From: Thomas Gleixner
> Sent: 05 February 2021 17:34
>
> On Wed, Jan 20 2021 at 10:51, Yejune Deng wrote:
> > In pps_fill_timex(), use memset and offsetof instead of '= 0'.
> >
> > Signed-off-by: Yejune Deng <[email protected]>
> > ---
> > kernel/time/ntp.c | 13 +++++--------
> > 1 file changed, 5 insertions(+), 8 deletions(-)
> >
> > diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
> > index 87389b9e21ab..3416c0381104 100644
> > --- a/kernel/time/ntp.c
> > +++ b/kernel/time/ntp.c
> > @@ -225,14 +225,11 @@ static inline int is_error_status(int status)
> > static inline void pps_fill_timex(struct __kernel_timex *txc)
> > {
> > /* PPS is not implemented, so these are zero */
> > - txc->ppsfreq = 0;
> > - txc->jitter = 0;
> > - txc->shift = 0;
> > - txc->stabil = 0;
> > - txc->jitcnt = 0;
> > - txc->calcnt = 0;
> > - txc->errcnt = 0;
> > - txc->stbcnt = 0;
> > + int offset, len;
> > +
> > + offset = offsetof(struct __kernel_timex, ppsfreq);
> > + len = offsetof(struct __kernel_timex, tai) - offset;
> > + memset(txc + offset, 0, len);
>
> That zeros bytes at a memory location which is
>
> (offset) * sizeof(struct __kernel_timex)
>
> bytes away from txc. How did this every boot?
>
> And no, even if you fix that pointer math problem then this kind of
> calculation from the middle of a struct is error prone.
It is also, at best, a code size optimisation.
If memset() is actually called (not inlined) then you get a whole
lot of tests against the size and alignment before any writes
of zero happen - which will be the same ones as in the inline code.
It can be worth using memcpy to copy part of a structure
but usually for one with lots of small fields (especially bitfields).
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT,
UK
Registration No: 1397386 (Wales)
