Beispiel #1
0
static void kernel_receive(struct sk_buff *skb)
{
    struct nlmsghdr *nlh = NULL;
    int len;
    nlh = nlmsg_hdr(skb);
    len = skb->len;

    while(nlmsg_ok(nlh, len)) {
        if (down_trylock(&receive_sem)) {
            return;
        }

        if (nlh->nlmsg_type == NL_U_PID) {
            write_lock_bh(&user_proc.lock);
            user_proc.pid = nlh->nlmsg_pid;
            write_unlock_bh(&user_proc.lock);
        } else if (nlh->nlmsg_type == NL_CLOSE) {
            write_lock_bh(&user_proc.lock);
            if (nlh->nlmsg_pid == user_proc.pid) {
                user_proc.pid = 0;
            }
            write_unlock_bh(&user_proc.lock);
        }

        up(&receive_sem);
        nlh = nlmsg_next(nlh, &len);
    }

}
/**
 * Netlink handler, called when a user space process writes into the netlink.
 *
 * Note:
 * - The socket buffer skb given as argument might contain multiple Netlink
 *   packet. We ack each of them.
 */
static void kct_receive_skb(struct sk_buff *skb)
{
	struct nlmsghdr *nlh = NULL;
	int len = 0;
	int err = 0;

	pr_debug("%s: message received on the socket.\n", __func__);

	nlh = nlmsg_hdr(skb);
	len = skb->len;

	while (nlmsg_ok(nlh, len)) {
		err = kct_receive_msg(skb, nlh);
		 if (err || (nlh->nlmsg_flags & NLM_F_ACK))
			netlink_ack(skb, nlh, err);

		nlh = nlmsg_next(nlh, &len);
	}
}
Beispiel #3
0
static int nl_recv_adpa(struct nl_handle *handle, struct sockaddr_nl *nla, unsigned char **buf, struct ucred **creds)
{
         int n=0;
	 int flags = 0;
         static int page_size = 0;
         struct iovec iov;
         struct msghdr msg = {
                 .msg_name = (void *)nla,
                 .msg_namelen = sizeof(struct sockaddr_nl),
                 .msg_iov = &iov,
                 .msg_iovlen = 1,
                 .msg_control = NULL,
                 .msg_controllen = 0,
                 .msg_flags = 0,
         };
	 struct cmsghdr *cmsg;
 	 if (handle->h_flags & NL_MSG_PEEK)
                 flags |= MSG_PEEK;
         if (page_size == 0)
                 page_size = getpagesize();
 	 
         iov.iov_len = page_size;
         iov.iov_base = *buf = calloc(1, iov.iov_len);
		
         if (handle->h_flags & NL_SOCK_PASSCRED) {
                 msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred*));
                 msg.msg_control = calloc(1, msg.msg_controllen);
         }
 	 
 retry:
	 n = recvmsg(listenfd, &msg, flags);
	 msg.msg_namelen = sizeof(struct sockaddr_nl);
	 if (!n)
                goto abort;
	 else if (n < 0) {
                 if (errno == EINTR) {
                         printf("recvmsg() returned EINTR, retrying\n");
                         goto retry;
                 } else if(errno == EAGAIN) {
                         printf("recvmsg() returned EAGAIN, aborting\n");
                         goto abort;
		}else {
			 free(msg.msg_control);
                         free(*buf);
                         printf("recvmsg failed");
			 goto abort;
		}
		
         }
	  if (iov.iov_len < n ||
             msg.msg_flags & MSG_TRUNC) {
	         printf("iov & msg.msg_flags & MSG_TRUNC\n");
                 /* Provided buffer is not long enough, enlarge it
                  * and try again. */
                 iov.iov_len *= 2;
                 iov.iov_base = *buf = realloc(*buf, iov.iov_len);
                 goto retry;
	 }
	 else if (msg.msg_flags & MSG_CTRUNC) {
		 printf("msg.msg_flags & MSG_TRUNC\n");
                 msg.msg_controllen *= 2;
                 msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
                 goto retry;
         } else if (flags != 0) {
		  printf("flags");
                 /* Buffer is big enough, do the actual reading */
                 flags = 0;
                 goto retry;
         }
	 if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
		 printf("socket address size mismatch\n");
		 goto abort;
         }
	  for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
                 if (cmsg->cmsg_level == SOL_SOCKET) {
                         *creds = calloc(1, sizeof(struct ucred*));
                         memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred*));
                         break;
                 }
         }
	 free(msg.msg_control);
	 close(listenfd);
	 return n;
abort:
         free(msg.msg_control);
         free(*buf);
         return 0;
}
static int nl_recvmsgs_adpa(struct nl_handle *handle, struct nl_cb *cb)
{
	int n, err = 0;
	struct nlmsghdr *hdr;
	int multipart = 0;
	struct sockaddr_nl nla = {0};
	struct nl_msg *msg = NULL;
	struct ucred *creds = NULL;
	unsigned char *buf = NULL;
continue_reading:
	
	n = nl_recv_adpa(handle,&nla,&buf,&creds);
	if(n <= 0)
		return n;
	hdr = (struct nlmsghdr *)buf;
	while (nlmsg_ok(hdr, n)) {
		printf("nlmsg_ok\n");
		nlmsg_free(msg);
		msg = nlmsg_convert(hdr);
		if (!msg) {
			printf("msg error\n");
			err=-1;
			//err = nl_errno(ENOMEM);
			goto out;
		}

		nlmsg_set_proto(msg, handle->h_proto);
		//nlmsg_set_src(msg, &nla);
		/*if (creds)
			nlmsg_set_creds(msg, creds);
		if (cb->cb_set[NL_CB_MSG_IN])
			NL_CB_CALL(cb, NL_CB_MSG_IN, msg);
		if (cb->cb_set[NL_CB_SEQ_CHECK])
			NL_CB_CALL(cb, NL_CB_SEQ_CHECK, msg);
		else if (hdr->nlmsg_seq != handle->h_seq_expect){
			if (cb->cb_set[NL_CB_INVALID])
				NL_CB_CALL(cb, NL_CB_INVALID, msg);
			else {
				//err=-1;
				//err = nl_error(EINVAL,
					//"Sequence number mismatch");
				printf("Sequence number mismatch\n");				
				//goto out;
			}
		}
		if (hdr->nlmsg_type == NLMSG_DONE ||
		    hdr->nlmsg_type == NLMSG_ERROR ||
		    hdr->nlmsg_type == NLMSG_NOOP ||
		    hdr->nlmsg_type == NLMSG_OVERRUN) {
			handle->h_seq_expect++;
			//NL_DBG(3, "recvmsgs(%p): Increased expected " \
			       "sequence number to %d\n",
			  //     handle, handle->h_seq_expect);
		}

		if (hdr->nlmsg_flags & NLM_F_MULTI)
			multipart = 1;
	
		if (hdr->nlmsg_flags & NLM_F_ACK) {
			if (cb->cb_set[NL_CB_SEND_ACK])
				NL_CB_CALL(cb, NL_CB_SEND_ACK, msg);
			else {
				
