[paraslash.git] / net.c

/*
 * Copyright (C) 2005-2008 Andre Noll <maan@systemlinux.org>
 *
 * Licensed under the GPL v2. For licencing details see COPYING.
 */

/** \file net.c Networking-related helper functions. */

#include <netdb.h>

/* At least NetBSD needs these. */
#ifndef AI_V4MAPPED
#define AI_V4MAPPED 0
#endif
#ifndef AI_ALL
#define AI_ALL 0
#endif
#ifndef AI_ADDRCONFIG
#define AI_ADDRCONFIG 0
#endif


#include "para.h"
#include "error.h"
#include "net.h"
#include "string.h"


/** Information about one encrypted connection. */
struct crypt_data {
        /** Function used to decrypt received data. */
        crypt_function *recv;
        /** Function used to encrypt data to be sent. */
        crypt_function *send;
        /**
         * Context-dependent data (crypt keys), passed verbatim to the above
         * crypt functions.
         */
        void *private_data;
};
/** Array holding per fd crypt data. */
static struct crypt_data *crypt_data_array;
/** Current size of the crypt data array. */
static unsigned cda_size = 0;

/**
 * Activate encryption for one file descriptor.
 *
 * \param fd The file descriptor.
 * \param recv_f The function used for decrypting received data.
 * \param send_f The function used for encrypting before sending.
 * \param private_data User data supplied by the caller.
 */
void enable_crypt(int fd, crypt_function *recv_f, crypt_function *send_f,
        void *private_data)
{
        if (fd + 1 > cda_size) {
                crypt_data_array = para_realloc(crypt_data_array,
                        (fd + 1) * sizeof(struct crypt_data));
                memset(crypt_data_array + cda_size, 0,
                        (fd + 1 - cda_size) * sizeof(struct crypt_data));
                cda_size = fd + 1;
        }
        crypt_data_array[fd].recv = recv_f;
        crypt_data_array[fd].send = send_f;
        crypt_data_array[fd].private_data = private_data;
        PARA_INFO_LOG("rc4 encryption activated for fd %d\n", fd);
}

/**
 * Deactivate encryption for a given fd.
 *
 * \param fd The file descriptor.
 *
 * This must be called if and only if \p fd was activated via enable_crypt().
 */
void disable_crypt(int fd)
{
        if (cda_size < fd + 1)
                return;
        crypt_data_array[fd].recv = NULL;
        crypt_data_array[fd].send = NULL;
        crypt_data_array[fd].private_data = NULL;
}


/**
 * Determine the socket type for a given layer-4 protocol.
 *
 * \param l4type The symbolic name of the transport-layer protocol.
 *
 * \sa ip(7), socket(2)
 */
static inline int sock_type(const unsigned l4type)
{
        switch (l4type) {
        case IPPROTO_UDP:       return SOCK_DGRAM;
        case IPPROTO_TCP:       return SOCK_STREAM;
        case IPPROTO_DCCP:      return SOCK_DCCP;
        }
        return -1;              /* not supported here */
}

/**
 * Pretty-print transport-layer name.
 */
static const char *layer4_name(const unsigned l4type)
{
        switch (l4type) {
        case IPPROTO_UDP:       return "UDP";
        case IPPROTO_TCP:       return "TCP";
        case IPPROTO_DCCP:      return "DCCP";
        }
        return "UNKNOWN PROTOCOL";
}

/**
 * Resolve IPv4/IPv6 address and create a ready-to-use active or passive socket.
 *
 * \param l3type The layer-3 type (\p AF_INET, \p AF_INET6, \p AF_UNSPEC).
 * \param l4type The layer-4 type (\p IPPROTO_xxx).
 * \param passive Whether this is a passive (1) or active (0) socket.
 * \param host Remote or local hostname or IPv/6 address string.
 * \param port_number Decimal port number.
 *
 * This creates a ready-made IPv4/v6 socket structure after looking up the
 * necessary parameters. The interpretation of \a host depends on the value of
 * \a passive:
 *      - on a passive socket host is interpreted as an interface IPv4/6 address
 *        (can be left NULL);
 *      - on an active socket, \a host is the peer DNS name or IPv4/6 address
 *        to connect to;
 *      - \a port_number is in either case the numeric port number (not service
 *        string).
 *
 * Furthermore, bind(2) is called on passive sockets, and connect(2) on active
 * sockets. The algorithm tries all possible address combinations until it
 * succeeds.
 *
 * \return This function returns 1 on success and \a -E_ADDRESS_LOOKUP when no
 * matching connection could be set up (with details in the error log).
 *
 *  \sa ipv6(7), getaddrinfo(3), bind(2), connect(2).
 */
int makesock(unsigned l3type, unsigned l4type, int passive,
                const char *host, unsigned short port_number)
{
        struct addrinfo *local = NULL, *src,
                        *remote = NULL, *dst, hints;
        char            *port = make_message("%u", port_number);
        int             rc, on = 1, sockfd = -1,
                        socktype = sock_type(l4type);

