Re: [patch/rfc 0/2] vmemmap for s390

[Heiko Carstens <heiko.carstens@de.ibm.com>]

On Mon, Dec 04, 2006 at 10:47:14AM +0900, KAMEZAWA Hiroyuki wrote:
> On Fri, 1 Dec 2006 15:05:42 +0100
> Heiko Carstens <heiko.carstens@de.ibm.com> wrote:
> 
> > This is the s390 implementation (both 31 and 64 bit) of a virtual memmap.
> > ia64 was used as a blueprint of course. I hope I incorporated everything
> > I read lately on linux-mm wrt. vmemmap.
> > So I post this as an RFC, since I most probably have forgotten something,
> > or did something wrong. Comments highly appreciated.
> >
> > This patchset is against linux-2.6.19-rc6-mm2.
> >
> > Patch 1 is sort of unrelated to the vmemmap patch but still needed, so
> > that the patch applies.
> > Patch 2 is the vmemmap implementation.
> >
> 
> - Could you divide Patch 2 into a few pieces ?
>   * setup, vmemmap pagetable creation, shared memory codes , etc...

Yes, will do.

> - Do you need vmemmap for 32 bits ? (just a question)

Yes, because of two reasons:

- this way we can use the same code for both 31 and 64 bit, which simplifies
  maintenance.

- if we convert to Mel Gorman's 'add_active_range' interface then we still
  need a way to add/initialize struct pages. Mel's new interface limits the
  size of the created mem_map to the largest valid pfn passed via
  'add_active_range'. Hence there is no way to add additional struct pages
  to the end of the initial mem_map, even if 'mem=...' was given via the
  kernel command line.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>