh;
		const struct inet_diag_req_v2 idr;
	} req = {
		.nlh = {
			.nlmsg_len = sizeof(req),
			.nlmsg_type = SOCK_DIAG_BY_FAMILY,
			.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST
		},
		.idr = {
			.sdiag_family = family,
			.sdiag_protocol = proto,
			.idiag_states = -1
		}
	};
	return send_query(tcp, fd, &req, sizeof(req));
}

static int
inet_parse_response(const void *const data, const int data_len,
		    const unsigned long inode, void *opaque_data)
{
	const char *const proto_name = opaque_data;
	const struct inet_diag_msg *const diag_msg = data;
	static const char zero_addr[sizeof(struct in6_addr)];
	socklen_t addr_size, text_size;

	if (data_len < (int) NLMSG_LENGTH(sizeof(*diag_msg)))
		return -1;
	if (diag_msg->idiag_inode != inode)
		return 0;

	switch (diag_msg->idiag_family) {
		case AF_INET:
			addr_size = sizeof(struct in_addr);
			text_size = INET_ADDRSTRLEN;
			break;
		case AF_INET6:
			addr_size = sizeof(struct in6_addr);
			text_size = INET6_ADDRSTRLEN;
			break;
		default:
			return -1;
	}

	char src_buf[text_size];
	char *details;

	/* open/closing brackets for IPv6 addresses */
	const char *ob = diag_msg->idiag_family == AF_INET6 ? "[" : "";
	const char *cb = diag_msg->idiag_family == AF_INET6 ? "]" : "";

	if (!inet_ntop(diag_msg->idiag_family, diag_msg->id.idiag_src,
		       src_buf, text_size))
		return -1;

	if (diag_msg->id.idiag_dport ||
	    memcmp(zero_addr, diag_msg->id.idiag_dst, addr_size)) {
		char dst_buf[text_size];

		if (!inet_ntop(diag_msg->idiag_family, diag_msg->id.idiag_dst,
			       dst_buf, text_size))
			return -1;

		if (asprintf(&details, "%s:[%s%s%s:%u->%s%s%s:%u]", proto_name,
			     ob, src_buf, cb, ntohs(diag_msg->id.idiag_sport),
			     ob, dst_buf, cb, ntohs(diag_msg->id.idiag_dport))
		    < 0)
			return false;
	} else {
		if (asprintf(&details, "%s:[%s%s%s:%u]",
			     proto_name, ob, src_buf, cb,
			     ntohs(diag_msg->id.idiag_sport)) < 0)
			return false;
	}

	return cache_inode_details(inode, details, false);
}

static bool
receive_responses(struct tcb *tcp, const int fd, const unsigned long inode,
		  const unsigned long expected_msg_type,
		  int (*parser)(const void *, int,
				unsigned long, void *),
		  void *opaque_data)
{
	static union {
		struct nlmsghdr hdr;
		long buf[8192 / sizeof(long)];
	} hdr_buf;

	struct sockaddr_nl nladdr = {
		.nl_family = AF_NETLINK
	};
	struct iovec iov = {
		.iov_base = hdr_buf.buf,
		.iov_len = sizeof(hdr_buf.buf)
	};
	int flags = 0;

	for (;;) {
		struct msghdr msg = {
			.msg_name = &nladdr,
			.msg_namelen = sizeof(nladdr),
			.msg_iov = &iov,
			.msg_iovlen = 1
		};

		ssize_t ret = recvmsg(fd, &msg, flags);
		if (ret < 0) {
			if (errno == EINTR)
				continue;
			return false;
		}

		const struct nlmsghdr *h = &hdr_buf.hdr;
		if (!NLMSG_OK(h, ret))
			return false;
		for (; NLMSG_OK(h, ret); h = NLMSG_NEXT(h, ret)) {
			if (h->nlmsg_type != expected_msg_type)
				return false;
			const int rc = parser(NLMSG_DATA(h),
					      h->nlmsg_len, inode, opaque_data);
			if (rc > 0)
				return true;
			if (rc < 0)
				return false;
		}
		flags = MSG_DONTWAIT;
	}
}

