Re: [RFC PATCH 2/2] zsmalloc: chain-length configuration should consider other metrics

[Sergey Senozhatsky]

On (26/01/12 20:56), Yosry Ahmed wrote:
> > > Looks alright to me too.
> > > 
> > > > > +		size_t sizes[2];
> > > > > +
> > > > > +		/* this object spans two pages */
> > > > > +		sizes[0] = PAGE_SIZE - off;
> > > > > +		sizes[1] = mem_len - sizes[0];
> > > > > +
> > > > > +		sg_init_table(sg, 2);
> > > > > +		sg_set_page(sg, zpdesc_page(zpdesc), sizes[0], off);
> > > > > +
> > > > > +		zpdesc = get_next_zpdesc(zpdesc);
> > > > > +		sg = sg_next(sg);
> > > > 
> > > > Is this stateful? Will the SG list be returned pointing at the second
> > > > page now?
> > > 
> > > It makes no difference because we just called sg_init_table(sg, 2),
> > > so sg_next(sg) is equivalent to &sg[1].
> > 
> > I did it this way for (sort of) consistency: sg next follows zpdesc next.
> 
> Makes sense, I think I confused myself earlier. Do you plan to switch
> the existing interfaces to use SG lists for both zswap and zram? Or does
> zram still need the old interfaces?

That's a good question.  I don't know.  I was thinking of providing
SG-list based API for zswap first and keeping zram as is, for now.
zram doesn't use crypto API at the moment so supporting new interface
will come with extra work.