From 838188511cb2085d8980956b270cc68ed9911102 Mon Sep 17 00:00:00 2001 From: Kaarle Ritvanen Date: Thu, 4 Aug 2016 12:26:56 +0200 Subject: [PATCH 11/12] ntpd: postpone hostname resolution if fails on startup Signed-off-by: Kaarle Ritvanen --- networking/ntpd.c | 130 +++++++++++++++++++++++++++--------------------------- 1 file changed, 64 insertions(+), 66 deletions(-) diff --git a/networking/ntpd.c b/networking/ntpd.c index b7fa5dce9..c6d07adac 100644 --- a/networking/ntpd.c +++ b/networking/ntpd.c @@ -155,6 +155,7 @@ #define RETRY_INTERVAL 32 /* on send/recv error, retry in N secs (need to be power of 2) */ #define NOREPLY_INTERVAL 512 /* sent, but got no reply: cap next query by this many seconds */ #define RESPONSE_INTERVAL 16 /* wait for reply up to N secs */ +#define HOSTNAME_INTERVAL 4 /* hostname lookup failed. Wait N secs for next try */ /* Step threshold (sec). std ntpd uses 0.128. */ @@ -293,6 +294,7 @@ typedef struct { typedef struct { len_and_sockaddr *p_lsa; + char *p_hostname; char *p_dotted; int p_fd; int datapoint_idx; @@ -318,7 +320,6 @@ typedef struct { datapoint_t filter_datapoint[NUM_DATAPOINTS]; /* last sent packet: */ msg_t p_xmt_msg; - char p_hostname[1]; } peer_t; @@ -791,27 +792,17 @@ reset_peer_stats(peer_t *p, double offset) } static void -resolve_peer_hostname(peer_t *p, int loop_on_fail) -{ - len_and_sockaddr *lsa; - - again: - lsa = host2sockaddr(p->p_hostname, 123); - if (!lsa) { - /* error message already emitted by host2sockaddr() */ - if (!loop_on_fail) - return; -//FIXME: do this to avoid infinite looping on typo in a hostname? -//well... in which case, what is a good value for loop_on_fail? - //if (--loop_on_fail == 0) - // xfunc_die(); - sleep(5); - goto again; +resolve_peer_hostname(peer_t *p) { + len_and_sockaddr *lsa = host2sockaddr(p->p_hostname, 123); + if (lsa) { + if (p->p_lsa) { + free(p->p_lsa); + free(p->p_dotted); + } + p->p_lsa = lsa; + p->p_dotted = xmalloc_sockaddr2dotted_noport(&lsa->u.sa); } - free(p->p_lsa); - free(p->p_dotted); - p->p_lsa = lsa; - p->p_dotted = xmalloc_sockaddr2dotted_noport(&lsa->u.sa); + set_next(p, lsa ? 0 : HOSTNAME_INTERVAL); } static void @@ -820,28 +811,29 @@ add_peers(const char *s) llist_t *item; peer_t *p; - p = xzalloc(sizeof(*p) + strlen(s)); - strcpy(p->p_hostname, s); - resolve_peer_hostname(p, /*loop_on_fail=*/ 1); + p = xzalloc(sizeof(*p)); + p->p_hostname = xstrdup(s); + resolve_peer_hostname(p); /* Names like N..pool.ntp.org are randomly resolved * to a pool of machines. Sometimes different N's resolve to the same IP. * It is not useful to have two peers with same IP. We skip duplicates. */ - for (item = G.ntp_peers; item != NULL; item = item->link) { - peer_t *pp = (peer_t *) item->data; - if (strcmp(p->p_dotted, pp->p_dotted) == 0) { - bb_error_msg("duplicate peer %s (%s)", s, p->p_dotted); - free(p->p_lsa); - free(p->p_dotted); - free(p); - return; + if (p->p_lsa) + for (item = G.ntp_peers; item != NULL; item = item->link) { + peer_t *pp = (peer_t *) item->data; + if (pp->p_lsa && strcmp(p->p_dotted, pp->p_dotted) == 0) { + bb_error_msg("duplicate peer %s (%s)", s, p->p_dotted); + free(p->p_hostname); + free(p->p_lsa); + free(p->p_dotted); + free(p); + return; + } } - } p->p_fd = -1; p->p_xmt_msg.m_status = MODE_CLIENT | (NTP_VERSION << 3); - p->next_action_time = G.cur_time; /* = set_next(p, 0); */ reset_peer_stats(p, STEP_THRESHOLD); llist_add_to(&G.ntp_peers, p); @@ -2378,44 +2370,50 @@ int ntpd_main(int argc UNUSED_PARAM, char **argv) for (item = G.ntp_peers; item != NULL; item = item->link) { peer_t *p = (peer_t *) item->data; - if (p->next_action_time <= G.cur_time) { - if (p->p_fd == -1) { - /* Time to send new req */ - if (--cnt == 0) { - VERB4 bb_error_msg("disabling burst mode"); - G.polladj_count = 0; - G.poll_exp = MINPOLL; + if (p->p_lsa) { + + if (p->next_action_time <= G.cur_time) { + if (p->p_fd == -1) { + /* Time to send new req */ + if (--cnt == 0) { + VERB4 bb_error_msg("disabling burst mode"); + G.polladj_count = 0; + G.poll_exp = MINPOLL; + } + send_query_to_peer(p); + } else { + /* Timed out waiting for reply */ + close(p->p_fd); + p->p_fd = -1; + /* If poll interval is small, increase it */ + if (G.poll_exp < BIGPOLL) + adjust_poll(MINPOLL); + timeout = poll_interval(NOREPLY_INTERVAL); + bb_error_msg("timed out waiting for %s, reach 0x%02x, next query in %us", + p->p_dotted, p->reachable_bits, timeout); + + /* What if don't see it because it changed its IP? */ + if (p->reachable_bits == 0) + resolve_peer_hostname(p); + + set_next(p, timeout); } - send_query_to_peer(p); - } else { - /* Timed out waiting for reply */ - close(p->p_fd); - p->p_fd = -1; - /* If poll interval is small, increase it */ - if (G.poll_exp < BIGPOLL) - adjust_poll(MINPOLL); - timeout = poll_interval(NOREPLY_INTERVAL); - bb_error_msg("timed out waiting for %s, reach 0x%02x, next query in %us", - p->p_dotted, p->reachable_bits, timeout); - - /* What if don't see it because it changed its IP? */ - if (p->reachable_bits == 0) - resolve_peer_hostname(p, /*loop_on_fail=*/ 0); - - set_next(p, timeout); } - } + + if (p->p_fd >= 0) { + /* Wait for reply from this peer */ + pfd[i].fd = p->p_fd; + pfd[i].events = POLLIN; + idx2peer[i] = p; + i++; + } + + } else + resolve_peer_hostname(p); if (p->next_action_time < nextaction) nextaction = p->next_action_time; - if (p->p_fd >= 0) { - /* Wait for reply from this peer */ - pfd[i].fd = p->p_fd; - pfd[i].events = POLLIN; - idx2peer[i] = p; - i++; - } } timeout = nextaction - G.cur_time; -- 2.11.0