			}
		}
		if (hdr->nlmsg_type == NLMSG_DONE) {
			multipart = 0;
			if (cb->cb_set[NL_CB_FINISH]) {
				NL_CB_CALL(cb, NL_CB_FINISH, msg);
			}
		}else if (hdr->nlmsg_type == NLMSG_NOOP) {
			if (cb->cb_set[NL_CB_SKIPPED])
				NL_CB_CALL(cb, NL_CB_SKIPPED, msg);
			else {
				printf("goto skip\n");
				goto skip;
			     }
		}else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
			if (cb->cb_set[NL_CB_OVERRUN])
				NL_CB_CALL(cb, NL_CB_OVERRUN, msg);
			else {
				printf("Overrun\n");
				err=-1;
			//	err = nl_error(EOVERFLOW, "Overrun");
				goto out;
			}
		}

		
		else if (hdr->nlmsg_type == NLMSG_ERROR) {
			struct nlmsgerr *e = nlmsg_data(hdr);
			printf("nlmsg_next3\n");
			if (hdr->nlmsg_len < nlmsg_msg_size(sizeof(*e))) {
				if (cb->cb_set[NL_CB_INVALID])
					NL_CB_CALL(cb, NL_CB_INVALID, msg);
				else {
					printf("Truncated error\n");
					err=-1;
					//err = nl_error(EINVAL,
					  //      "Truncated error message");
					goto out;
				}
			} else if (e->error) {		
			    if (cb->cb_err) {
					printf("cb->cb_err\n");
					err = cb->cb_err(&nla, e,
						   cb->cb_err_arg);
					if (err < 0) {
						printf("err<0\n");
						goto out;
					}
					else if (err == NL_SKIP) {
						printf("err:goto skip\n");
						goto skip;
					}
					else if (err == NL_STOP) {
						printf("err:Netlink Error\n");
						err=-1;
						//err = nl_error(-e->error,
						//         "Netlink Error");
						goto out;
					}
				} else {
				printf("Netlink Error\n");
				err=-1;
				//err = nl_error(-e->error,
					//	  "Netlink Error");
				goto out;
				}
			} else if (cb->cb_set[NL_CB_ACK]) {
				NL_CB_CALL(cb, NL_CB_ACK, msg);
				goto skip;
			}
		} */
		//else {
			if (cb->cb_set[NL_CB_VALID]) {
				NL_CB_CALL(cb, NL_CB_VALID, msg);
			}
	    	//}
skip:
		err = 0;
		hdr = nlmsg_next(hdr, &n);
	}
	nlmsg_free(msg);
	free(buf);
	free(creds);
	buf = NULL;
	msg = NULL;
	creds = NULL;
	//if (multipart) {
//		/* Multipart message not yet complete, continue reading */
	//	goto continue_reading;
	//}
stop:
	err = 0;
out:
	nlmsg_free(msg);
	free(buf);
	free(creds);
	return err;
}
Beispiel #4
0
/**
 * Receive data from netlink socket
 * @arg sk		Netlink socket.
 * @arg nla		Destination pointer for peer's netlink address.
 * @arg buf		Destination pointer for message content.
 * @arg creds		Destination pointer for credentials.
 *
 * Receives a netlink message, allocates a buffer in \c *buf and
 * stores the message content. The peer's netlink address is stored
 * in \c *nla. The caller is responsible for freeing the buffer allocated
 * in \c *buf if a positive value is returned.  Interrupted system calls
 * are handled by repeating the read. The input buffer size is determined
 * by peeking before the actual read is done.
 *
 * A non-blocking sockets causes the function to return immediately with
 * a return value of 0 if no data is available.
 *
 * @return Number of octets read, 0 on EOF or a negative error code.
 */
int nl_recv(struct nl_sock *sk, struct sockaddr_nl *nla,
	    unsigned char **buf, struct ucred **creds)
{
	int n;
	int flags = 0;
	static int page_size = 0;
	struct iovec iov;
	struct msghdr msg = {
		.msg_name = (void *) nla,
		.msg_namelen = sizeof(struct sockaddr_nl),
		.msg_iov = &iov,
		.msg_iovlen = 1,
		.msg_control = NULL,
		.msg_controllen = 0,
		.msg_flags = 0,
	};
	struct cmsghdr *cmsg;

	memset(nla, 0, sizeof(*nla));

	if (sk->s_flags & NL_MSG_PEEK)
		flags |= MSG_PEEK;

	if (page_size == 0)
		page_size = getpagesize();

	iov.iov_len = page_size;
	iov.iov_base = *buf = malloc(iov.iov_len);

	if (sk->s_flags & NL_SOCK_PASSCRED) {
		msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred));
		msg.msg_control = calloc(1, msg.msg_controllen);
	}
retry:

	n = recvmsg(sk->s_fd, &msg, flags);
	if (!n)
		goto abort;
	else if (n < 0) {
		if (errno == EINTR) {
			NL_DBG(3, "recvmsg() returned EINTR, retrying\n");
			goto retry;
		} else if (errno == EAGAIN) {
			NL_DBG(3, "recvmsg() returned EAGAIN, aborting\n");
			goto abort;
		} else {
			free(msg.msg_control);
			free(*buf);
			return -nl_syserr2nlerr(errno);
		}
	}

	if (iov.iov_len < n ||
	    msg.msg_flags & MSG_TRUNC) {
		/* Provided buffer is not long enough, enlarge it
		 * and try again. */
		iov.iov_len *= 2;
		iov.iov_base = *buf = realloc(*buf, iov.iov_len);
		goto retry;
	} else if (msg.msg_flags & MSG_CTRUNC) {
		msg.msg_controllen *= 2;
		msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
		goto retry;
	} else if (flags != 0) {
		/* Buffer is big enough, do the actual reading */
		flags = 0;
		goto retry;
	}

	if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
		free(msg.msg_control);
		free(*buf);
		return -NLE_NOADDR;
	}

	for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
		if (cmsg->cmsg_level == SOL_SOCKET &&
		    cmsg->cmsg_type == SCM_CREDENTIALS) {
			if (creds) {
				*creds = calloc(1, sizeof(struct ucred));
				memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred));
			}
			break;
		}
	}

	free(msg.msg_control);
	return n;

abort:
	free(msg.msg_control);
	free(*buf);
	return 0;
}

#define NL_CB_CALL(cb, type, msg) \
do { \
	err = nl_cb_call(cb, type, msg); \
	switch (err) { \
	case NL_OK: \
		err = 0; \
		break; \
	case NL_SKIP: \
		goto skip; \
	case NL_STOP: \
		goto stop; \
	default: \
		goto out; \
	} \
} while (0)

static int recvmsgs(struct nl_sock *sk, struct nl_cb *cb)
{
	int n, err = 0, multipart = 0, interrupted = 0;
	unsigned char *buf = NULL;
	struct nlmsghdr *hdr;
	struct sockaddr_nl nla = {0};
	struct nl_msg *msg = NULL;
	struct ucred *creds = NULL;

continue_reading:
	NL_DBG(3, "Attempting to read from %p\n", sk);
	if (cb->cb_recv_ow)
		n = cb->cb_recv_ow(sk, &nla, &buf, &creds);
	else
		n = nl_recv(sk, &nla, &buf, &creds);

	if (n <= 0)
		return n;