        /* Set up address hint structure */
        memset(&hints, 0, sizeof(hints));
        hints.ai_family = l3type;
        /* getaddrinfo does not really work well with SOCK_DCCP */
        if (socktype == SOCK_DGRAM || socktype == SOCK_STREAM)
                hints.ai_socktype = socktype;

        /* only use addresses available on the host */
        hints.ai_flags = AI_ADDRCONFIG;
        if (l3type == AF_INET6)
                /* use v4-mapped-v6 if no v6 addresses found */
                hints.ai_flags |= AI_V4MAPPED | AI_ALL;

        if (passive && host == NULL)
                hints.ai_flags |= AI_PASSIVE;

        /* Obtain local/remote address information */
        if ((rc = getaddrinfo(host, port, &hints, passive ? &local : &remote))) {
                PARA_ERROR_LOG("can not resolve %s address %s#%s: %s.\n",
                                layer4_name(l4type),
                                host?  : (passive? "[loopback]" : "[localhost]"),
                                port, gai_strerror(rc));
                return -E_ADDRESS_LOOKUP;
        }

        /* Iterate over all src/dst combination, exhausting dst first */
        for (src = local, dst = remote; src != NULL || dst != NULL; /* no op */ ) {
                if (src && dst && src->ai_family == AF_INET
                                && dst->ai_family == AF_INET6)
                        goto get_next_dst; /* v4 -> v6 is not possible */

                sockfd = socket(src ? src->ai_family : dst->ai_family,
                        socktype, l4type);
                if (sockfd < 0)
                        goto get_next_dst;

                /*
                 * Set those options that need to be set before establishing
                 * the connection. Reuse the address on passive (listening)
                 * sockets to avoid failure on restart.
                 */
                if (passive && setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR,
                                &on, sizeof(on)) == -1) {
                        PARA_ERROR_LOG("can not set SO_REUSEADDR: %s\n",
                                strerror(errno));
                        return -ERRNO_TO_PARA_ERROR(errno);
                }

                if (src) {
                        if (bind(sockfd, src->ai_addr, src->ai_addrlen) < 0) {
                                close(sockfd);
                                goto get_next_src;
                        }
                        if (!dst) /* bind-only completed successfully */
                                break;
                }

                if (dst && connect(sockfd, dst->ai_addr, dst->ai_addrlen) == 0)
                        break; /* connection completed successfully */
                close(sockfd);
get_next_dst:
                if (dst && (dst = dst->ai_next))
                        continue;
get_next_src:
                if (src && (src = src->ai_next)) /* restart inner loop */
                        dst = remote;
        }
        if (local)
                freeaddrinfo(local);
        if (remote)
                freeaddrinfo(remote);

        if (src == NULL && dst == NULL) {
                PARA_ERROR_LOG("can not create %s socket %s#%s.\n",
                        layer4_name(l4type), host?  : (passive?
                        "[loopback]" : "[localhost]"), port);
                return -ERRNO_TO_PARA_ERROR(errno);
        }
        return sockfd;
}

