Message ID | 1416878180-25312-1-git-send-email-pang.xunlei@linaro.org |
---|---|
State | New |
Headers | show |
On Mon, Nov 24, 2014 at 5:16 PM, pang.xunlei <pang.xunlei@linaro.org> wrote: > Ideally, __clocksource_updatefreq_scale, selects the largest shift > value possible for a clocksource. This results in the mult memember > of struct clocksource being particularly large, although not so large > that NTP would adjust the clock to cause it to overflow. > > That said, nothing actually prohibits an overflow from occurring, its > just that it "shouldn't" occur. > > So while very unlikely, and so far never observed, the value of > (cs->mult+cs->maxadj) may have a chance to reach very near 0xFFFFFFFF, > so there is a possibility it may overflow when doing NTP positive > adjustment > > See the following detail: When NTP slewes the clock, kernel goes > through update_wall_time()->...->timekeeping_apply_adjustment(): > tk->tkr.mult += mult_adj; > > Since there is no guard against it, its possible tk->tkr.mult may > overflow during this operation. > > This patch avoids any possible mult overflow by judging the overflow > case before adding mult_adj to mult, also adds the WARNING message > when capturing such case. > > Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> > --- > Fix the problem in the former patch catched by Fengguang's 0day robot: > [time] WARNING: CPU: 0 PID: 1 at kernel/time/timekeeping.c:1337 update_wall_time() > > kernel/time/timekeeping.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c > index 0aef92a..9a1482e 100644 > --- a/kernel/time/timekeeping.c > +++ b/kernel/time/timekeeping.c > @@ -1390,6 +1390,12 @@ static __always_inline void timekeeping_apply_adjustment(struct timekeeper *tk, > * > * XXX - TODO: Doc ntp_error calculation. > */ > + if (mult_adj > 0 && (tk->tkr.mult + mult_adj < mult_adj)) { > + /* NTP adjustment caused clocksource mult overflow */ > + WARN_ON_ONCE(1); > + return; > + } > + > tk->tkr.mult += mult_adj; > tk->xtime_interval += interval; > tk->tkr.xtime_nsec -= offset; So the original problematic patch is already queued in -tip, so next time its best to just provide the additional fix on-top of the problematic one (instead of sending a entirely new patch). I've got a fix implementing the same, which went through testing today, and I'll send it to Thomas in a moment. thanks -john -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 0aef92a..9a1482e 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -1390,6 +1390,12 @@ static __always_inline void timekeeping_apply_adjustment(struct timekeeper *tk, * * XXX - TODO: Doc ntp_error calculation. */ + if (mult_adj > 0 && (tk->tkr.mult + mult_adj < mult_adj)) { + /* NTP adjustment caused clocksource mult overflow */ + WARN_ON_ONCE(1); + return; + } + tk->tkr.mult += mult_adj; tk->xtime_interval += interval; tk->tkr.xtime_nsec -= offset;
Ideally, __clocksource_updatefreq_scale, selects the largest shift value possible for a clocksource. This results in the mult memember of struct clocksource being particularly large, although not so large that NTP would adjust the clock to cause it to overflow. That said, nothing actually prohibits an overflow from occurring, its just that it "shouldn't" occur. So while very unlikely, and so far never observed, the value of (cs->mult+cs->maxadj) may have a chance to reach very near 0xFFFFFFFF, so there is a possibility it may overflow when doing NTP positive adjustment See the following detail: When NTP slewes the clock, kernel goes through update_wall_time()->...->timekeeping_apply_adjustment(): tk->tkr.mult += mult_adj; Since there is no guard against it, its possible tk->tkr.mult may overflow during this operation. This patch avoids any possible mult overflow by judging the overflow case before adding mult_adj to mult, also adds the WARNING message when capturing such case. Signed-off-by: pang.xunlei <pang.xunlei@linaro.org> --- Fix the problem in the former patch catched by Fengguang's 0day robot: [time] WARNING: CPU: 0 PID: 1 at kernel/time/timekeeping.c:1337 update_wall_time() kernel/time/timekeeping.c | 6 ++++++ 1 file changed, 6 insertions(+)