	NL_DBG(3, "recvmsgs(%p): Read %d bytes\n", sk, n);

	hdr = (struct nlmsghdr *) buf;
	while (nlmsg_ok(hdr, n)) {
		NL_DBG(3, "recvmsgs(%p): Processing valid message...\n", sk);

		nlmsg_free(msg);
		msg = nlmsg_convert(hdr);
		if (!msg) {
			err = -NLE_NOMEM;
			goto out;
		}

		nlmsg_set_proto(msg, sk->s_proto);
		nlmsg_set_src(msg, &nla);
		if (creds)
			nlmsg_set_creds(msg, creds);

		/* Raw callback is the first, it gives the most control
		 * to the user and he can do his very own parsing. */
		if (cb->cb_set[NL_CB_MSG_IN])
			NL_CB_CALL(cb, NL_CB_MSG_IN, msg);

		/* Sequence number checking. The check may be done by
		 * the user, otherwise a very simple check is applied
		 * enforcing strict ordering */
		if (cb->cb_set[NL_CB_SEQ_CHECK]) {
			NL_CB_CALL(cb, NL_CB_SEQ_CHECK, msg);

		/* Only do sequence checking if auto-ack mode is enabled */
		} else if (!(sk->s_flags & NL_NO_AUTO_ACK)) {
			if (hdr->nlmsg_seq != sk->s_seq_expect) {
				if (cb->cb_set[NL_CB_INVALID])
					NL_CB_CALL(cb, NL_CB_INVALID, msg);
				else {
					err = -NLE_SEQ_MISMATCH;
					goto out;
				}
			}
		}

		if (hdr->nlmsg_type == NLMSG_DONE ||
		    hdr->nlmsg_type == NLMSG_ERROR ||
		    hdr->nlmsg_type == NLMSG_NOOP ||
		    hdr->nlmsg_type == NLMSG_OVERRUN) {
			/* We can't check for !NLM_F_MULTI since some netlink
			 * users in the kernel are broken. */
			sk->s_seq_expect++;
			NL_DBG(3, "recvmsgs(%p): Increased expected " \
			       "sequence number to %d\n",
			       sk, sk->s_seq_expect);
		}

		if (hdr->nlmsg_flags & NLM_F_MULTI)
			multipart = 1;

		if (hdr->nlmsg_flags & NLM_F_DUMP_INTR) {
			if (cb->cb_set[NL_CB_DUMP_INTR])
				NL_CB_CALL(cb, NL_CB_DUMP_INTR, msg);
			else {
				/*
				 * We have to continue reading to clear
				 * all messages until a NLMSG_DONE is
				 * received and report the inconsistency.
				 */
				interrupted = 1;
			}
		}
	
		/* Other side wishes to see an ack for this message */
		if (hdr->nlmsg_flags & NLM_F_ACK) {
			if (cb->cb_set[NL_CB_SEND_ACK])
				NL_CB_CALL(cb, NL_CB_SEND_ACK, msg);
			else {
				/* FIXME: implement */
			}
		}

		/* messages terminates a multpart message, this is
		 * usually the end of a message and therefore we slip
		 * out of the loop by default. the user may overrule
		 * this action by skipping this packet. */
		if (hdr->nlmsg_type == NLMSG_DONE) {
			multipart = 0;
			if (cb->cb_set[NL_CB_FINISH])
				NL_CB_CALL(cb, NL_CB_FINISH, msg);
		}

		/* Message to be ignored, the default action is to
		 * skip this message if no callback is specified. The
		 * user may overrule this action by returning
		 * NL_PROCEED. */
		else if (hdr->nlmsg_type == NLMSG_NOOP) {
			if (cb->cb_set[NL_CB_SKIPPED])
				NL_CB_CALL(cb, NL_CB_SKIPPED, msg);
			else
				goto skip;
		}

		/* Data got lost, report back to user. The default action is to
		 * quit parsing. The user may overrule this action by retuning
		 * NL_SKIP or NL_PROCEED (dangerous) */
		else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
			if (cb->cb_set[NL_CB_OVERRUN])
				NL_CB_CALL(cb, NL_CB_OVERRUN, msg);
			else {
				err = -NLE_MSG_OVERFLOW;
				goto out;
			}
		}

		/* Message carries a nlmsgerr */
		else if (hdr->nlmsg_type == NLMSG_ERROR) {
			struct nlmsgerr *e = nlmsg_data(hdr);

			if (hdr->nlmsg_len < nlmsg_size(sizeof(*e))) {
				/* Truncated error message, the default action
				 * is to stop parsing. The user may overrule
				 * this action by returning NL_SKIP or
				 * NL_PROCEED (dangerous) */
				if (cb->cb_set[NL_CB_INVALID])
					NL_CB_CALL(cb, NL_CB_INVALID, msg);
				else {
					err = -NLE_MSG_TRUNC;
					goto out;
				}
			} else if (e->error) {
				/* Error message reported back from kernel. */
				if (cb->cb_err) {
					err = cb->cb_err(&nla, e,
							   cb->cb_err_arg);
					if (err < 0)
						goto out;
					else if (err == NL_SKIP)
						goto skip;
					else if (err == NL_STOP) {
						err = -nl_syserr2nlerr(e->error);
						goto out;
					}
				} else {
					err = -nl_syserr2nlerr(e->error);
					goto out;
				}
			} else if (cb->cb_set[NL_CB_ACK])
				NL_CB_CALL(cb, NL_CB_ACK, msg);
		} else {
			/* Valid message (not checking for MULTIPART bit to
			 * get along with broken kernels. NL_SKIP has no
			 * effect on this.  */
			if (cb->cb_set[NL_CB_VALID])
				NL_CB_CALL(cb, NL_CB_VALID, msg);
		}
skip:
		err = 0;
		hdr = nlmsg_next(hdr, &n);
	}
	
	nlmsg_free(msg);
	free(buf);
	free(creds);
	buf = NULL;
	msg = NULL;
	creds = NULL;

	if (multipart) {
		/* Multipart message not yet complete, continue reading */
		goto continue_reading;
	}
stop:
	err = 0;
out:
	nlmsg_free(msg);
	free(buf);
	free(creds);

	if (interrupted)
		err = -NLE_DUMP_INTR;

	return err;
}
Beispiel #5
0
/**
 * Send netlink message with control over sendmsg() message header.
 * @arg handle		Netlink handle.
 * @arg msg		Netlink message to be sent.
 * @arg hdr		Sendmsg() message header.
 * @return Number of characters sent on sucess or a negative error code.
 */
int nl_sendmsg(struct nl_handle *handle, struct nl_msg *msg, struct msghdr *hdr)
{
	struct nl_cb *cb;
	int ret;

	struct iovec iov = {
		.iov_base = (void *) nlmsg_hdr(msg),
		.iov_len = nlmsg_hdr(msg)->nlmsg_len,
	};

	hdr->msg_iov = &iov;
	hdr->msg_iovlen = 1;

	nlmsg_set_src(msg, &handle->h_local);

	cb = nl_handle_get_cb(handle);
	if (cb->cb_set[NL_CB_MSG_OUT])
		if (nl_cb_call(cb, NL_CB_MSG_OUT, msg) != NL_PROCEED)
			return 0;

	ret = sendmsg(handle->h_fd, hdr, 0);
	if (ret < 0)
		return nl_errno(errno);

	return ret;
}


/**
 * Send netlink message.
 * @arg handle		Netlink handle
 * @arg msg		Netlink message to be sent.
 * @see nl_sendmsg()
 * @return Number of characters sent on success or a negative error code.
 */
int nl_send(struct nl_handle *handle, struct nl_msg *msg)
{
	struct sockaddr_nl *dst;
	struct ucred *creds;
	
	struct msghdr hdr = {
		.msg_name = (void *) &handle->h_peer,
		.msg_namelen = sizeof(struct sockaddr_nl),
	};