/**
 * Create a passive / listening socket.
 *
 * \param l3type The network-layer type (\p AF_xxx).
 * \param l4type The transport-layer type (\p IPPROTO_xxx).
 * \param port The decimal port number to listen on.
 *
 * \return Positive integer (socket descriptor) on success, negative value
 * otherwise.
 *
 * \sa makesock(), ip(7), ipv6(7), bind(2), listen(2).
 */
int para_listen(unsigned l3type, unsigned l4type, unsigned short port)
{
        int ret, fd = makesock(l3type, l4type, 1, NULL, port);

        if (fd > 0) {
                ret = listen(fd, BACKLOG);
                if (ret < 0) {
                        close(fd);
                        return -ERRNO_TO_PARA_ERROR(errno);
                }
                PARA_INFO_LOG("listening on %s port %u, fd %d\n",
                        layer4_name(l4type), port, fd);
        }
        return fd;
}

/**
 * Print numeric host and port number (beware - uses static char).
 *
 * \param sa The IPv4/IPv6 socket address to use.
 * \param len The length of \p sa.
 *
 * \sa getnameinfo(3).
 */
static char *host_and_port(struct sockaddr *sa, socklen_t len)
{
        static char output[NI_MAXHOST + NI_MAXSERV + 2];
        char hbuf[NI_MAXHOST], sbuf[NI_MAXSERV];
        int ret;

        ret = getnameinfo(sa, len, hbuf, sizeof(hbuf), sbuf, sizeof(sbuf),
                NI_NUMERICHOST | NI_NUMERICSERV);
        if (ret) {
                PARA_WARNING_LOG("hostname lookup error (%s).\n",
                        gai_strerror(ret));
                sprintf(output, "(unknown)");
        } else
                sprintf(output, "%s#%s", hbuf, sbuf);
        return output;
}

/**
 * Look up the local or remote side of a connected socket structure.
 *
 * \param fd The socket descriptor of the connected socket.
 * \param getname Either \p getsockname() for local, or \p getpeername() for
 * remote side.
 *
 * \return A static character string identifying hostname and port of the
 * chosen side.
 *
 * \sa getsockname(2), getpeername(2).
 */
static char *__get_sock_name(int fd, int (*getname)(int, struct sockaddr*,
                socklen_t *))
{
        struct sockaddr_storage ss;
        socklen_t sslen = sizeof(ss);

        if (getname(fd, (struct sockaddr *)&ss, &sslen) < 0) {
                static char *dont_know = "(don't know)";
                PARA_ERROR_LOG("can not determine address from fd %d: %s\n",
                        fd, strerror(errno));
                return dont_know;
        }
        return host_and_port((struct sockaddr *)&ss, sslen);
}

char *local_name(int sockfd)
{
        return __get_sock_name(sockfd, getsockname);
}

char *remote_name(int sockfd)
{
        return __get_sock_name(sockfd, getpeername);
}

/**
 * Extract IPv4 or IPv6-mapped-IPv4 address from sockaddr_storage.
 * \param ss Container of IPv4/6 address
 * \return Extracted IPv4 address (different from 0) or 0 if unsuccessful.
 *
 * \sa RFC 3493
 */
struct in_addr extract_v4_addr(const struct sockaddr_storage *ss)
{
        struct in_addr ia = {.s_addr = 0};

        if (ss->ss_family == AF_INET)
                 ia.s_addr = ((struct sockaddr_in *)ss)->sin_addr.s_addr;
        if (ss->ss_family == AF_INET6) {
                const struct in6_addr v6_addr = ((struct sockaddr_in6 *)ss)->sin6_addr;

                if (IN6_IS_ADDR_V4MAPPED(&v6_addr))
                        memcpy(&ia.s_addr, &(v6_addr.s6_addr[12]), 4);
        }
        return ia;
}

/*
 * Send out a buffer, resend on short writes.
 *
 * \param fd The file descriptor.
 * \param buf The buffer to be sent.
 * \param len The length of \a buf.
 *
 * \return Standard. In any case, the number of bytes actually sent is stored
 * in \a len.
 */
static int sendall(int fd, const char *buf, size_t *len)
{
        size_t total = *len;

        assert(total);
        *len = 0;
        while (*len < total) {
                int ret = send(fd, buf + *len, total - *len, 0);
                if (ret == -1)
                        return -ERRNO_TO_PARA_ERROR(errno);
                *len += ret;
        }
        return 1;
}

