On Thu, Oct 1, 2015 at 6:26 AM, Patrik Flykt <patrik.fl...@linux.intel.com> wrote:
> > Hi, > > On Wed, 2015-09-30 at 22:08 -0700, Naveen Singh wrote: > > From: Naveen Singh <nav...@nestlabs.com> > > > > Current NTP code is written with an assumption that timeserver is > > always an IPv4 address. If there is an IPv6 timeserver then the socket > > operation would fail with error as Permission denied (13). This change in > > ntp.c ensures that code works fine with both IPv4 and IPv6 address. > > --- > > src/ntp.c | 129 > ++++++++++++++++++++++++++++++++++++++++++++++++-------------- > > 1 file changed, 101 insertions(+), 28 deletions(-) > > This was patch version 2, so the next one is version 3. Add the version > to the Subject: line as [PATCH v3], this can be done with > 'git ... --subject-prefix="PATCH v3" ...'. > > > diff --git a/src/ntp.c b/src/ntp.c > > index 2c313a4..a55365d 100644 > > --- a/src/ntp.c > > +++ b/src/ntp.c > > @@ -18,7 +18,6 @@ > > * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA > 02110-1301 USA > > * > > */ > > - > > #ifdef HAVE_CONFIG_H > > #include <config.h> > > #endif > > @@ -34,6 +33,7 @@ > > #include <netinet/in.h> > > #include <netinet/ip.h> > > #include <arpa/inet.h> > > +#include <netdb.h> > > > > #include <glib.h> > > > > @@ -117,12 +117,12 @@ static struct timespec mtx_time; > > static int transmit_fd = 0; > > > > static char *timeserver = NULL; > > -static struct sockaddr_in timeserver_addr; > > +static struct sockaddr_storage timeserver_addr; > > static gint poll_id = 0; > > static gint timeout_id = 0; > > static guint retries = 0; > > > > -static void send_packet(int fd, const char *server, uint32_t timeout); > > +static void send_packet(int fd, const char *server, uint32_t timeout, > int family); > > > > static void next_server(void) > > { > > @@ -143,17 +143,21 @@ static gboolean send_timeout(gpointer user_data) > > if (retries++ == NTP_SEND_RETRIES) > > next_server(); > > else > > - send_packet(transmit_fd, timeserver, timeout << 1); > > + send_packet(transmit_fd, timeserver, timeout << 1, > timeserver_addr.ss_family); > > > > return FALSE; > > } > > > > -static void send_packet(int fd, const char *server, uint32_t timeout) > > +static void send_packet(int fd, const char *server, uint32_t timeout, > int family) > > Instead of supplying a char *server, why don't we give a struct > sockaddr* instead? What's the purpose of carrying the server > "name" (nah, the IP address as a string), if it is already resolved in > the calling function? BTW, it is available as timeserver_addr, so do we > really need any of this information here? > > > { > > struct ntp_msg msg; > > - struct sockaddr_in addr; > > + struct sockaddr_in in4addr; > > + struct sockaddr_in6 in6addr; > > + struct sockaddr * addr; > > struct timeval transmit_timeval; > > ssize_t len; > > + int size; > > + unsigned char * addrptr; > > > > /* > > * At some point, we could specify the actual system precision > with: > > @@ -168,10 +172,29 @@ static void send_packet(int fd, const char > *server, uint32_t timeout) > > msg.poll = 10; // max > > msg.precision = NTP_PRECISION_S; > > > > - memset(&addr, 0, sizeof(addr)); > > - addr.sin_family = AF_INET; > > - addr.sin_port = htons(123); > > - addr.sin_addr.s_addr = inet_addr(server); > > + if (family == AF_INET) { > > + memset(&in4addr, 0, sizeof(in4addr)); > > + in4addr.sin_family = AF_INET; > > + in4addr.sin_port = htons(123); > > + size = sizeof(in4addr); > > + addrptr = (unsigned char *)&in4addr.sin_addr.s_addr; > > + addr = (struct sockaddr *)&in4addr; > > + } else if (family == AF_INET6){ > > + memset(&in6addr, 0, sizeof(in6addr)); > > + in6addr.sin6_family = AF_INET6; > > + in6addr.sin6_port = htons(123); > > + size = sizeof(in6addr); > > + addrptr = in6addr.sin6_addr.