	/* Overwrite destination if specified in the message itself, defaults
	 * to the peer address of the handle.
	 */
	dst = nlmsg_get_dst(msg);
	if (dst->nl_family == AF_NETLINK)
		hdr.msg_name = dst;

	/* Add credentials if present. */
	creds = nlmsg_get_creds(msg);
	if (creds != NULL) {
		char buf[CMSG_SPACE(sizeof(struct ucred))];
		struct cmsghdr *cmsg;

		hdr.msg_control = buf;
		hdr.msg_controllen = sizeof(buf);

		cmsg = CMSG_FIRSTHDR(&hdr);
		cmsg->cmsg_level = SOL_SOCKET;
		cmsg->cmsg_type = SCM_CREDENTIALS;
		cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
		memcpy(CMSG_DATA(cmsg), creds, sizeof(struct ucred));
	}

	return nl_sendmsg(handle, msg, &hdr);
}

/**
 * Send netlink message and check & extend header values as needed.
 * @arg handle		Netlink handle.
 * @arg msg		Netlink message to be sent.
 *
 * Checks the netlink message \c nlh for completness and extends it
 * as required before sending it out. Checked fields include pid,
 * sequence nr, and flags.
 *
 * @see nl_send()
 * @return Number of characters sent or a negative error code.
 */
int nl_send_auto_complete(struct nl_handle *handle, struct nl_msg *msg)
{
	struct nlmsghdr *nlh;

	nlh = nlmsg_hdr(msg);
	if (nlh->nlmsg_pid == 0)
		nlh->nlmsg_pid = handle->h_local.nl_pid;

	if (nlh->nlmsg_seq == 0)
		nlh->nlmsg_seq = handle->h_seq_next++;
	
	nlh->nlmsg_flags |= (NLM_F_REQUEST | NLM_F_ACK);

	if (handle->h_cb->cb_send_ow)
		return handle->h_cb->cb_send_ow(handle, msg);
	else
		return nl_send(handle, msg);
}

/**
 * Send simple netlink message using nl_send_auto_complete()
 * @arg handle		Netlink handle.
 * @arg type		Netlink message type.
 * @arg flags		Netlink message flags.
 * @arg buf		Data buffer.
 * @arg size		Size of data buffer.
 *
 * Builds a netlink message with the specified type and flags and
 * appends the specified data as payload to the message.
 *
 * @see nl_send_auto_complete()
 * @return Number of characters sent on success or a negative error code.
 */
int nl_send_simple(struct nl_handle *handle, int type, int flags, void *buf,
		   size_t size)
{
	int err;
	struct nl_msg *msg;
	struct nlmsghdr nlh = {
		.nlmsg_len = nlmsg_msg_size(0),
		.nlmsg_type = type,
		.nlmsg_flags = flags,
	};

	msg = nlmsg_build(&nlh);
	if (!msg)
		return nl_errno(ENOMEM);

	if (buf && size)
		nlmsg_append(msg, buf, size, 1);

	err = nl_send_auto_complete(handle, msg);
	nlmsg_free(msg);

	return err;
}

/** @} */

/**
 * @name Receive
 * @{
 */

/**
 * Receive netlink message from netlink socket.
 * @arg handle		Netlink handle.
 * @arg nla		Destination pointer for peer's netlink address.
 * @arg buf		Destination pointer for message content.
 * @arg creds		Destination pointer for credentials.
 *
 * Receives a netlink message, allocates a buffer in \c *buf and
 * stores the message content. The peer's netlink address is stored
 * in \c *nla. The caller is responsible for freeing the buffer allocated
 * in \c *buf if a positive value is returned.  Interruped system calls
 * are handled by repeating the read. The input buffer size is determined
 * by peeking before the actual read is done.
 *
 * A non-blocking sockets causes the function to return immediately if
 * no data is available.
 *
 * @return Number of octets read, 0 on EOF or a negative error code.
 */
int nl_recv(struct nl_handle *handle, struct sockaddr_nl *nla,
	    unsigned char **buf, struct ucred **creds)
{
	int n;
	int flags = MSG_PEEK;

	struct iovec iov = {
		.iov_len = 4096,
	};

	struct msghdr msg = {
		.msg_name = (void *) nla,
		.msg_namelen = sizeof(sizeof(struct sockaddr_nl)),
		.msg_iov = &iov,
		.msg_iovlen = 1,
		.msg_control = NULL,
		.msg_controllen = 0,
		.msg_flags = 0,
	};
	struct cmsghdr *cmsg;

	iov.iov_base = *buf = calloc(1, iov.iov_len);

	if (handle->h_flags & NL_SOCK_PASSCRED) {
		msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred));
		msg.msg_control = calloc(1, msg.msg_controllen);
	}
retry:

	if ((n = recvmsg(handle->h_fd, &msg, flags)) <= 0) {
		if (!n)
			goto abort;
		else if (n < 0) {
			if (errno == EINTR)
				goto retry;
			else if (errno == EAGAIN)
				goto abort;
			else {
				free(msg.msg_control);
				free(*buf);
				return nl_error(errno, "recvmsg failed");
			}
		}
	}
	
	if (iov.iov_len < n) {
		/* Provided buffer is not long enough, enlarge it
		 * and try again. */
		iov.iov_len *= 2;
		iov.iov_base = *buf = realloc(*buf, iov.iov_len);
		goto retry;
	} else if (msg.msg_flags & MSG_CTRUNC) {
		msg.msg_controllen *= 2;
		msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
		goto retry;
	} else if (flags != 0) {
		/* Buffer is big enough, do the actual reading */
		flags = 0;
		goto retry;
	}

	if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
		free(msg.msg_control);
		free(*buf);
		return nl_error(EADDRNOTAVAIL, "socket address size mismatch");
	}

	for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
		if (cmsg->cmsg_level == SOL_SOCKET &&
		    cmsg->cmsg_type == SCM_CREDENTIALS) {
			*creds = calloc(1, sizeof(struct ucred));
			memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred));
			break;
		}
	}

	free(msg.msg_control);
	return n;

abort:
	free(msg.msg_control);
	free(*buf);
	return 0;
}


/**
 * Receive a set of messages from a netlink socket.
 * @arg handle		netlink handle
 * @arg cb		set of callbacks to control the behaviour.
 *
 * Repeatedly calls nl_recv() and parses the messages as netlink
 * messages. Stops reading if one of the callbacks returns
 * NL_EXIT or nl_recv returns either 0 or a negative error code.
 *
 * A non-blocking sockets causes the function to return immediately if
 * no data is available.
 *
 * @return 0 on success or a negative error code from nl_recv().
 */
int nl_recvmsgs(struct nl_handle *handle, struct nl_cb *cb)
{
	int n, err = 0;
	unsigned char *buf = NULL;
	struct nlmsghdr *hdr;
	struct sockaddr_nl nla = {0};
	struct nl_msg *msg = NULL;
	struct ucred *creds = NULL;

continue_reading:
	if (cb->cb_recv_ow)
		n = cb->cb_recv_ow(handle, &nla, &buf, &creds);
	else
		n = nl_recv(handle, &nla, &buf, &creds);

	if (n <= 0)
		return n;