/**
 * Encrypt and send a binary buffer.
 *
 * \param fd The file descriptor.
 * \param buf The buffer to be encrypted and sent.
 * \param len The length of \a buf.
 *
 * Check if encryption is available. If yes, encrypt the given buffer.  Send
 * out the buffer, encrypted or not, and try to resend the remaing part in case
 * of short writes.
 *
 * \return Standard.
 */
int send_bin_buffer(int fd, const char *buf, size_t len)
{
        int ret;
        crypt_function *cf = NULL;

        if (!len)
                PARA_CRIT_LOG("%s", "len == 0\n");
        if (fd + 1 <= cda_size)
                cf = crypt_data_array[fd].send;
        if (cf) {
                void *private = crypt_data_array[fd].private_data;
                /* RC4 may write more than len to the output buffer */
                unsigned char *outbuf = para_malloc(ROUND_UP(len, 8));
                (*cf)(len, (unsigned char *)buf, outbuf, private);
                ret = sendall(fd, (char *)outbuf, &len);
                free(outbuf);
        } else
                ret = sendall(fd, buf, &len);
        return ret;
}

/**
 * Encrypt and send null terminated buffer.
 *
 * \param fd The file descriptor.
 * \param buf The null-terminated buffer to be send.
 *
 * This is equivalent to send_bin_buffer(fd, buf, strlen(buf)).
 *
 * \return Standard.
 */
int send_buffer(int fd, const char *buf)
{
        return send_bin_buffer(fd, buf, strlen(buf));
}


/**
 * Send and encrypt a buffer given by a format string.
 *
 * \param fd The file descriptor.
 * \param fmt A format string.
 *
 * \return Standard.
 */
__printf_2_3 int send_va_buffer(int fd, const char *fmt, ...)
{
        char *msg;
        int ret;

        PARA_VSPRINTF(fmt, msg);
        ret = send_buffer(fd, msg);
        free(msg);
        return ret;
}

/**
 * Receive and decrypt.
 *
 * \param fd The file descriptor.
 * \param buf The buffer to write the decrypted data to.
 * \param size The size of \a buf.
 *
 * Receive at most \a size bytes from file descriptor \a fd. If encryption is
 * available, decrypt the received buffer.
 *
 * \return The number of bytes received on success, negative on errors.
 *
 * \sa recv(2)
 */
__must_check int recv_bin_buffer(int fd, char *buf, size_t size)
{
        ssize_t n;
        crypt_function *cf = NULL;

        if (fd + 1 <= cda_size)
                cf = crypt_data_array[fd].recv;
        if (cf) {
                unsigned char *tmp = para_malloc(size);
                void *private = crypt_data_array[fd].private_data;
                n = recv(fd, tmp, size, 0);
                if (n > 0) {
                        size_t numbytes = n;
                        unsigned char *b = (unsigned char *)buf;
                        (*cf)(numbytes, tmp, b, private);
                }
                free(tmp);
        } else
                n = recv(fd, buf, size, 0);
        if (n == -1)
                return -ERRNO_TO_PARA_ERROR(errno);
        return n;
}

/**
 * Receive, decrypt and write terminating NULL byte.
 *
 * \param fd The file descriptor.
 * \param buf The buffer to write the decrypted data to.
 * \param size The size of \a buf.
 *
 * Read and decrypt at most \a size - 1 bytes from file descriptor \a fd and
 * write a NULL byte at the end of the received data.
 *
 * \return The return value of the underlying call to \a recv_bin_buffer().
 *
 * \sa recv_bin_buffer()
 */
int recv_buffer(int fd, char *buf, size_t size)
{
        int n;

        assert(size);
        n = recv_bin_buffer(fd, buf, size - 1);
        if (n >= 0)
                buf[n] = '\0';
        else
                *buf = '\0';
        return n;
}

/**
 * Wrapper around the accept system call.
 *
 * \param fd The listening socket.
 * \param addr Structure which is filled in with the address of the peer socket.
 * \param size Should contain the size of the structure pointed to by \a addr.
 *
 * Accept incoming connections on \a addr. Retry if interrupted.
 *
 * \return The new file descriptor on success, negative on errors.
 *
 * \sa accept(2).
 */
int para_accept(int fd, void *addr, socklen_t size)
{
        int new_fd;

        do
                new_fd = accept(fd, (struct sockaddr *) addr, &size);
        while (new_fd < 0 && errno == EINTR);
        return new_fd < 0? -ERRNO_TO_PARA_ERROR(errno) : new_fd;
}

