On Sun, Dec 06, 2015 at 10:07:01AM +0100, Thomas Gleixner wrote: > On Sat, 5 Dec 2015, Sasha Levin wrote: > > On 12/05/2015 12:10 PM, Thomas Gleixner wrote: > > > That's not a canonical timeval. timeval_valid() is what you want to > > > check it. Or has adjtimex some magic exception here?
You can't use timeval_valid for ADJ_SETOFFSET, because the seconds field can also be negative (when setting the time back). > > Nope, it looks like timeval_valid() is indeed what I've needed to use. > > > > Is there a reason ntp_validate_timex() doesn't do timeval_valid() too > > for at least the ADJ_SETOFFSET case? If not, I'll add it in. So you should not use timeval_valid, and the original patch is also not right. The rule is: The value of a timeval is the sum of its fields, but the field tv_usec must always be non-negative. We had a discussion about this a year or two ago. Maybe I can find it again. The overflow is a latent problem, and the patch should: 1. return error in case (txc->time.tv_usec >= USEC_PER_SEC) 2. remove the redundant test in timekeeping_inject_offset. Thanks, Richard -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/