/**************************************************************************** * netutils/ntpclient/ntpclient.c * * Copyright (C) 2014 Gregory Nutt. All rights reserved. * Author: Gregory Nutt * * 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. Neither the name NuttX nor the names of its contributors may be * used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * ****************************************************************************/ /**************************************************************************** * Included Files ****************************************************************************/ #include #include #include #include #include #include #include #include #include #include #include #include "ntpv3.h" /**************************************************************************** * Pre-processor Definitions ****************************************************************************/ /* NTP Time is seconds since 1900. Convert to Unix time which is seconds * since 1970 */ #define NTP2UNIX_TRANLSLATION 2208988800u #define NTP_VERSION 3 /**************************************************************************** * Private Types ****************************************************************************/ /* This enumeration describes the state of the NTP daemon */ enum ntpc_daemon_e { NTP_NOT_RUNNING = 0, NTP_STARTED, NTP_RUNNING, NTP_STOP_REQUESTED, NTP_STOPPED }; /* This type describes the state of the NTP client daemon. Only once * instance of the NTP daemon is permitted in this implementation. */ struct ntpc_daemon_s { volatile uint8_t state; /* See enum ntpc_daemon_e */ sem_t interlock; /* Used to synchronize start and stop events */ pid_t pid; /* Task ID of the NTP daemon */ }; /**************************************************************************** * Private Data ****************************************************************************/ /* This type describes the state of the NTP client daemon. Only once * instance of the NTP daemon is permitted in this implementation. This * limitation is due only to this global data structure. */ static struct ntpc_daemon_s g_ntpc_daemon; /**************************************************************************** * Private Functions ****************************************************************************/ /**************************************************************************** * Name: ntpc_getuint32 * * Description: * Return the big-endian, 4-byte value in network (big-endian) order. * ****************************************************************************/ static inline uint32_t ntpc_getuint32(FAR uint8_t *ptr) { /* Network order is big-endian; host order is irrelevant */ return (uint32_t)ptr[3] | /* MS byte appears first in data stream */ ((uint32_t)ptr[2] << 8) | ((uint32_t)ptr[1] << 16) | ((uint32_t)ptr[0] << 24); } /**************************************************************************** * Name: ntpc_settime * * Description: * Given the NTP time in seconds, set the system time * ****************************************************************************/ static void ntpc_settime(FAR uint8_t *timestamp) { struct timespec tp; time_t seconds; uint32_t frac; uint32_t nsec; #ifdef CONFIG_HAVE_LONG_LONG uint64_t tmp; #else uint32_t a16; uint32_t b0; uint32_t t32; uint32_t t16; uint32_t t0; #endif /* NTP timestamps are represented as a 64-bit fixed-point number, in * seconds relative to 0000 UT on 1 January 1900. The integer part is * in the first 32 bits and the fraction part in the last 32 bits, as * shown in the following diagram. * * 0 1 2 3 * 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ * | Integer Part | * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ * | Fraction Part | * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ */ seconds = ntpc_getuint32(timestamp); /* Translate seconds to account for the difference in the origin time */ if (seconds > NTP2UNIX_TRANLSLATION) { seconds -= NTP2UNIX_TRANLSLATION; } /* Conversion of the fractional part to nanoseconds: * * NSec = (f * 1,000,000,000) / 4,294,967,296 * = (f * (5**9 * 2**9) / (2**32) * = (f * 5**9) / (2**23) * = (f * 1,953,125) / 8,388,608 */ frac = ntpc_getuint32(timestamp + 4); #ifdef CONFIG_HAVE_LONG_LONG /* if we have 64-bit long long values, then the computation is easy */ tmp = ((uint64_t)frac * 1953125) >> 23; nsec = (uint32_t)tmp; #else /* If we don't have 64 bit integer types, then the calculation is a little * more complex: * * Let f = a << 16 + b * 1,953,125 = 0x1d << 16 + 0xcd65 * NSec << 23 = ((a << 16) + b) * ((0x1d << 16) + 0xcd65) * = (a << 16) * 0x1d << 16) + * (a << 16) * 0xcd65 + * b * 0x1d << 16) + * b * 0xcd65; */ /* Break the fractional part up into two values */ a16 = frac >> 16; b0 = frac & 0xffff; /* Get the b32 and b0 terms * * t32 = (a << 16) * 0x1d << 16) * t0 = b * 0xcd65 */ t32 = 0x001d * a16; t0 = 0xcd65 * b0; /* Get the first b16 term * * (a << 16) * 0xcd65 */ t16 = 0xcd65 * a16; /* Add the upper 16-bits to the b32 accumulator */ t32 += (t16 >> 16); /* Add the lower 16-bits to the b0 accumulator, handling carry to the b32 * accumulator */ t16 <<= 16; if (t0 > (0xffffffff - t16)) { t32++; } t0 += t16; /* Get the second b16 term * * b * (0x1d << 16) */ t16 = 0x001d * b0; /* Add the upper 16-bits to the b32 accumulator */ t32 += (t16 >> 16); /* Add the lower 16-bits to the b0 accumulator, handling carry to the b32 * accumulator */ t16 <<= 16; if (t0 > (0xffffffff - t16)) { t32++; } t0 += t16; /* t32 and t0 represent the 64 bit product. Now shift right by 23 bits to * accomplish the divide by by 2**23. */ nsec = (t32 << (32 - 23)) + (t0 >> 23); #endif /* Set the system time */ tp.tv_sec = seconds; tp.tv_nsec = nsec; clock_settime(CLOCK_REALTIME, &tp); svdbg("Set time to %lu seconds: %d\n", (unsigned long)tp.tv_sec, ret); } /**************************************************************************** * Name: ntpc_daemon * * Description: * This the the NTP client daemon. This is a *very* minimal * implementation! An NTP request is and the system clock is set when the * response is received * ****************************************************************************/ static int ntpc_daemon(int argc, char **argv) { struct sockaddr_in server; struct ntp_datagram_s xmit; struct ntp_datagram_s recv; struct timeval tv; socklen_t socklen; ssize_t nbytes; int exitcode = EXIT_SUCCESS; int ret; int sd; /* Indicate that we have started */ g_ntpc_daemon.state = NTP_RUNNING; sem_post(&g_ntpc_daemon.interlock); /* Create a datagram socket */ sd = socket(AF_INET, SOCK_DGRAM, 0); if (sd < 0) { ndbg("ERROR: socket failed: %d\n", errno); g_ntpc_daemon.state = NTP_STOPPED; sem_post(&g_ntpc_daemon.interlock); return EXIT_FAILURE; } /* Setup a receive timeout on the socket */ tv.tv_sec = 5; tv.tv_usec = 0; ret = setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(struct timeval)); if (ret < 0) { ndbg("ERROR: setsockopt failed: %d\n", errno); g_ntpc_daemon.state = NTP_STOPPED; sem_post(&g_ntpc_daemon.interlock); return EXIT_FAILURE; } /* Setup or sockaddr_in struct with information about the server we are * going to ask the the time from. */ memset(&server, 0, sizeof(struct sockaddr_in)); server.sin_family = AF_INET; server.sin_port = htons(CONFIG_NETUTILS_NTPCLIENT_PORTNO); server.sin_addr.s_addr = htonl(CONFIG_NETUTILS_NTPCLIENT_SERVERIP); /* Here we do the communication with the NTP server. This is a very simple * client architecture. A request is sent and then a NTP packet is received * and used to set the current time. * * NOTE that the scheduler is locked whenever this loop runs. That * assures both: (1) that there are no asynchronous stop requests and * (2) that we are not suspended while in critical moments when we about * to set the new time. This sounds harsh, but this function is suspended * most of the time either: (1) sending a datagram, (2) receiving a datagram, * or (3) waiting for the next poll cycle. * * TODO: The first datagram that is sent is usually lost. That is because * the MAC address of the NTP server is not in the ARP table. This is * particularly bad here because the request will not be sent again until * the long delay expires leaving the system with bad time for a long time * initially. Solutions: * * 1. Fix send logic so that it assures that the ARP request has been * sent and the entry is in the ARP table before sending the packet * (best). * 2. Add some ad hoc logic here so that there is no delay until at least * one good time is received. */ sched_lock(); while (g_ntpc_daemon.state != NTP_STOP_REQUESTED) { /* Format the transmit datagram */ memset(&xmit, 0, sizeof(xmit)); xmit.lvm = MKLVM(0, 