static bool
unix_send_query(struct tcb *tcp, const int fd, const unsigned long inode)
{
	/*
	 * The kernel bug was fixed in mainline by commit v4.5-rc6~35^2~11
	 * and backported to stable/linux-4.4.y by commit v4.4.4~297.
	 */
	const uint16_t dump_flag =
		os_release < KERNEL_VERSION(4, 4, 4) ? NLM_F_DUMP : 0;

	struct {
		const struct nlmsghdr nlh;
		const struct unix_diag_req udr;
	} req = {
		.nlh = {
			.nlmsg_len = sizeof(req),
			.nlmsg_type = SOCK_DIAG_BY_FAMILY,
			.nlmsg_flags = NLM_F_REQUEST | dump_flag
		},
		.udr = {
			.sdiag_family = AF_UNIX,
			.udiag_ino = inode,
			.udiag_states = -1,
			.udiag_show = UDIAG_SHOW_NAME | UDIAG_SHOW_PEER,
			.udiag_cookie = { ~0U, ~0U }
		}
	};
	return send_query(tcp, fd, &req, sizeof(req));
}

static int
unix_parse_response(const void *data, const int data_len,
		    const unsigned long inode, void *opaque_data)
{
	const char *proto_name = opaque_data;
	const struct unix_diag_msg *diag_msg = data;
	struct rtattr *attr;
	int rta_len = data_len - NLMSG_LENGTH(sizeof(*diag_msg));
	uint32_t peer = 0;
	size_t path_len = 0;
	char path[UNIX_PATH_MAX + 1];

	if (rta_len < 0)
		return -1;
	if (diag_msg->udiag_ino != inode)
		return 0;
	if (diag_msg->udiag_family != AF_UNIX)
		return -1;

	for (attr = (struct rtattr *) (diag_msg + 1);
	     RTA_OK(attr, rta_len);
	     attr = RTA_NEXT(attr, rta_len)) {
		switch (attr->rta_type) {
		case UNIX_DIAG_NAME:
			if (!path_len) {
				path_len = RTA_PAYLOAD(attr);
				if (path_len > UNIX_PATH_MAX)
					path_len = UNIX_PATH_MAX;
				memcpy(path, RTA_DATA(attr), path_len);
				path[path_len] = '\0';
			}
			break;
		case UNIX_DIAG_PEER:
			if (RTA_PAYLOAD(attr) >= 4)
				peer = *(uint32_t *) RTA_DATA(attr);
			break;
		}
	}

	/*
	 * print obtained information in the following format:
	 * "UNIX:[" SELF_INODE [ "->" PEER_INODE ][ "," SOCKET_FILE ] "]"
	 */
	if (!peer && !path_len)
		return -1;

	char peer_str[3 + sizeof(peer) * 3];
	if (peer)
		xsprintf(peer_str, "->%u", peer);
	else
		peer_str[0] = '\0';

	const char *path_str;
	if (path_len) {
		char *outstr = alloca(4 * path_len + 4);

		outstr[0] = ',';
		if (path[0] == '\0') {
			outstr[1] = '@';
			string_quote(path + 1, outstr + 2,
				     path_len - 1, QUOTE_0_TERMINATED, NULL);
		} else {
			string_quote(path, outstr + 1,
				     path_len, QUOTE_0_TERMINATED, NULL);
		}
		path_str = outstr;
	} else {
		path_str = "";
	}

	char *details;
	if (asprintf(&details, "%s:[%lu%s%s]", proto_name, inode,
		     peer_str, path_str) < 0)
		return -1;

