Re: memfill

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2017-02-08 at 10:04 -0800, Matthew Wilcox wrote:
> On Tue, Feb 07, 2017 at 11:07:34AM -0800, James Bottomley wrote:
> > > /** 
> > >  * memfill - Fill a region of memory with the given value
> > >  * @s: Pointer to the start of the region.
> > >  * @v: The word to fill the region with.
> > >  * @n: The size of the region.
> > >  * 
> > >  * Differs from memset() in that it fills with an unsigned long
> > > instead of 
> > >  * a byte.  The pointer and the size must be aligned to unsigned
> > > long.
> > >  */
> > > void memfill(unsigned long *s, unsigned long v, size_t n)
> > 
> > If we're going to do this, are you sure we wouldn't be wanting a
> > string
> > fill instead of a memfill (because filling either by byte or long
> > looks
> > a bit restrictive) assuming static strings that we can tell the
> > compile
> > time size of, it would be easy for generic code to optimise.
> 
> When you say 'string fill', do you mean this?
> 
> void memfill(void *dst, size_t dsz, void *src, size_t ssz)
> {
> 	if (ssz == 1) {
> 		memset(dst, *(unsigned char *)src, dsz);
> 	} else if (ssz == sizeof(short)) {
> 		memset_s(dst, *(unsigned short *)src, dsz);
> 	} else if (ssz == sizeof(int)) {
> 		memset_i(dst, *(unsigned int *)src, dsz);
> 	} else if (ssz == sizeof(long)) {
> 		memset_l(dst, *(unsigned long *)src, dsz);
> 	} else {
> 		while (dsz >= ssz) {
> 			memcpy(dst, src, ssz);
> 			dsz -= ssz;
> 			dst += ssz;
> 		}
> 		if (dsz)
> 			memcpy(dst. src. dsz);
> 	}
> }
> 
> (with memset_[sil] being obvious, I hope).  Possibly some variation 
> on this to optimise compile-time constant dsz.
> 
> I have no objection to that.  Indeed, it would match Lars Wirzenius'
> memfill() definition (if not implementation) which makes me quite 
> happy.

Yes, that's about it.  My only qualm looking at the proposal was if
memfill is genuinely useful to something why would it only want to fill
in units of sizeof(long).  On the other hand, we've been operating for
decades without it, so perhaps memset_l is the only use case?

James