	NL_DBG(3, "recvmsgs(%p): Read %d bytes\n", handle, n);

	hdr = (struct nlmsghdr *) buf;
	while (nlmsg_ok(hdr, n)) {
		NL_DBG(3, "recgmsgs(%p): Processing valid message...\n",
		       handle);

		nlmsg_free(msg);
		msg = nlmsg_convert(hdr);
		if (!msg) {
			err = nl_errno(ENOMEM);
			goto out;
		}

		nlmsg_set_proto(msg, handle->h_proto);
		nlmsg_set_src(msg, &nla);
		if (creds)
			nlmsg_set_creds(msg, creds);

		/* Raw callback is the first, it gives the most control
		 * to the user and he can do his very own parsing. */
		if (cb->cb_set[NL_CB_MSG_IN]) {
			err = nl_cb_call(cb, NL_CB_MSG_IN, msg);
			if (err == NL_SKIP)
				goto skip;
			else if (err == NL_EXIT || err < 0)
				goto out;
		}

		/* Sequence number checking. The check may be done by
		 * the user, otherwise a very simple check is applied
		 * enforcing strict ordering */
		if (cb->cb_set[NL_CB_SEQ_CHECK]) {
			err = nl_cb_call(cb, NL_CB_SEQ_CHECK, msg);
			if (err == NL_SKIP)
				goto skip;
			else if (err == NL_EXIT || err < 0)
				goto out;
		} else if (hdr->nlmsg_seq != handle->h_seq_expect) {
			if (cb->cb_set[NL_CB_INVALID]) {
				err = nl_cb_call(cb, NL_CB_INVALID, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			} else
				goto out;
		}

		if (hdr->nlmsg_type == NLMSG_DONE ||
		    hdr->nlmsg_type == NLMSG_ERROR ||
		    hdr->nlmsg_type == NLMSG_NOOP ||
		    hdr->nlmsg_type == NLMSG_OVERRUN) {
			/* We can't check for !NLM_F_MULTI since some netlink
			 * users in the kernel are broken. */
			handle->h_seq_expect++;
			NL_DBG(3, "recvmsgs(%p): Increased expected " \
			       "sequence number to %d\n",
			       handle, handle->h_seq_expect);
		}
	
		/* Other side wishes to see an ack for this message */
		if (hdr->nlmsg_flags & NLM_F_ACK) {
			if (cb->cb_set[NL_CB_SEND_ACK]) {
				err = nl_cb_call(cb, NL_CB_SEND_ACK, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			} else {
				/* FIXME: implement */
			}
		}

		/* messages terminates a multpart message, this is
		 * usually the end of a message and therefore we slip
		 * out of the loop by default. the user may overrule
		 * this action by skipping this packet. */
		if (hdr->nlmsg_type == NLMSG_DONE) {
			if (cb->cb_set[NL_CB_FINISH]) {
				err = nl_cb_call(cb, NL_CB_FINISH, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			}
			err = 0;
			goto out;
		}

		/* Message to be ignored, the default action is to
		 * skip this message if no callback is specified. The
		 * user may overrule this action by returning
		 * NL_PROCEED. */
		else if (hdr->nlmsg_type == NLMSG_NOOP) {
			if (cb->cb_set[NL_CB_SKIPPED]) {
				err = nl_cb_call(cb, NL_CB_SKIPPED, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			} else
				goto skip;
		}

		/* Data got lost, report back to user. The default action is to
		 * quit parsing. The user may overrule this action by retuning
		 * NL_SKIP or NL_PROCEED (dangerous) */
		else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
			if (cb->cb_set[NL_CB_OVERRUN]) {
				err = nl_cb_call(cb, NL_CB_OVERRUN, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			} else
				goto out;
		}

		/* Message carries a nlmsgerr */
		else if (hdr->nlmsg_type == NLMSG_ERROR) {
			struct nlmsgerr *e = nlmsg_data(hdr);

			if (hdr->nlmsg_len < nlmsg_msg_size(sizeof(*e))) {
				/* Truncated error message, the default action
				 * is to stop parsing. The user may overrule
				 * this action by returning NL_SKIP or
				 * NL_PROCEED (dangerous) */
				if (cb->cb_set[NL_CB_INVALID]) {
					err = nl_cb_call(cb, NL_CB_INVALID,
							 msg);
					if (err == NL_SKIP)
						goto skip;
					else if (err == NL_EXIT || err < 0)
						goto out;
				} else
					goto out;
			} else if (e->error) {
				/* Error message reported back from kernel. */
				if (cb->cb_err) {
					err = cb->cb_err(&nla, e,
							   cb->cb_err_arg);
					if (err == NL_SKIP)
						goto skip;
					else if (err == NL_EXIT || err < 0) {
						nl_error(-e->error,
							 "Netlink Error");
						err = e->error;
						goto out;
					}
				} else {
					nl_error(-e->error, "Netlink Error");
					err = e->error;
					goto out;
				}
			} else if (cb->cb_set[NL_CB_ACK]) {
				/* ACK */
				err = nl_cb_call(cb, NL_CB_ACK, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			}
		} else {
			/* Valid message (not checking for MULTIPART bit to
			 * get along with broken kernels. NL_SKIP has no
			 * effect on this.  */
			if (cb->cb_set[NL_CB_VALID]) {
				err = nl_cb_call(cb, NL_CB_VALID, msg);
				if (err == NL_SKIP)
					goto skip;
				else if (err == NL_EXIT || err < 0)
					goto out;
			}
		}
skip:
		hdr = nlmsg_next(hdr, &n);
	}
	
	nlmsg_free(msg);
	free(buf);
	free(creds);
	buf = NULL;
	msg = NULL;
	creds = NULL;

