picnoir/Systemd
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

time-sync-wait: trivial style fixes

Browse source
This commit is contained in:
Zbigniew Jędrzejewski-Szmek 2018-04-07 19:19:49 +02:00
parent 30dd293c88
commit 0fad9daf4b
1 changed files with 11 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

24
src/time-wait-sync/time-wait-sync.c
View file

@ -47,8 +47,7 @@ static void clock_state_release(ClockState *sp) {
sp->fd = safe_close(sp->fd);
}
static int clock_state_update(ClockState *sp,
sd_event *event);
static int clock_state_update(ClockState *sp, sd_event *event);
static int io_handler(sd_event_source * s,
int fd,
@ -73,22 +72,22 @@ static int clock_state_update(ClockState *sp,
clock_state_release(sp);
/* The kernel supports cancelling timers whenever its realtime clock is "set" (which can happen in a variety of
* ways, generally adjustments of at least 500 ms). The way this module works is we set up a timer that will
* wake when it the clock is set, and when that happens we read the clock synchronization state from the return
* ways, generally adjustments of at least 500 ms). The way this module works is we set up a timer that will
* wake when the clock is set, and when that happens we read the clock synchronization state from the return
* value of adjtimex(2), which supports the NTP time adjustment protocol.
*
* The kernel determines whether the clock is synchronized using driver-specific tests, based on time
* information passed by an application, generally through adjtimex(2). If the application asserts the clock
* information passed by an application, generally through adjtimex(2). If the application asserts the clock
* is synchronized, but does not also do something that "sets the clock", the timer will not be cancelled and
* synchronization will not be detected. Should this behavior be observed with a time synchronization provider
* synchronization will not be detected. Should this behavior be observed with a time synchronization provider
* this code might be reworked to do a periodic check as well.
*
* Similarly, this service will never complete if the application sets the time without also providing
* information that adjtimex(2) can use to determine that the clock is synchronized.
*
* Well-behaved implementations including systemd-timesyncd should not produce either situation. For timesyncd
* the initial set of the timestamp uses settimeofday(2), which sets the clock but does not mark it
* synchronized. When an NTP source is selected it sets the clock again with clock_adjtime(2) which does mark
* Well-behaved implementations including systemd-timesyncd should not produce either situation. For timesyncd
* the initial setting of the time uses settimeofday(2), which sets the clock but does not mark it
* synchronized. When an NTP source is selected it sets the clock again with clock_adjtime(2) which does mark
* it synchronized. */
r = timerfd_create(CLOCK_REALTIME, TFD_NONBLOCK | TFD_CLOEXEC);
if (r < 0) {
@ -119,7 +118,7 @@ static int clock_state_update(ClockState *sp,
log_info("adjtime state %d status %x time %s", sp->adjtime_state, tx.status, ts);
if (sp->adjtime_state == TIME_ERROR) {
/* Not synchronized. Do a one-shot wait on the descriptor and inform the caller we need to keep
/* Not synchronized. Do a one-shot wait on the descriptor and inform the caller we need to keep
* running. */
r = sd_event_add_io(event, &sp->event_source, sp->fd,
EPOLLIN, io_handler, sp);
@ -140,8 +139,7 @@ static int clock_state_update(ClockState *sp,
return r;
}
int main(int argc,
char * argv[]) {
int main(int argc, char * argv[]) {
int r;
_cleanup_(sd_event_unrefp) sd_event *event;
ClockState state = {
@ -177,7 +175,7 @@ int main(int argc,
r = clock_state_update(&state, event);
if (r > 0) {
r = sd_event_loop(event);
if (0 > r)
if (r < 0)
log_error_errno(r, "Failed in event loop: %m");
else if (state.adjtime_state == TIME_ERROR) {
log_error("Event loop terminated without synchronizing");