Re: [swsusp] Rework image freeing

[Andrew Morton <akpm@osdl.org>]

Pavel Machek <pavel@ucw.cz> wrote:
>
> Hi!
> 
> > >
> > > +	for_each_zone(zone) {
> > >  +		for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn) {
> > >  +			struct page * page;
> > >  +			page = pfn_to_page(zone_pfn + zone->zone_start_pfn);
> > >  +			if (PageNosave(page) && PageNosaveFree(page)) {
> > >  +				ClearPageNosave(page);
> > >  +				ClearPageNosaveFree(page);
> > >  +				free_page((long) page_address(page));
> > >  +			}
> > 
> > There doesn't seem to be much point in converting the pageframe address to
> > a virtual address, then back to a pageframe address.  Why not just do
> > put_page() here?
> 
> I do not know what put_page does, and free_page() has some friendly
> BUG_ONs. But if I should just do put_page(page); instead of
> free_page((long) page_address(page)), that is okay with me.

I don't know what alloc_pages() and free_page[s]() are supposed to do either
really.  They date from when Linus was in diapers.  They talk in terms of
kernel virtual addresses and will explode loudly if used for highmem pages.

All modern code will use page*'s or pfn's.

> [Sound of harddrive seeking and CPU fan going up as test kernel
> compiles].
> 

Ah, I wondered what it was.