	/* Multipart message not yet complete, continue reading */
	goto continue_reading;

out:
	nlmsg_free(msg);
	free(buf);
	free(creds);

	return err;
}

/**
 * Receive a set of message from a netlink socket using handlers in nl_handle.
 * @arg handle		netlink handle
 *
 * Calls nl_recvmsgs() with the handlers configured in the netlink handle.
 */
int nl_recvmsgs_def(struct nl_handle *handle)
{
	if (handle->h_cb->cb_recvmsgs_ow)
		return handle->h_cb->cb_recvmsgs_ow(handle, handle->h_cb);
	else
		return nl_recvmsgs(handle, handle->h_cb);
}

static int ack_wait_handler(struct nl_msg *msg, void *arg)
{
	return NL_EXIT;
}

/**
 * Wait for ACK.
 * @arg handle		netlink handle
 * @pre The netlink socket must be in blocking state.
 *
 * Waits until an ACK is received for the latest not yet acknowledged
 * netlink message.
 */
int nl_wait_for_ack(struct nl_handle *handle)
{
	int err;
	struct nl_cb *cb = nl_cb_clone(nl_handle_get_cb(handle));

	nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, ack_wait_handler, NULL);

	err = nl_recvmsgs(handle, cb);
	nl_cb_destroy(cb);

	return err;
}

/** @} */

/**
 * @name Netlink Family Translations
 * @{
 */

static struct trans_tbl nlfamilies[] = {
	__ADD(NETLINK_ROUTE,route)
	__ADD(NETLINK_W1,w1)
	__ADD(NETLINK_USERSOCK,usersock)
	__ADD(NETLINK_FIREWALL,firewall)
	__ADD(NETLINK_INET_DIAG,inetdiag)
	__ADD(NETLINK_NFLOG,nflog)
	__ADD(NETLINK_XFRM,xfrm)
	__ADD(NETLINK_SELINUX,selinux)
	__ADD(NETLINK_ISCSI,iscsi)
	__ADD(NETLINK_AUDIT,audit)
	__ADD(NETLINK_FIB_LOOKUP,fib_lookup)
	__ADD(NETLINK_CONNECTOR,connector)
	__ADD(NETLINK_NETFILTER,netfilter)
	__ADD(NETLINK_IP6_FW,ip6_fw)
	__ADD(NETLINK_DNRTMSG,dnrtmsg)
	__ADD(NETLINK_KOBJECT_UEVENT,kobject_uevent)
	__ADD(NETLINK_GENERIC,generic)
};

/**
 * Convert netlink family to character string.
 * @arg family		Netlink family.
 * @arg buf		Destination buffer.
 * @arg size		Size of destination buffer.
 *
 * Converts a netlink family to a character string and stores it in
 * the specified destination buffer.
 *
 * @return The destination buffer or the family encoded in hexidecimal
 *         form if no match was found.
 */
char * nl_nlfamily2str(int family, char *buf, size_t size)
{
	return __type2str(family, buf, size, nlfamilies,
			  ARRAY_SIZE(nlfamilies));
}

/**
 * Convert character string to netlink family.
 * @arg name		Name of netlink family.
 *
 * Converts the provided character string specifying a netlink
 * family to the corresponding numeric value.
 *
 * @return Numeric netlink family or a negative value if no match was found.
 */
int nl_str2nlfamily(const char *name)
{
	return __str2type(name, nlfamilies, ARRAY_SIZE(nlfamilies));
}
/**
 * Send netlink message.
 * @arg sk		Netlink socket.
 * @arg msg		Netlink message to be sent.
 * @arg iov		iovec to be sent.
 * @arg iovlen		number of struct iovec to be sent.
 * @see nl_sendmsg()
 * @return Number of characters sent on success or a negative error code.
 */
int nl_send_iovec(struct nl_sock *sk, struct nl_msg *msg, struct iovec *iov, unsigned iovlen)
{
	struct sockaddr_nl *dst;
	struct ucred *creds;
	struct msghdr hdr = {
		.msg_name = (void *) &sk->s_peer,
		.msg_namelen = sizeof(struct sockaddr_nl),
		.msg_iov = iov,
		.msg_iovlen = iovlen,
	};

	/* Overwrite destination if specified in the message itself, defaults
	 * to the peer address of the socket.
	 */
	dst = nlmsg_get_dst(msg);
	if (dst->nl_family == AF_NETLINK)
		hdr.msg_name = dst;

	/* Add credentials if present. */
	creds = nlmsg_get_creds(msg);
	if (creds != NULL) {
		char buf[CMSG_SPACE(sizeof(struct ucred))];
		struct cmsghdr *cmsg;

		hdr.msg_control = buf;
		hdr.msg_controllen = sizeof(buf);

		cmsg = CMSG_FIRSTHDR(&hdr);
		cmsg->cmsg_level = SOL_SOCKET;
		cmsg->cmsg_type = SCM_CREDENTIALS;
		cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
		memcpy(CMSG_DATA(cmsg), creds, sizeof(struct ucred));
	}

	return nl_sendmsg(sk, msg, &hdr);
}



/**
* Send netlink message.
* @arg sk		Netlink socket.
* @arg msg		Netlink message to be sent.
* @see nl_sendmsg()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send(struct nl_sock *sk, struct nl_msg *msg)
{
	struct iovec iov = {
		.iov_base = (void *) nlmsg_hdr(msg),
		.iov_len = nlmsg_hdr(msg)->nlmsg_len,
	};

	return nl_send_iovec(sk, msg, &iov, 1);
}

void nl_auto_complete(struct nl_sock *sk, struct nl_msg *msg)
{
	struct nlmsghdr *nlh;

	nlh = nlmsg_hdr(msg);
	if (nlh->nlmsg_pid == 0)
		nlh->nlmsg_pid = sk->s_local.nl_pid;

	if (nlh->nlmsg_seq == 0)
		nlh->nlmsg_seq = sk->s_seq_next++;

	if (msg->nm_protocol == -1)
		msg->nm_protocol = sk->s_proto;

	nlh->nlmsg_flags |= NLM_F_REQUEST;

	if (!(sk->s_flags & NL_NO_AUTO_ACK))
		nlh->nlmsg_flags |= NLM_F_ACK;
}

/**
 * Send netlink message and check & extend header values as needed.
 * @arg sk		Netlink socket.
 * @arg msg		Netlink message to be sent.
 *
 * Checks the netlink message \c nlh for completness and extends it
 * as required before sending it out. Checked fields include pid,
 * sequence nr, and flags.
 *
 * @see nl_send()
 * @return Number of characters sent or a negative error code.
 */
int nl_send_auto_complete(struct nl_sock *sk, struct nl_msg *msg)
{
	struct nl_cb *cb = sk->s_cb;

	nl_auto_complete(sk, msg);

	if (cb->cb_send_ow)
		return cb->cb_send_ow(sk, msg);
	else
		return nl_send(sk, msg);
}

