[PATCH] xen: Fix selfballooning and ensure it doesn't go too far

[PATCH] xen: Fix selfballooning and ensure it doesn't go too far

[Dan Magenheimer]

[PATCH] xen: Fix selfballooning and ensure it doesn't go too far

The balloon driver's "current_pages" is very different from
totalram_pages.  Self-ballooning needs to be driven by
the latter.  Also, Committed_AS doesn't account for pages
used by the kernel so enforce a floor for when there
are little or no user-space threads using memory.

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>

diff --git a/drivers/xen/xen-selfballoon.c b/drivers/xen/xen-selfballoon.c
index 6ea852e..e3e48f1 100644
--- a/drivers/xen/xen-selfballoon.c
+++ b/drivers/xen/xen-selfballoon.c
@@ -195,14 +195,13 @@ __setup("selfballooning", xen_selfballooning_setup);
  */
 static void selfballoon_process(struct work_struct *work)
 {
-	unsigned long cur_pages, goal_pages, tgt_pages;
+	unsigned long cur_pages, goal_pages, tgt_pages, floor_pages;
 	bool reset_timer = false;
 
 	if (xen_selfballooning_enabled) {
-		cur_pages = balloon_stats.current_pages;
+		cur_pages = totalram_pages;
 		tgt_pages = cur_pages; /* default is no change */
-		goal_pages = percpu_counter_read_positive(&vm_committed_as) +
-			balloon_stats.current_pages - totalram_pages;
+		goal_pages = percpu_counter_read_positive(&vm_committed_as);
 #ifdef CONFIG_FRONTSWAP
 		/* allow space for frontswap pages to be repatriated */
 		if (frontswap_selfshrinking && frontswap_enabled)
@@ -217,7 +216,13 @@ static void selfballoon_process(struct work_struct *work)
 				((goal_pages - cur_pages) /
 				  selfballoon_uphysteresis);
 		/* else if cur_pages == goal_pages, no change */
-		balloon_set_new_target(tgt_pages);
+		floor_pages = totalreserve_pages +
+				(roundup_pow_of_two(max_pfn) >> 5);
+		/* don't balloon too far, lest OOMs occur... */
+		if (tgt_pages < floor_pages)
+			tgt_pages = floor_pages;
+		balloon_set_new_target(tgt_pages +
+			balloon_stats.current_pages - totalram_pages);
 		reset_timer = true;
 	}
 #ifdef CONFIG_FRONTSWAP

Re: [PATCH] xen: Fix selfballooning and ensure it doesn't go too far

[Konrad Rzeszutek Wilk]

On Sat, Sep 24, 2011 at 01:58:08PM -0700, Dan Magenheimer wrote:
> [PATCH] xen: Fix selfballooning and ensure it doesn't go too far
> 
> The balloon driver's "current_pages" is very different from
> totalram_pages.  Self-ballooning needs to be driven by
> the latter.  Also, Committed_AS doesn't account for pages
> used by the kernel so enforce a floor for when there
> are little or no user-space threads using memory.

Hey Dan,
..
> +		floor_pages = totalreserve_pages +
> +				(roundup_pow_of_two(max_pfn) >> 5);

Would it make sense to make the shift be a runtime argument
in case some users report problems with that calculation?

> +		/* don't balloon too far, lest OOMs occur... */
> +		if (tgt_pages < floor_pages)
> +			tgt_pages = floor_pages;
> +		balloon_set_new_target(tgt_pages +
> +			balloon_stats.current_pages - totalram_pages);
>  		reset_timer = true;
>  	}
>  #ifdef CONFIG_FRONTSWAP

Otherwise it looks OK to me. Would you like me to queue it up
for 3.1-rc7?

RE: [PATCH] xen: Fix selfballooning and ensure it doesn't go too far

[Dan Magenheimer]

> From: Konrad Rzeszutek Wilk
> Sent: Monday, September 26, 2011 11:23 AM
> To: Dan Magenheimer
> Cc: linux-kernel@vger.kernel.org; xen-devel@lists.xensource.com; David Vrabel; Jeremy Fitzhardinge
> Subject: Re: [PATCH] xen: Fix selfballooning and ensure it doesn't go too far
> 
> On Sat, Sep 24, 2011 at 01:58:08PM -0700, Dan Magenheimer wrote:
> > [PATCH] xen: Fix selfballooning and ensure it doesn't go too far
> >
> > The balloon driver's "current_pages" is very different from
> > totalram_pages.  Self-ballooning needs to be driven by
> > the latter.  Also, Committed_AS doesn't account for pages
> > used by the kernel so enforce a floor for when there
> > are little or no user-space threads using memory.
> 
> Hey Dan,
> ..
> > +		floor_pages = totalreserve_pages +
> > +				(roundup_pow_of_two(max_pfn) >> 5);
> 
> Would it make sense to make the shift be a runtime argument
> in case some users report problems with that calculation?

Good idea.  I'll take a look at that.

> > +		/* don't balloon too far, lest OOMs occur... */
> > +		if (tgt_pages < floor_pages)
> > +			tgt_pages = floor_pages;
> > +		balloon_set_new_target(tgt_pages +
> > +			balloon_stats.current_pages - totalram_pages);
> >  		reset_timer = true;
> >  	}
> >  #ifdef CONFIG_FRONTSWAP
> 
> Otherwise it looks OK to me. Would you like me to queue it up
> for 3.1-rc7?

Will let you know when I have that change done/tested.

Thanks,
Dan