__in6_u.__u6_addr8; > > + addr = (struct sockaddr *)&in6addr; > > + } else { > > + DBG("wrong family type"); > > + return; > > + } > > + if (inet_pton(family, server, addrptr) == 0) > > + { > > + DBG("cannot convert ip address string"); > > + return; > > + } > > The above is unnecessary. You already have timeserver_addr set or the > function gets a sockaddr * to use. > > > gettimeofday(&transmit_timeval, NULL); > > clock_gettime(CLOCK_MONOTONIC, &mtx_time); > > @@ -180,7 +203,7 @@ static void send_packet(int fd, const char *server, > uint32_t timeout) > > msg.xmttime.fraction = htonl(transmit_timeval.tv_usec * 1000); > > > > len = sendto(fd, &msg, sizeof(msg), MSG_DONTWAIT, > > - &addr, sizeof(addr)); > > + addr, size); > > Use the stored timeserver_addr or supplied sockaddr * family information > to set the size to either sizeof(struct sockaddr_in) or sizeof(struct > sockaddr_in6). > > > if (len < 0) { > > connman_error("Time request for server %s failed (%d/%s)", > > server, errno, strerror(errno)); > > @@ -213,7 +236,7 @@ static gboolean next_poll(gpointer user_data) > > if (!timeserver || transmit_fd == 0) > > return FALSE; > > > > - send_packet(transmit_fd, timeserver, NTP_SEND_TIMEOUT); > > + send_packet(transmit_fd, timeserver, NTP_SEND_TIMEOUT, > timeserver_addr.ss_family); > > > > return FALSE; > > } > > @@ -363,7 +386,7 @@ static gboolean received_data(GIOChannel *channel, > GIOCondition condition, > > gpointer user_data) > > { > > unsigned char buf[128]; > > - struct sockaddr_in sender_addr; > > + struct sockaddr_storage sender_addr; > > struct msghdr msg; > > struct iovec iov; > > struct cmsghdr *cmsg; > > @@ -396,9 +419,18 @@ static gboolean received_data(GIOChannel *channel, > GIOCondition condition, > > if (len < 0) > > return TRUE; > > > > - if (timeserver_addr.sin_addr.s_addr != sender_addr.sin_addr.s_addr) > > - /* only accept messages from the timeserver */ > > + if (sender_addr.ss_family == AF_INET) { > > + if (((struct sockaddr_in > *)×erver_addr)->sin_addr.s_addr != ((struct sockaddr_in > *)&sender_addr)->sin_addr.s_addr) > > + return TRUE; > > + } else if(sender_addr.ss_family == AF_INET6) { > > + if (memcmp(((struct sockaddr_in6 > *)×erver_addr)->sin6_addr.__in6_u.__u6_addr8, > > + ((struct sockaddr_in6 > *)&sender_addr)->sin6_addr.__in6_u.__u6_addr8, > > + 16) != 0) > > Use memcmp for both of these. First check address families, if they > match the size can be deduced (for example as above) and have only a > single memcmp to do the matching. > > > + return TRUE; > > + } else { > > + connman_error("Not a valid family type"); > > return TRUE; > > + } > > > > tv = NULL; > > clock_gettime(CLOCK_MONOTONIC, &mrx_time); > > @@ -422,36 +454,78 @@ static gboolean received_data(GIOChannel *channel, > GIOCondition condition, > > static void start_ntp(char *server) > > { > > GIOChannel *channel; > > - struct sockaddr_in addr; > > + struct sockaddr * addr; > > + struct sockaddr_in in4addr; > > + struct sockaddr_in6 in6addr; > > + struct addrinfo hint; > > + struct addrinfo *info; > > + int family; > > int tos = IPTOS_LOWDELAY, timestamp = 1; > > + int size; > > + int ret; > > > > if (!server) > > return; > > > > - DBG("server %s", server); > > + memset(&hint, 0, sizeof(hint)); > > + hint.ai_family = AF_UNSPEC; > > + hint.ai_flags = AI_NUMERICHOST; > > + ret = getaddrinfo(server, NULL, &hint, &info); > > + > > + if (ret) { > > + connman_error("cannot get server info"); > > + return; > > + } > > + > > + family = info->ai_family; > > + freeaddrinfo(info); > > Before throwing away info, the address is already available as > ->sin6_addr or ->sin_addr. See man getaddrinfo. Store the address and > the address family in timeserver_addr. > > > + > > + DBG("server %s family %d", server, family); > > > > if (channel_watch > 0) > > goto send; > > > > - transmit_fd = socket(PF_INET, SOCK_DGRAM | SOCK_CLOEXEC, 0); > > - if (transmit_fd < 0) { > > + transmit_fd = socket(family, SOCK_DGRAM | SOCK_CLOEXEC, 0); > > + if (transmit_fd <= 0) { > > connman_error("Failed to open time server socket"); > > return; > > } > > > > - memset(&addr, 0, sizeof(addr)); > > - addr.sin_family = AF_INET; > > + if (family == AF_INET) { > > + memset(&in4addr, 0, sizeof(in4addr)); > > + in4addr.sin_family = family; > > + addr = (struct sockaddr *)&in4addr; > > + size = sizeof(in4addr); > > + if (inet_pton(family, server, &(((struct sockaddr_in > *)×erver_addr)->sin_addr.s_addr)) == 0) { > > + DBG("cannot convert ipv4 address string"); > > + } > > + } else if (family == AF_INET6) { > > + memset(&in6addr, 0, sizeof(in6addr)); > > + in6addr.sin6_family = family; > > + addr = (struct sockaddr *)&in6addr; > > + size = sizeof(in6addr); > > + if (inet_pton(family, server, ((struct sockaddr_in6 > *)×erver_addr)->sin6_addr.__in6_u.__u6_addr8) == 0) { > > + DBG("cannot convert ipv6 address string"); > > + } > > + } else { > > + connman_error("Family not correct"); > > + return; > > + } > > No need for the above, look into the 'info' supplied to getaddrinfo() > above. > > > - if (bind(transmit_fd, (struct sockaddr *) &addr, sizeof(addr)) < > 0) { > > + timeserver_addr.ss_family = family; > > + > > + if (bind(transmit_fd, (struct sockaddr *) addr, size) < 0) { > > connman_error("Failed to bind time server socket"); > > close(transmit_fd); > > return; > > } > > > > - if (setsockopt(transmit_fd, IPPROTO_IP, IP_TOS, &tos, sizeof(tos)) > < 0) { > > - connman_error("Failed to set type of service option"); > > - close(transmit_fd); > > - return; > > + if (family == AF_INET) { > > + if (setsockopt(transmit_fd, IPPROTO_IP, IP_TOS, &tos, > sizeof(tos)) < 0) { > > + connman_error("Failed to set type of service > option"); > > + close(transmit_fd); > > + return; > > + } > > } > > > > if (setsockopt(transmit_fd, SOL_SOCKET, SO_TIMESTAMP, ×tamp, > > @@ -479,7 +553,7 @@ static void start_ntp(char *server) > > g_io_channel_unref(channel); > > > > send: > > - send_packet(transmit_fd, server, NTP_SEND_TIMEOUT); > > + send_packet(transmit_fd, server, NTP_SEND_TIMEOUT, family); > > } > > > > int __connman_ntp_start(char *server) > > @@ -493,7 +567,6 @@ int __connman_ntp_start(char *server) > > g_free(timeserver); > > > > timeserver = g_strdup(server); > > - timeserver_addr.sin_addr.s_addr = inet_addr(server); > > > > start_ntp(timeserver); > > > > Patrik > > I really did not want to diverge too much from the original code (in terms > of changing the function signatures and logic) and that is why i did not > want to pass the address directly. These all are valid sugestions and I > will send you the version 3 patch after doing the testing. > _______________________________________________ > connman mailing list > connman@connman.net > https://lists.connman.net/mailman/listinfo/connman > _______________________________________________ connman mailing list connman@connman.net https://lists.connman.net/mailman/listinfo/connman