/**
 * Send simple netlink message using nl_send_auto_complete()
 * @arg sk		Netlink socket.
 * @arg type		Netlink message type.
 * @arg flags		Netlink message flags.
 * @arg buf		Data buffer.
 * @arg size		Size of data buffer.
 *
 * Builds a netlink message with the specified type and flags and
 * appends the specified data as payload to the message.
 *
 * @see nl_send_auto_complete()
 * @return Number of characters sent on success or a negative error code.
 */
int nl_send_simple(struct nl_sock *sk, int type, int flags, void *buf,
		   size_t size)
{
	int err;
	struct nl_msg *msg;

	msg = nlmsg_alloc_simple(type, flags);
	if (!msg)
		return -NLE_NOMEM;

	if (buf && size) {
		err = nlmsg_append(msg, buf, size, NLMSG_ALIGNTO);
		if (err < 0)
			goto errout;
	}
	

	err = nl_send_auto_complete(sk, msg);
errout:
	nlmsg_free(msg);

	return err;
}

/** @} */

/**
 * @name Receive
 * @{
 */

/**
 * Receive data from netlink socket
 * @arg sk		Netlink socket.
 * @arg nla		Destination pointer for peer's netlink address.
 * @arg buf		Destination pointer for message content.
 * @arg creds		Destination pointer for credentials.
 *
 * Receives a netlink message, allocates a buffer in \c *buf and
 * stores the message content. The peer's netlink address is stored
 * in \c *nla. The caller is responsible for freeing the buffer allocated
 * in \c *buf if a positive value is returned.  Interruped system calls
 * are handled by repeating the read. The input buffer size is determined
 * by peeking before the actual read is done.
 *
 * A non-blocking sockets causes the function to return immediately with
 * a return value of 0 if no data is available.
 *
 * @return Number of octets read, 0 on EOF or a negative error code.
 */
int nl_recv(struct nl_sock *sk, struct sockaddr_nl *nla,
	    unsigned char **buf, struct ucred **creds)
{
	int n;
	int flags = 0;
	static int page_size = 0;
	struct iovec iov;
	struct msghdr msg = {
		.msg_name = (void *) nla,
		.msg_namelen = sizeof(struct sockaddr_nl),
		.msg_iov = &iov,
		.msg_iovlen = 1,
		.msg_control = NULL,
		.msg_controllen = 0,
		.msg_flags = 0,
	};
	struct cmsghdr *cmsg;

	if (sk->s_flags & NL_MSG_PEEK)
		flags |= MSG_PEEK;

	if (page_size == 0)
		page_size = getpagesize();

	iov.iov_len = page_size;
	iov.iov_base = *buf = malloc(iov.iov_len);

	if (sk->s_flags & NL_SOCK_PASSCRED) {
		msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred));
		msg.msg_control = calloc(1, msg.msg_controllen);
	}
retry:

	n = recvmsg(sk->s_fd, &msg, flags);
	if (!n)
		goto abort;
	else if (n < 0) {
		if (errno == EINTR) {
			NL_DBG(3, "recvmsg() returned EINTR, retrying\n");
			goto retry;
		} else if (errno == EAGAIN) {
			NL_DBG(3, "recvmsg() returned EAGAIN, aborting\n");
			goto abort;
		} else {
			free(msg.msg_control);
			free(*buf);
			return -nl_syserr2nlerr(errno);
		}
	}

	if (iov.iov_len < n ||
	    msg.msg_flags & MSG_TRUNC) {
		/* Provided buffer is not long enough, enlarge it
		 * and try again. */
		iov.iov_len *= 2;
		iov.iov_base = *buf = realloc(*buf, iov.iov_len);
		goto retry;
	} else if (msg.msg_flags & MSG_CTRUNC) {
		msg.msg_controllen *= 2;
		msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
		goto retry;
	} else if (flags != 0) {
		/* Buffer is big enough, do the actual reading */
		flags = 0;
		goto retry;
	}

	if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
		free(msg.msg_control);
		free(*buf);
		return -NLE_NOADDR;
	}

	for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
		if (cmsg->cmsg_level == SOL_SOCKET &&
		    cmsg->cmsg_type == SCM_CREDENTIALS) {
			*creds = calloc(1, sizeof(struct ucred));
			memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred));
			break;
		}
	}

	free(msg.msg_control);
	return n;

abort:
	free(msg.msg_control);
	free(*buf);
	return 0;
}

#define NL_CB_CALL(cb, type, msg) \
do { \
	err = nl_cb_call(cb, type, msg); \
	switch (err) { \
	case NL_OK: \
		err = 0; \
		break; \
	case NL_SKIP: \
		goto skip; \
	case NL_STOP: \
		goto stop; \
	default: \
		goto out; \
	} \
} while (0)

static int recvmsgs(struct nl_sock *sk, struct nl_cb *cb)
{
	int n, err = 0, multipart = 0;
	unsigned char *buf = NULL;
	struct nlmsghdr *hdr;
	struct sockaddr_nl nla = {0};
	struct nl_msg *msg = NULL;
	struct ucred *creds = NULL;

continue_reading:
	NL_DBG(3, "Attempting to read from %p\n", sk);
	if (cb->cb_recv_ow)
		n = cb->cb_recv_ow(sk, &nla, &buf, &creds);
	else
		n = nl_recv(sk, &nla, &buf, &creds);

	if (n <= 0)
		return n;

	NL_DBG(3, "recvmsgs(%p): Read %d bytes\n", sk, n);

	hdr = (struct nlmsghdr *) buf;
	while (nlmsg_ok(hdr, n)) {
		NL_DBG(3, "recgmsgs(%p): Processing valid message...\n", sk);

		nlmsg_free(msg);
		msg = nlmsg_convert(hdr);
		if (!msg) {
			err = -NLE_NOMEM;
			goto out;
		}

		nlmsg_set_proto(msg, sk->s_proto);
		nlmsg_set_src(msg, &nla);
		if (creds)
			nlmsg_set_creds(msg, creds);

		/* Raw callback is the first, it gives the most control
		 * to the user and he can do his very own parsing. */
		if (cb->cb_set[NL_CB_MSG_IN])
			NL_CB_CALL(cb, NL_CB_MSG_IN, msg);

		/* Sequence number checking. The check may be done by
		 * the user, otherwise a very simple check is applied
		 * enforcing strict ordering */
		if (cb->cb_set[NL_CB_SEQ_CHECK])
			NL_CB_CALL(cb, NL_CB_SEQ_CHECK, msg);
		else if (hdr->nlmsg_seq != sk->s_seq_expect) {
			if (cb->cb_set[NL_CB_INVALID])
				NL_CB_CALL(cb, NL_CB_INVALID, msg);
			else {
				err = -NLE_SEQ_MISMATCH;
				goto out;
			}
		}

		if (hdr->nlmsg_type == NLMSG_DONE ||
		    hdr->nlmsg_type == NLMSG_ERROR ||
		    hdr->nlmsg_type == NLMSG_NOOP ||
		    hdr->nlmsg_type == NLMSG_OVERRUN) {
			/* We can't check for !NLM_F_MULTI since some netlink
			 * users in the kernel are broken. */
			sk->s_seq_expect++;
			NL_DBG(3, "recvmsgs(%p): Increased expected " \
			       "sequence number to %d\n",
			       sk, sk->s_seq_expect);
		}

		if (hdr->nlmsg_flags & NLM_F_MULTI)
			multipart = 1;
	
		/* Other side wishes to see an ack for this message */
		if (hdr->nlmsg_flags & NLM_F_ACK) {
			if (cb->cb_set[NL_CB_SEND_ACK])
				NL_CB_CALL(cb, NL_CB_SEND_ACK, msg);
			else {
				/* FIXME: implement */
			}
		}

