RE: Subject: [PATCH V6 1/4] mm: frontswap: swap data structure changes

[Dan Magenheimer <dan.magenheimer@oracle.com>]

> From: Jan Beulich [mailto:JBeulich@novell.com]
> Sent: Tuesday, August 09, 2011 10:18 AM
> To: Dan Magenheimer
> Cc: hannes@cmpxchg.org; jackdachef@gmail.com; hughd@google.com; jeremy@goop.org; npiggin@kernel.dk;
> linux-mm@kvack.org; akpm@linux-foundation.org; sjenning@linux.vnet.ibm.com; Chris Mason; Konrad Wilk;
> Kurt Hackel; riel@redhat.com; ngupta@vflare.org; linux-kernel@vger.kernel.org; matthew@wil.cx
> Subject: RE: Subject: [PATCH V6 1/4] mm: frontswap: swap data structure changes
> 
> >>> On 09.08.11 at 17:03, Dan Magenheimer <dan.magenheimer@oracle.com> wrote:
> >> > --- linux/include/linux/swap.h	2011-08-08 08:19:25.880690134 -0600
> >> > +++ frontswap/include/linux/swap.h	2011-08-08 08:59:03.952691415 -0600
> >> > @@ -194,6 +194,8 @@ struct swap_info_struct {
> >> >  	struct block_device *bdev;	/* swap device or bdev of swap file */
> >> >  	struct file *swap_file;		/* seldom referenced */
> >> >  	unsigned int old_block_size;	/* seldom referenced */
> >>
> >> #ifdef CONFIG_FRONTSWAP
> >>
> >> > +	unsigned long *frontswap_map;	/* frontswap in-use, one bit per page */
> >> > +	unsigned int frontswap_pages;	/* frontswap pages in-use counter */
> >>
> >> #endif
> >>
> >> (to eliminate any overhead with that config option unset)
> >>
> >> Jan
> >
> > Hi Jan --
> >
> > Thanks for the review!
> >
> > As noted in the commit comment, if these structure elements are
> > not put inside an #ifdef CONFIG_FRONTSWAP, it becomes
> > unnecessary to clutter the core swap code with several ifdefs.
> > The cost is one pointer and one unsigned int per allocated
> > swap device (often no more than one swap device per system),
> > so the code clarity seemed more important than the tiny
> > additional runtime space cost.
> >
> > Do you disagree?
> 
> Not necessarily - I just know that in other similar occasions (partly
> internally to our company) I was asked to make sure turned off
> features would not leave *any* run time foot print whatsoever.
> 
> Jan

Hi Jan --

With two extra static inlines in frontswap.h (frontswap_map_get()
and frontswap_map_set(), I've managed to both avoid the extra swap struct
members for frontswap_map and frontswap_pages when CONFIG_FRONTSWAP is
disabled AND avoid the #ifdef CONFIG_FRONTSWAP clutter in swapfile.h.

I'll post a V7 soon... let me know what you think!

Thanks,
Dan

--
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/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>