3, NTP_VERSION); svdbg("Sending a NTP packet\n"); ret = sendto(sd, &xmit, sizeof(struct ntp_datagram_s), 0, (FAR struct sockaddr *)&server, sizeof(struct sockaddr_in)); if (ret < 0) { /* Check if we received a signal. That is not an error but * other error events will terminate the client. */ int errval = errno; if (errval != EINTR) { ndbg("ERROR: sendto() failed: %d\n", errval); exitcode = EXIT_FAILURE; break; } /* Go back to the top of the loop if we were interrupted * by a signal. The signal might mean that we were * requested to stop(?) */ continue; } /* Attempt to receive a packet (with a timeout that was set up via * setsockopt() above) */ socklen = sizeof(struct sockaddr_in); nbytes = recvfrom(sd, (void *)&recv, sizeof(struct ntp_datagram_s), 0, (FAR struct sockaddr *)&server, &socklen); /* Check if the received message was long enough to be a valid NTP * datagram. */ if (nbytes >= (ssize_t)NTP_DATAGRAM_MINSIZE) { svdbg("Setting time\n"); ntpc_settime(recv.recvtimestamp); } /* Check for errors. Note that properly received, short datagrams * are simply ignored. */ else if (nbytes < 0) { /* Check if we received a signal. That is not an error but * other error events will terminate the client. */ int errval = errno; if (errval != EINTR) { ndbg("ERROR: recvfrom() failed: %d\n", errval); exitcode = EXIT_FAILURE; break; } } /* A full implementation of an NTP client would require much more. I * think we can skip most of that here. */ if (g_ntpc_daemon.state == NTP_RUNNING) { svdbg("Waiting for %d seconds\n", CONFIG_NETUTILS_NTPCLIENT_POLLDELAYSEC); (void)sleep(CONFIG_NETUTILS_NTPCLIENT_POLLDELAYSEC); } } /* The NTP client is terminating */ sched_unlock(); g_ntpc_daemon.state = NTP_STOPPED; sem_post(&g_ntpc_daemon.interlock); return exitcode; } /**************************************************************************** * Public Functions ****************************************************************************/ /**************************************************************************** * Name: ntpc_start * * Description: * Start the NTP daemon * * Returned Value: * On success, the non-negative task ID of the NTPC daemon is returned; * On failure, a negated errno value is returned. * ****************************************************************************/ int ntpc_start(void) { /* Is the NTP in a non-running state? */ sched_lock(); if (g_ntpc_daemon.state == NTP_NOT_RUNNING || g_ntpc_daemon.state == NTP_STOPPED) { /* Is this the first time that the NTP daemon has been started? */ if (g_ntpc_daemon.state == NTP_NOT_RUNNING) { /* Yes... then we will need to initialize the state structure */ sem_init(&g_ntpc_daemon.interlock, 0, 0); } /* Start the NTP daemon */ g_ntpc_daemon.state = NTP_STARTED; g_ntpc_daemon.pid = task_create("NTP daemon", CONFIG_NETUTILS_NTPCLIENT_SERVERPRIO, CONFIG_NETUTILS_NTPCLIENT_STACKSIZE, ntpc_daemon, NULL); /* Handle failures to start the NTP daemon */ if (g_ntpc_daemon.pid < 0) { int errval = errno; DEBUGASSERT(errval > 0); g_ntpc_daemon.state = NTP_STOPPED; ndbg("ERROR: Failed to start the NTP daemon\n", errval); return -errval; } /* Wait for any daemon state change */ do { (void)sem_wait(&g_ntpc_daemon.interlock); } while (g_ntpc_daemon.state == NTP_STARTED); } sched_unlock(); return g_ntpc_daemon.pid; } /**************************************************************************** * Name: ntpc_stop * * Description: * Stop the NTP daemon * * Returned Value: * Zero on success; a negated errno value on failure. The current * implementation only returns success. * ****************************************************************************/ #ifndef CONFIG_DISABLE_SIGNALS int ntpc_stop(void) { int ret; /* Is the NTP in a running state? */ sched_lock(); if (g_ntpc_daemon.state == NTP_STARTED || g_ntpc_daemon.state == NTP_RUNNING) { /* Yes.. request that the daemon stop. */ g_ntpc_daemon.state = NTP_STOP_REQUESTED; /* Wait for any daemon state change */ do { /* Signal the NTP client */ ret = kill(g_ntpc_daemon.pid, CONFIG_NETUTILS_NTPCLIENT_SIGWAKEUP); if (ret < 0) { ndbg("ERROR: kill pid %d failed: %d\n", g_ntpc_daemon.pid, errno); break; } /* Wait for the NTP client to respond to the stop request */ (void)sem_wait(&g_ntpc_daemon.interlock); } while (g_ntpc_daemon.state == NTP_STOP_REQUESTED); } sched_unlock(); return OK; } #endif