Re: [PATCH 4/9] move x86's simple heap management to common code

[Christoffer Dall <christoffer.dall@linaro.org>]

On Thu, Jan 02, 2014 at 04:17:08PM +0100, Andrew Jones wrote:
> On Sat, Dec 28, 2013 at 10:30:24PM -0800, Christoffer Dall wrote:
> > > +
> > > +	if (s != (s & ~(pagesize - 1))) {
> > 
> > you're just testing 'if (s & (pagesize -1))' right?
> 
> yeah, I'll simplify that.
> 
> > 
> > > +		s += pagesize;
> > > +		s &= ~(pagesize - 1);
> > > +		p = (void *)s;
> > > +	}
> > 
> > a one-line comment on this block saying 'page-align start of heap would
> > be nice.
> 
> added
> 
> > 
> > > +
> > > +	while (size >= pagesize) {
> > > +		*(void **)p = free_head;
> > > +		free_head = p;
> > > +		p += pagesize;
> > > +		size -= pagesize;
> > > +	}
> > 
> > you could also be nice and comment this block of code, saying something
> > like "set up linked list of free pages using the pages themselves as the
> > data structure" if you should feel so inclined.
> 
> added "link free pages"
> 
> > 
> > why are you not trusting start to be page aligned but you are trusting
> > size to be?  If size is smaller than pagesize then this loop will go
> > nuts won't it?
> 
> I don't see how. As soon as size is less than pagesize we won't
> [re]enter the loop, and thus it can never go negative (big positive).
> 
yeah you're right. duh.