Linus Torvalds <> wrote:

> > David Howells (7):
> >       Handle leap seconds in mktime64()
> This one is completely wrong.
> Leap seconds are inserted *at* the minute, not at the secodn before the 
> minute.
> So this code:
> +       /* Handle leap seconds */
> +       if (sec == 60)
> +               sec = 59;
> is just complete crap. Making the whole commit bogus and wrong.

I did ask on ksummit-discuss beforehand.  The advice was to treat hh:mm:60 as
hh:mm:59 rather than hh:mm+1:00.  Unless we actually support leap seconds as
distinct time_t values, it has to be one or the other.

> The code did the right thing wrt leap seconds before, without having
> any magical and incorrect special case. That commit makes it instead
> have two seconds of xx:xx:59.

... as opposed to two seconds of xx:xx+1:00.  You can argue it either way -
and arguably both are equally wrong since neither maps correctly to reality.

> The fact that people add extra code to make things extra wrong is
> annoying. The patch is marked as being cc'd to John Stultz, but I
> assume it was never acked, because I doubt he would ack something like
> this.
> To make things worse, this whole series seems to have existed for less
> than one day, and then it was sent to me as a pull request, however
> buggy and non-acked it was.

I only asked James to pass the CVE-labelled commit on to you and didn't
include it in a patch series.  The rest I posted hoping for reviews.

> To make things EVEN *more* broken, this crap was marked for stable.

It will theoretically need to end up there anyway, since it is technically
possible for the bugs to prevent a kernel from booting - just not very likely.
To unsubscribe from this list: send the line "unsubscribe 
linux-security-module" in
the body of a message to
More majordomo info at

Reply via email to