	return cache_inode_details(inode, details, false);
}

static bool
netlink_send_query(struct tcb *tcp, const int fd, const unsigned long inode)
{
	struct {
		const struct nlmsghdr nlh;
		const struct netlink_diag_req ndr;
	} req = {
		.nlh = {
			.nlmsg_len = sizeof(req),
			.nlmsg_type = SOCK_DIAG_BY_FAMILY,
			.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST
		},
		.ndr = {
			.sdiag_family = AF_NETLINK,
			.sdiag_protocol = NDIAG_PROTO_ALL
		}
	};
	return send_query(tcp, fd, &req, sizeof(req));
}

static bool
packet_send_query(struct tcb *tcp, const int fd, const unsigned long inode)
{
	struct {
		const struct nlmsghdr nlh;
		const struct packet_diag_req ndr;
	} req = {
		.nlh = {
			.nlmsg_len = sizeof(req),
			.nlmsg_type = SOCK_DIAG_BY_FAMILY,
			.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST
		},
		.ndr = {
			.sdiag_family = AF_PACKET,
			.pdiag_show = PACKET_SHOW_INFO,
		}
	};
	return send_query(tcp, fd, &req, sizeof(req));
}

static int
netlink_parse_response(const void *data, const int data_len,
		       const unsigned long inode, void *opaque_data)
{
	const char *proto_name = opaque_data;
	const struct netlink_diag_msg *const diag_msg = data;
	const char *netlink_proto;
	char *details;

	if (data_len < (int) NLMSG_LENGTH(sizeof(*diag_msg)))
		return -1;
	if (diag_msg->ndiag_ino != inode)
		return 0;

	if (diag_msg->ndiag_family != AF_NETLINK)
		return -1;

	netlink_proto = xlookup(netlink_protocols,
				diag_msg->ndiag_protocol);

	if (netlink_proto) {
		netlink_proto = STR_STRIP_PREFIX(netlink_proto, "NETLINK_");
		if (asprintf(&details, "%s:[%s:%u]%c%c", proto_name,
			     netlink_proto, diag_msg->ndiag_portid,
			     '\0', diag_msg->ndiag_protocol) < 0)
			return -1;
	} else {
		if (asprintf(&details, "%s:[%u]%c%c", proto_name,
			     (unsigned) diag_msg->ndiag_protocol,
			     '\0', diag_msg->ndiag_protocol) < 0)
			return -1;
	}

	return cache_inode_details(inode, details, true);
}

struct packet_cb_data {
	const char *proto_name;
	const char *ret;
};

static int
packet_parse_response(const void *data, const int data_len,
		       const unsigned long inode, void *opaque_data)
{
	struct packet_cb_data *cb_data = opaque_data;
	const struct packet_diag_msg *const diag_msg = data;

	uint32_t have_ifindex = false;
	uint32_t ifindex = 0;
	const char *ifname = NULL;

	bool have_uid = false;
	uint32_t uid = 0;

	struct rtattr *attr;
	int rta_len = data_len - NLMSG_LENGTH(sizeof(*diag_msg));

	const char *packet_proto;
	const char *packet_type;

	char *details;

	if (rta_len < 0)
		return -1;
	if (diag_msg->pdiag_ino != inode)
		return 0;

	if (diag_msg->pdiag_family != AF_PACKET)
		return -1;

	for (attr = (struct rtattr *) (diag_msg + 1);
	     RTA_OK(attr, rta_len);
	     attr = RTA_NEXT(attr, rta_len)) {
		switch (attr->rta_type) {
		case PACKET_DIAG_INFO: {
			if (have_ifindex)
				break;

			struct packet_diag_info *pinfo;

			if (RTA_PAYLOAD(attr) >= sizeof(pinfo)) {
				pinfo = RTA_DATA(attr);

				ifindex = pinfo->pdi_index;
				ifname = get_ifname(ifindex);
				have_ifindex = true;
			}
			break;
		}
		case PACKET_DIAG_UID: {
			if (have_uid)
				break;

			if (RTA_PAYLOAD(attr) >= sizeof(uid)) {
				uid = *(uint32_t *) RTA_DATA(attr);
				have_ifindex = true;
			}
		}
		}
	}

	packet_proto = xlookup(ethernet_protocols, diag_msg->pdiag_num);
	packet_type = xlookup(socktypes, diag_msg->pdiag_type);

	char ifindex_num[sizeof(ifindex) * 3 + sizeof("dev ")];
	char proto_num[sizeof(diag_msg->pdiag_num) * 2 + sizeof("proto 0x")];
	char type_num[sizeof(diag_msg->pdiag_type) * 2 + sizeof("type 0x")];

