On Sat, 18 Mar 2017, Deepa Dinamani wrote:
> struct timespec is not y2038 safe. The plan is to
> get rid of all uses of timespec internally in the
> kernel. Replace uses of timespec with timespec64.
> The syscall interfaces will be changed in a separate
> series.
> Call to do_sys_setimeofday() is superfluous as all
> the necessary checks and functions are done by the
> underlying function do_sys_setimeofday64().
> Replace do_sys_setimeofday64() directly instead of
> do_sys_settimeofday(). The do_sys_setimeofday64()
> calls timespec64_valid() internally, which is the
> same as timespec_valid().
Again, this is hard to read and contains too much useless information,
struct timespec is not Y2038 safe on 32 bit machines and needs to be
replaced with struct timespec64.
do_sys_settimeofday() is a wrapper around do_sys_settimeofday64().
Convert all call sites to use do_sys_seetimeofday64() and remove it.
> @@ -109,7 +109,7 @@ COMPAT_SYSCALL_DEFINE2(settimeofday, struct
> compat_timeval __user *, tv,
> struct timezone __user *, tz)
> {
> struct timeval user_tv;
> - struct timespec new_ts;
> + struct timespec64 new_ts;
Please use the ordering I pointed out in the other mail.
Thanks,
tglx