/**
 * Prepare a structure for \p AF_UNIX socket addresses.
 *
 * \param u Pointer to the struct to be prepared.
 * \param name The socket pathname.
 *
 * This just copies \a name to the sun_path component of \a u.
 *
 * \return Positive on success, \p -E_NAME_TOO_LONG if \a name is longer
 * than \p UNIX_PATH_MAX.
 */
static int init_unix_addr(struct sockaddr_un *u, const char *name)
{
        if (strlen(name) >= UNIX_PATH_MAX)
                return -E_NAME_TOO_LONG;
        memset(u->sun_path, 0, UNIX_PATH_MAX);
        u->sun_family = PF_UNIX;
        strcpy(u->sun_path, name);
        return 1;
}

/**
 * Prepare, create, and bind a socket for local communication.
 *
 * \param name The socket pathname.
 * \param unix_addr Pointer to the \p AF_UNIX socket structure.
 * \param mode The desired mode of the socket.
 *
 * This function creates a local socket for sequenced, reliable,
 * two-way, connection-based byte streams.
 *
 * \return The file descriptor, on success, negative on errors.
 *
 * \sa socket(2)
 * \sa bind(2)
 * \sa chmod(2)
 */
int create_local_socket(const char *name, struct sockaddr_un *unix_addr,
                mode_t mode)
{
        int fd, ret;

        ret = init_unix_addr(unix_addr, name);
        if (ret < 0)
                return ret;
        ret = socket(PF_UNIX, SOCK_STREAM, 0);
        if (ret < 0)
                return -ERRNO_TO_PARA_ERROR(errno);
        fd = ret;
        ret = bind(fd, (struct sockaddr *) unix_addr, UNIX_PATH_MAX);
        if (ret < 0) {
                ret = -ERRNO_TO_PARA_ERROR(errno);
                goto err;
        }
        ret = -E_CHMOD;
        if (chmod(name, mode) < 0)
                goto err;
        return fd;
err:
        close(fd);
        return ret;
}

/**
 * Prepare, create, and connect to a Unix domain socket for local communication.
 *
 * \param name The socket pathname.
 *
 * This function creates a local socket for sequenced, reliable, two-way,
 * connection-based byte streams.
 *
 * \return The file descriptor, on success, negative on errors.
 *
 * \sa create_local_socket(), unix(7), connect(2).
 */
int create_remote_socket(const char *name)
{
        struct sockaddr_un unix_addr;
        int fd, ret;

        ret = init_unix_addr(&unix_addr, name);
        if (ret < 0)
                return ret;
        fd = socket(PF_UNIX, SOCK_STREAM, 0);
        if (fd < 0)
                return -ERRNO_TO_PARA_ERROR(errno);
        if (connect(fd, (struct sockaddr *)&unix_addr, sizeof(unix_addr)) == -1) {
                ret = -ERRNO_TO_PARA_ERROR(errno);
                goto err;
        }
        return fd;
err:
        close(fd);
        return ret;
}

#ifndef HAVE_UCRED
ssize_t send_cred_buffer(int sock, char *buf)
{
        return send_buffer(sock, buf);
}
int recv_cred_buffer(int fd, char *buf, size_t size)
{
        return recv_buffer(fd, buf, size) > 0? 1 : -E_RECVMSG;
}
#else /* HAVE_UCRED */
/**
 * Send \p NULL-terminated buffer and Unix credentials of the current process.
 *
 * \param sock The socket file descriptor.
 * \param buf The buffer to be sent.
 *
 * \return On success, this call returns the number of characters sent.  On
 * error, \p -E_SENDMSG is returned.
 *
 * \sa sendmsg(2), okir's Black Hats Manual.
 */
ssize_t send_cred_buffer(int sock, char *buf)
{
        char control[sizeof(struct cmsghdr) + 10];
        struct msghdr msg;
        struct cmsghdr *cmsg;
        static struct iovec iov;
        struct ucred c;
        int ret;