	//if (have_ifindex && !ifindex)
	//	return -1;

	if (!have_ifindex) {
		ifname = "???";
	} else if (have_ifindex && !ifname) {
		xsprintf(ifindex_num, "dev %u", ifindex);
		ifname = ifindex_num;
	}

	if (!packet_proto) {
		xsprintf(proto_num, "proto %#x", diag_msg->pdiag_num);
		packet_proto = proto_num;
	} else {
		packet_proto = STR_STRIP_PREFIX(packet_proto, "ETH_P_");
	}

	if (!packet_type) {
		xsprintf(type_num, "type %#x", diag_msg->pdiag_type);
		packet_type = type_num;
	} else {
		packet_type = STR_STRIP_PREFIX(packet_type, "SOCK_");
	}

	if (have_uid) {
		if (asprintf(&details, "%s:[%s:%s:%s:uid %u]",
		    cb_data->proto_name, ifname, packet_type,
		    packet_proto, uid) < 0)
			return -1;
	} else {
		if (asprintf(&details, "%s:[%s:%s:%s]",
		    cb_data->proto_name, ifname, packet_type, packet_proto) < 0)
			return -1;
	}

	cb_data->ret = details;

	if (have_ifindex && ifindex)
		return cache_inode_details(inode, details, false);

	return 1;
}

static const char *
unix_get(struct tcb *tcp, const int fd, const int family, const int proto,
	 const unsigned long inode, const char *name, bool data)
{
	if (data)
		return NULL;

	return unix_send_query(tcp, fd, inode)
		&& receive_responses(tcp, fd, inode, SOCK_DIAG_BY_FAMILY,
				     unix_parse_response, (void *) name)
		? get_sockaddr_by_inode_cached(inode) : NULL;
}

static const char *
inet_get(struct tcb *tcp, const int fd, const int family, const int protocol,
	 const unsigned long inode, const char *proto_name, bool data)
{
	if (data)
		return NULL;

	return inet_send_query(tcp, fd, family, protocol)
		&& receive_responses(tcp, fd, inode, SOCK_DIAG_BY_FAMILY,
				     inet_parse_response, (void *) proto_name)
		? get_sockaddr_by_inode_cached(inode) : NULL;
}

/* An additional fall-back cache for the cases sock_diag fails us, filled by
 * socket() calls */
enum {
	NLF_CACHE_BITS = 8,
	NLF_CACHE_SIZE = 1 << NLF_CACHE_BITS,
	NLF_CACHE_MASK = NLF_CACHE_SIZE - 1,
};
static const uint64_t NLF_CACHE_KEY_MASK = ~((uint64_t) NLF_CACHE_MASK);
static uint64_t netlink_cache[256];

void
set_netlink_family_cache_entry(uint64_t inode, uint8_t family)
{
	uint8_t idx = inode & NLF_CACHE_MASK;
	uint64_t val = (inode & NLF_CACHE_KEY_MASK) | family;

	netlink_cache[idx] = val;
}

int
get_netlink_family_cache_entry(uint64_t inode)
{
	uint8_t idx = inode & NLF_CACHE_MASK;
	uint64_t val = netlink_cache[idx];

	if (!((inode ^ val) & NLF_CACHE_KEY_MASK))
		return val & NLF_CACHE_MASK;

	return -1;
}

void
invalidate_netlink_family_cache_entry(uint64_t inode)
{
	uint8_t idx = inode & NLF_CACHE_MASK;
	uint64_t val = netlink_cache[idx];