		/* messages terminates a multpart message, this is
		 * usually the end of a message and therefore we slip
		 * out of the loop by default. the user may overrule
		 * this action by skipping this packet. */
		if (hdr->nlmsg_type == NLMSG_DONE) {
			multipart = 0;
			if (cb->cb_set[NL_CB_FINISH])
				NL_CB_CALL(cb, NL_CB_FINISH, msg);
		}

		/* Message to be ignored, the default action is to
		 * skip this message if no callback is specified. The
		 * user may overrule this action by returning
		 * NL_PROCEED. */
		else if (hdr->nlmsg_type == NLMSG_NOOP) {
			if (cb->cb_set[NL_CB_SKIPPED])
				NL_CB_CALL(cb, NL_CB_SKIPPED, msg);
			else
				goto skip;
		}

		/* Data got lost, report back to user. The default action is to
		 * quit parsing. The user may overrule this action by retuning
		 * NL_SKIP or NL_PROCEED (dangerous) */
		else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
			if (cb->cb_set[NL_CB_OVERRUN])
				NL_CB_CALL(cb, NL_CB_OVERRUN, msg);
			else {
				err = -NLE_MSG_OVERFLOW;
				goto out;
			}
		}

		/* Message carries a nlmsgerr */
		else if (hdr->nlmsg_type == NLMSG_ERROR) {
			struct nlmsgerr *e = nlmsg_data(hdr);

			if (hdr->nlmsg_len < nlmsg_msg_size(sizeof(*e))) {
				/* Truncated error message, the default action
				 * is to stop parsing. The user may overrule
				 * this action by returning NL_SKIP or
				 * NL_PROCEED (dangerous) */
				if (cb->cb_set[NL_CB_INVALID])
					NL_CB_CALL(cb, NL_CB_INVALID, msg);
				else {
					err = -NLE_MSG_TRUNC;
					goto out;
				}
			} else if (e->error) {
				/* Error message reported back from kernel. */
				if (cb->cb_err) {
					err = cb->cb_err(&nla, e,
							   cb->cb_err_arg);
					if (err < 0)
						goto out;
					else if (err == NL_SKIP)
						goto skip;
					else if (err == NL_STOP) {
						err = -nl_syserr2nlerr(e->error);
						goto out;
					}
				} else {
					err = -nl_syserr2nlerr(e->error);
					goto out;
				}
			} else if (cb->cb_set[NL_CB_ACK])
				NL_CB_CALL(cb, NL_CB_ACK, msg);
		} else {
			/* Valid message (not checking for MULTIPART bit to
			 * get along with broken kernels. NL_SKIP has no
			 * effect on this.  */
			if (cb->cb_set[NL_CB_VALID])
				NL_CB_CALL(cb, NL_CB_VALID, msg);
		}
skip:
		err = 0;
		hdr = nlmsg_next(hdr, &n);
	}
	
	nlmsg_free(msg);
	free(buf);
	free(creds);
	buf = NULL;
	msg = NULL;
	creds = NULL;

	if (multipart) {
		/* Multipart message not yet complete, continue reading */
		goto continue_reading;
	}
stop:
	err = 0;
out:
	nlmsg_free(msg);
	free(buf);
	free(creds);

	return err;
}
Beispiel #7
0
static int
tcpTable_load_netlink(void)
{
    /*  TODO: perhaps use permanent nl handle? */
    struct nl_handle *nl = nl_handle_alloc();
    struct inet_diag_req req = {
        .idiag_family = AF_INET,
        .idiag_states = TCP_ALL,
    };

    struct nl_msg *nm;

    struct sockaddr_nl peer;
    unsigned char *buf = NULL;
    int running = 1, len;

    if (nl == NULL) {
        DEBUGMSGTL(("mibII/tcpTable", "Failed to allocate netlink handle\n"));
        snmp_log(LOG_ERR, "snmpd: Failed to allocate netlink handle\n");
        return -1;
    }

    if (nl_connect(nl, NETLINK_INET_DIAG) < 0) {
        DEBUGMSGTL(("mibII/tcpTable", "Failed to connect to netlink: %s\n", nl_geterror()));
        snmp_log(LOG_ERR, "snmpd: Couldn't connect to netlink: %s\n", nl_geterror());
        nl_handle_destroy(nl);
        return -1;
    }

    nm = nlmsg_alloc_simple(TCPDIAG_GETSOCK, NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST);
    nlmsg_append(nm, &req, sizeof(struct inet_diag_req), 0);

    if (nl_send_auto_complete(nl, nm) < 0) {
        DEBUGMSGTL(("mibII/tcpTable", "nl_send_autocomplete(): %s\n", nl_geterror()));
        snmp_log(LOG_ERR, "snmpd: nl_send_autocomplete(): %s\n", nl_geterror());
        nl_handle_destroy(nl);
        return -1;
    }
    nlmsg_free(nm);

    while (running) {
        struct nlmsghdr *h;
        if ((len = nl_recv(nl, &peer, &buf, NULL)) <= 0) {
            DEBUGMSGTL(("mibII/tcpTable", "nl_recv(): %s\n", nl_geterror()));
            snmp_log(LOG_ERR, "snmpd: nl_recv(): %s\n", nl_geterror());
            nl_handle_destroy(nl);
            return -1;
        }

        h = (struct nlmsghdr*)buf;

        while (nlmsg_ok(h, len)) {
            struct inet_diag_msg *r = nlmsg_data(h);
            struct inpcb    pcb, *nnew;

            if (h->nlmsg_type == NLMSG_DONE) {
                running = 0;
                break;
            }

            r = nlmsg_data(h);

            if (r->idiag_family != AF_INET) {
                h = nlmsg_next(h, &len);
                continue;
            }

            memcpy(&pcb.inp_laddr.s_addr, r->id.idiag_src, r->idiag_family == AF_INET ? 4 : 6);
            memcpy(&pcb.inp_faddr.s_addr, r->id.idiag_dst, r->idiag_family == AF_INET ? 4 : 6);

            pcb.inp_lport = r->id.idiag_sport;
            pcb.inp_fport = r->id.idiag_dport;

            pcb.inp_state = (r->idiag_state & 0xf) < 12 ? linux_states[r->idiag_state & 0xf] : 2;
            if (pcb.inp_state == 5 /* established */ ||
                    pcb.inp_state == 8 /*  closeWait  */ )
                tcp_estab++;
            pcb.uid = r->idiag_uid;

            nnew = SNMP_MALLOC_TYPEDEF(struct inpcb);
            if (nnew == NULL) {
                running = 0;
                /*  XXX report malloc error and return -1? */
                break;
            }
            memcpy(nnew, &pcb, sizeof(struct inpcb));
            nnew->inp_next = tcp_head;
            tcp_head       = nnew;

            h = nlmsg_next(h, &len);
        }
        free(buf);
    }

    nl_handle_destroy(nl);

    if (tcp_head) {
        DEBUGMSGTL(("mibII/tcpTable", "Loaded TCP Table using netlink\n"));
        return 0;
    }
    DEBUGMSGTL(("mibII/tcpTable", "Failed to load TCP Table (netlink)\n"));
    return -1;
}