        /* Response data */
        iov.iov_base = buf;
        iov.iov_len  = strlen(buf);
        c.pid = getpid();
        c.uid = getuid();
        c.gid = getgid();
        /* compose the message */
        memset(&msg, 0, sizeof(msg));
        msg.msg_iov = &iov;
        msg.msg_iovlen = 1;
        msg.msg_control = control;
        msg.msg_controllen = sizeof(control);
        /* attach the ucred struct */
        cmsg = CMSG_FIRSTHDR(&msg);
        cmsg->cmsg_level = SOL_SOCKET;
        cmsg->cmsg_type = SCM_CREDENTIALS;
        cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
        *(struct ucred *)CMSG_DATA(cmsg) = c;
        msg.msg_controllen = cmsg->cmsg_len;
        ret = sendmsg(sock, &msg, 0);
        if (ret  < 0)
                ret = -E_SENDMSG;
        return ret;
}

static void dispose_fds(int *fds, unsigned num)
{
        int i;

        for (i = 0; i < num; i++)
                close(fds[i]);
}

/**
 * Receive a buffer and the Unix credentials of the sending process.
 *
 * \param fd the socket file descriptor.
 * \param buf the buffer to store the message.
 * \param size the size of \a buffer.
 *
 * \return negative on errors, the user id on success.
 *
 * \sa recvmsg(2), okir's Black Hats Manual.
 */
int recv_cred_buffer(int fd, char *buf, size_t size)
{
        char control[255];
        struct msghdr msg;
        struct cmsghdr *cmsg;
        struct iovec iov;
        int result = 0;
        int yes = 1;
        struct ucred cred;

        setsockopt(fd, SOL_SOCKET, SO_PASSCRED, &yes, sizeof(int));
        memset(&msg, 0, sizeof(msg));
        memset(buf, 0, size);
        iov.iov_base = buf;
        iov.iov_len = size;
        msg.msg_iov = &iov;
        msg.msg_iovlen = 1;
        msg.msg_control = control;
        msg.msg_controllen = sizeof(control);
        if (recvmsg(fd, &msg, 0) < 0)
                return -E_RECVMSG;
        result = -E_SCM_CREDENTIALS;
        cmsg = CMSG_FIRSTHDR(&msg);
        while (cmsg) {
                if (cmsg->cmsg_level == SOL_SOCKET && cmsg->cmsg_type
                                == SCM_CREDENTIALS) {
                        memcpy(&cred, CMSG_DATA(cmsg), sizeof(struct ucred));
                        result = cred.uid;
                } else
                        if (cmsg->cmsg_level == SOL_SOCKET
                                        && cmsg->cmsg_type == SCM_RIGHTS) {
                                dispose_fds((int *) CMSG_DATA(cmsg),
                                        (cmsg->cmsg_len - CMSG_LEN(0))
                                        / sizeof(int));
                        }
                cmsg = CMSG_NXTHDR(&msg, cmsg);
        }
        return result;
}
#endif /* HAVE_UCRED */

/**
 * Receive a buffer and check for a pattern.
 *
 * \param fd The file descriptor to receive from.
 * \param pattern The expected pattern.
 * \param bufsize The size of the internal buffer.
 *
 * \return Positive if \a pattern was received, negative otherwise.
 *
 * This function creates a buffer of size \a bufsize and tries
 * to receive at most \a bufsize bytes from file descriptor \a fd.
 * If at least \p strlen(\a pattern) bytes were received, the beginning of
 * the received buffer is compared with \a pattern, ignoring case.
 *
 * \sa recv_buffer(), \sa strncasecmp(3).
 */
int recv_pattern(int fd, const char *pattern, size_t bufsize)
{
        size_t len = strlen(pattern);
        char *buf = para_malloc(bufsize + 1);
        int ret = -E_RECV_PATTERN, n = recv_buffer(fd, buf, bufsize);

        if (n < len)
                goto out;
        if (strncasecmp(buf, pattern, len))
                goto out;
        ret = 1;
out:
        if (ret < 0) {
                PARA_NOTICE_LOG("n = %d, did not receive pattern '%s'\n", n,
                        pattern);
                if (n > 0)
                        PARA_NOTICE_LOG("recvd: %s\n", buf);
        }
        free(buf);
        return ret;
}