	if ((inode ^ val) & NLF_CACHE_KEY_MASK)
		return;

	netlink_cache[idx] = ~0LLU;
}

static const char *
netlink_get(struct tcb *tcp, const int fd, const int family, const int protocol,
	    const unsigned long inode, const char *proto_name, bool data)
{
	const char *ret = NULL;

	if (netlink_send_query(tcp, fd, inode)
	    && receive_responses(tcp, fd, inode, SOCK_DIAG_BY_FAMILY,
				 netlink_parse_response, (void *) proto_name)) {
		ret = (data ? get_sockdata_by_inode_cached
			    : get_sockaddr_by_inode_cached)(inode);

		if (ret)
			return ret;
	}

	if (!ret && data) {
		static char buf;

		int ret = get_netlink_family_cache_entry(inode);

		if (ret < 0)
			return NULL;

		buf = ret;

		return &buf;
	}

	return NULL;
}

static const char *
packet_get(struct tcb *tcp, const int fd, const int family, const int protocol,
	   const unsigned long inode, const char *proto_name, bool data)
{
	struct packet_cb_data cb_data = { proto_name };

	if (data)
		return NULL;

	return packet_send_query(tcp, fd, inode)
		&& receive_responses(tcp, fd, inode, SOCK_DIAG_BY_FAMILY,
				     packet_parse_response,
				     (void *) &cb_data)
		? cb_data.ret : NULL;
}

static const struct {
	const char *const name;
	const char * (*const get)(struct tcb *, int fd, int family,
				  int protocol, unsigned long inode,
				  const char *proto_name, bool data);
	int family;
	int proto;
	bool has_data;
} protocols[] = {
	[SOCK_PROTO_UNIX]	= { "UNIX",	unix_get,	AF_UNIX},
	/*
	 * inet_diag handlers are currently implemented only for TCP,
	 * UDP(lite), SCTP, RAW, and DCCP, but we try to resolve it for all
	 * protocols anyway, just in case.
	 */
	[SOCK_PROTO_TCP]	=
		{ "TCP",	inet_get, AF_INET,  IPPROTO_TCP },
	[SOCK_PROTO_UDP]	=
		{ "UDP",	inet_get, AF_INET,  IPPROTO_UDP },
	[SOCK_PROTO_UDPLITE]	=
		{ "UDPLITE",	inet_get, AF_INET,  IPPROTO_UDPLITE },
	[SOCK_PROTO_DCCP]	=
		{ "DCCP",	inet_get, AF_INET,  IPPROTO_DCCP },
	[SOCK_PROTO_SCTP]	=
		{ "SCTP",	inet_get, AF_INET,  IPPROTO_SCTP },
	[SOCK_PROTO_L2TP_IP]	=
		{ "L2TP/IP",	inet_get, AF_INET,  IPPROTO_L2TP },
	[SOCK_PROTO_PING]	=
		{ "PING",	inet_get, AF_INET,  IPPROTO_ICMP },
	[SOCK_PROTO_RAW]	=
		{ "RAW",	inet_get, AF_INET,  IPPROTO_RAW },
	[SOCK_PROTO_TCPv6]	=
		{ "TCPv6",	inet_get, AF_INET6, IPPROTO_TCP },
	[SOCK_PROTO_UDPv6]	=
		{ "UDPv6",	inet_get, AF_INET6, IPPROTO_UDP },
	[SOCK_PROTO_UDPLITEv6]	=
		{ "UDPLITEv6",	inet_get, AF_INET6, IPPROTO_UDPLITE },
	[SOCK_PROTO_DCCPv6]	=
		{ "DCCPv6",	inet_get, AF_INET6, IPPROTO_DCCP },
	[SOCK_PROTO_SCTPv6]	=
		{ "SCTPv6",	inet_get, AF_INET6, IPPROTO_SCTP },
	[SOCK_PROTO_L2TP_IPv6]	=
		{ "L2TP/IPv6",	inet_get, AF_INET6, IPPROTO_L2TP },
	[SOCK_PROTO_PINGv6]	=
		{ "PINGv6",	inet_get, AF_INET6, IPPROTO_ICMP },
	[SOCK_PROTO_RAWv6]	=
		{ "RAWv6",	inet_get, AF_INET6, IPPROTO_RAW },
	[SOCK_PROTO_NETLINK]	=
		{ "NETLINK",	netlink_get, AF_NETLINK, 0, true },
	[SOCK_PROTO_AX25]	= { "AX25",	NULL,		AF_AX25 },
	[SOCK_PROTO_DDP]	= { "DDP",	NULL,		AF_APPLETALK },
	[SOCK_PROTO_NETROM]	= { "NETROM",	NULL,		AF_NETROM },
	[SOCK_PROTO_PACKET]	= { "PACKET",	packet_get,	AF_PACKET },
	[SOCK_PROTO_ROSE]	= { "ROSE",	NULL,		AF_ROSE },
	[SOCK_PROTO_X25]	= { "X25",	NULL,		AF_X25 },
};

enum sock_proto
get_proto_by_name(const char *const name)
{
	unsigned int i;
	for (i = (unsigned int) SOCK_PROTO_UNKNOWN + 1;
	     i < ARRAY_SIZE(protocols); ++i) {
		if (protocols[i].name && !strcmp(name, protocols[i].name))
			return (enum sock_proto) i;
	}
	return SOCK_PROTO_UNKNOWN;
}

int
get_family_by_proto(enum sock_proto proto)
{
	if ((size_t) proto < ARRAY_SIZE(protocols))
		return protocols[proto].family;

	return AF_UNSPEC;
}

static const char *
get_sockaddr_by_inode_uncached(struct tcb *tcp, const unsigned long inode,
			       const enum sock_proto proto, bool data)
{
	if ((unsigned int) proto >= ARRAY_SIZE(protocols) ||
	    (proto != SOCK_PROTO_UNKNOWN && !protocols[proto].get))
		return NULL;

	const int fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_SOCK_DIAG);
	if (fd < 0)
		return NULL;
	const char *details = NULL;

	if (proto != SOCK_PROTO_UNKNOWN) {
		details = protocols[proto].get(tcp, fd, protocols[proto].family,
					       protocols[proto].proto, inode,
					       protocols[proto].name, data);
	} else {
		unsigned int i;
		for (i = (unsigned int) SOCK_PROTO_UNKNOWN + 1;
		     i < ARRAY_SIZE(protocols); ++i) {
			if (!protocols[i].get)
				continue;
			details = protocols[i].get(tcp, fd,
						   protocols[proto].family,
						   protocols[proto].proto,
						   inode,
						   protocols[proto].name, data);
			if (details)
				break;
		}
	}

	close(fd);
	return details;
}

static bool
print_sockaddr_by_inode_uncached(struct tcb *tcp, const unsigned long inode,
				 const enum sock_proto proto)
{
	const char *details = get_sockaddr_by_inode_uncached(tcp, inode, proto,
							     false);

	if (details) {
		tprints(details);
		return true;
	}

	if ((unsigned int) proto < ARRAY_SIZE(protocols) &&
	    protocols[proto].name) {
		tprintf("%s:[%lu]", protocols[proto].name, inode);
		return true;
	}

	return false;
}

/* Given an inode number of a socket, return its protocol details.  */
const char *
get_sockaddr_by_inode(struct tcb *const tcp, const int fd,
		      const unsigned long inode)
{
	const char *details = get_sockaddr_by_inode_cached(inode);
	return details ? details :
		get_sockaddr_by_inode_uncached(tcp, inode, getfdproto(tcp, fd),
					       false);
}

const char *
get_sockdata_by_inode(struct tcb *const tcp, const int fd,
		      const unsigned long inode)
{
	const char *data = get_sockdata_by_inode_cached(inode);

	return data ? data :
		get_sockaddr_by_inode_uncached(tcp, inode, getfdproto(tcp, fd),
					       true);
}

/* Given an inode number of a socket, print out its protocol details.  */
bool
print_sockaddr_by_inode(struct tcb *const tcp, const int fd,
			const unsigned long inode)
{
	return print_sockaddr_by_inode_cached(inode) ? true :
		print_sockaddr_by_inode_uncached(tcp, inode,
						 getfdproto(tcp, fd));
}

/*
 * Managing the cache for decoding communications of Netlink GENERIC protocol
 *
 * As name shown Netlink GENERIC protocol is generic protocol. The
 * numbers of msg types used in the protocol are not defined
 * statically. Kernel defines them on demand.  So the xlat converted
 * from header files doesn't help for decoding the protocol. Following
 * codes are building xlat(dyxlat) at runtime.
 */
static bool
genl_query_families(struct tcb *tcp, const int fd, int id)
{
	struct genl_req {
		const struct nlmsghdr nlh;
		struct genlmsghdr gnlh;
		struct nlattr ATTRIBUTE_ALIGNED(NLA_ALIGNTO) nlah;
		uint16_t id;
	} req = {
		.nlh = {
			.nlmsg_len = sizeof(req),
			.nlmsg_type = GENL_ID_CTRL,
			.nlmsg_flags = NLM_F_REQUEST
				       | (id < 0 ? NLM_F_DUMP : 0),
		},
		.gnlh = {
			.cmd = CTRL_CMD_GETFAMILY,
		},
		.nlah = {
			.nla_len = offsetofend(struct genl_req, id)
				    - offsetof(struct genl_req, nlah),
			.nla_type = CTRL_ATTR_FAMILY_ID,
		},
		.id = id,
	};
	return send_query(tcp, fd, &req, sizeof(req));
}

static int
genl_parse_families_response(const void *const data,
			     const int data_len, const unsigned long inode,
			     void *opaque_data)
{
	struct dyxlat *const dyxlat = opaque_data;
	const struct genlmsghdr *const gnlh = data;
	struct rtattr *attr;
	int rta_len = data_len - NLMSG_LENGTH(sizeof(*gnlh));

	char *name = NULL;
	unsigned int name_len = 0;
	uint16_t *id = NULL;

	if (rta_len < 0)
		return -1;
	if (gnlh->cmd != CTRL_CMD_NEWFAMILY)
		return -1;
	if (gnlh->version != 2)
		return -1;

	for (attr = (struct rtattr *) (gnlh + 1);
	     RTA_OK(attr, rta_len);
	     attr = RTA_NEXT(attr, rta_len)) {
		switch (attr->rta_type) {
		case CTRL_ATTR_FAMILY_NAME:
			if (!name) {
				name = RTA_DATA(attr);
				name_len = RTA_PAYLOAD(attr);
			}
			break;
		case CTRL_ATTR_FAMILY_ID:
			if (!id && RTA_PAYLOAD(attr) == sizeof(*id))
				id = RTA_DATA(attr);
			break;
		}

		if (name && id) {
			dyxlat_add_pair(dyxlat, *id, name, name_len);
			name = NULL;
			id = NULL;
		}
	}

	return 0;
}

const char *
genl_get_family_name(struct tcb *tcp, uint16_t id)
{
	static struct dyxlat *dyxlat;

	const char *name = dyxlat ? xlookup(dyxlat_get(dyxlat), id) : NULL;
	if (name)
		return name;

	bool init = !dyxlat;
	if (init)
		dyxlat = dyxlat_alloc(0);

	int fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_GENERIC);
	if (fd < 0)
		return NULL;

	if (genl_query_families(tcp, fd, init ? -1 : id))
		receive_responses(tcp, fd, 0, GENL_ID_CTRL,
				  genl_parse_families_response, dyxlat);
	close(fd);

	return xlookup(dyxlat_get(dyxlat), id);
}