[PATCH] mm/vmscan.c:shrink_list(): check PageSwapCache() after add_to_swap()

[PATCH] mm/vmscan.c:shrink_list(): check PageSwapCache() after add_to_swap()

[Nikita Danilov]

Hello,

shrink_list() checks PageSwapCache() before calling add_to_swap(), this
means that anonymous page that is going to be added to the swap right
now these checks return false and:

 (*) it will be unaccounted for in nr_mapped, and

 (*) it won't be written to the swap if gfp_flags include __GFP_IO but
     not __GFP_FS.

(Both will happen only on the next round of scanning.)

Patch below just moves may_enter_fs initialization down. I am not sure
about (*nr_mapped) increase though.

Comments?

Nikita.
diff -puN -b mm/vmscan.c~check-PageSwapCache-after-add-to-swap mm/vmscan.c
--- i386/mm/vmscan.c~check-PageSwapCache-after-add-to-swap	Wed Jan 28 19:22:14 2004
+++ i386-nikita/mm/vmscan.c	Wed Jan 28 19:36:16 2004
@@ -380,8 +380,6 @@ shrink_list(struct list_head *page_list,
 			(*nr_mapped)++;
 
 		BUG_ON(PageActive(page));
-		may_enter_fs = (gfp_mask & __GFP_FS) ||
-				(PageSwapCache(page) && (gfp_mask & __GFP_IO));
 
 		if (PageWriteback(page))
 			goto keep_locked;
@@ -412,6 +410,9 @@ shrink_list(struct list_head *page_list,
 		}
 #endif /* CONFIG_SWAP */
 
+		may_enter_fs = (gfp_mask & __GFP_FS) ||
+				(PageSwapCache(page) && (gfp_mask & __GFP_IO));
+
 		/*
 		 * The page is mapped into the page tables of one or more
 		 * processes. Try to unmap it here.

_
--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [PATCH] mm/vmscan.c:shrink_list(): check PageSwapCache() after add_to_swap()

[Andrew Morton]

Nikita Danilov <Nikita@Namesys.COM> wrote:
>
> Hello,
> 
> shrink_list() checks PageSwapCache() before calling add_to_swap(), this
> means that anonymous page that is going to be added to the swap right
> now these checks return false and:
> 
>  (*) it will be unaccounted for in nr_mapped, and
> 
>  (*) it won't be written to the swap if gfp_flags include __GFP_IO but
>      not __GFP_FS.
> 
> (Both will happen only on the next round of scanning.)

OK.  Does it make a measurable change in any benchmarks?

> Patch below just moves may_enter_fs initialization down. I am not sure
> about (*nr_mapped) increase though.

nr_mapped seems OK.
--
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:"aart@kvack.org"> aart@kvack.org </a>

[BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Nick Piggin]

Andrew Morton wrote:

>Nikita Danilov <Nikita@Namesys.COM> wrote:
>
>>Hello,
>>
>>shrink_list() checks PageSwapCache() before calling add_to_swap(), this
>>means that anonymous page that is going to be added to the swap right
>>now these checks return false and:
>>
>> (*) it will be unaccounted for in nr_mapped, and
>>
>> (*) it won't be written to the swap if gfp_flags include __GFP_IO but
>>     not __GFP_FS.
>>
>>(Both will happen only on the next round of scanning.)
>>
>
>OK.  Does it make a measurable change in any benchmarks?
>
>

Small big significantly better on kbuild when tested on top of the other
two patches (dont-rotate-active-list and my mapped-fair).

With this patch as well, we are now as good or better than 2.4 on
medium and heavy swapping kbuilds and much better than stock 2.6
with light swapping loads (not as good as 2.4 but close).

http://www.kerneltrap.org/~npiggin/vm/3/



--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Rik van Riel]

On Fri, 30 Jan 2004, Nick Piggin wrote:

> Small big significantly better on kbuild when tested on top of the other
> two patches (dont-rotate-active-list and my mapped-fair).

Where can I grab those ?

> With this patch as well, we are now as good or better than 2.4 on
> medium and heavy swapping kbuilds and much better than stock 2.6
> with light swapping loads (not as good as 2.4 but close).
> 
> http://www.kerneltrap.org/~npiggin/vm/3/

Neat!  Does it have any side effects to interactive
desktop behaviour ?

-- 
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." - Brian W. Kernighan

--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Nick Piggin]

Rik van Riel wrote:

>On Fri, 30 Jan 2004, Nick Piggin wrote:
>
>
>>Small big significantly better on kbuild when tested on top of the other
>>two patches (dont-rotate-active-list and my mapped-fair).
>>
>
>Where can I grab those ?
>
>

I've put the all here for now.
They are against the latest mm with the RSS patch backed out.

http://kerneltrap.org/~npiggin/vm/

>>With this patch as well, we are now as good or better than 2.4 on
>>medium and heavy swapping kbuilds and much better than stock 2.6
>>with light swapping loads (not as good as 2.4 but close).
>>
>>http://www.kerneltrap.org/~npiggin/vm/3/
>>
>
>Neat!  Does it have any side effects to interactive
>desktop behaviour ?
>
>

To be honest I haven't tried them on my desktop. I would guess
they should be fine.

--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Andrew Morton]

Nick Piggin <piggin@cyberone.com.au> wrote:
>
> They are against the latest mm with the RSS patch backed out.

Is the rss patch in mm2 still misbehaving?
--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Nick Piggin]

Andrew Morton wrote:

>Nick Piggin <piggin@cyberone.com.au> wrote:
>
>>They are against the latest mm with the RSS patch backed out.
>>
>
>Is the rss patch in mm2 still misbehaving?
>
>

I haven't had a look yet Andrew, I'll give it a try soon.
Sorry, I actually meant 2.6.2-rc2-mm1.

--
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:"aart@kvack.org"> aart@kvack.org </a>

Re: [BENCHMARKS] 2.6 kbuild results (with add_to_swap patch)

[Nick Piggin]

Rik van Riel wrote:

>On Fri, 30 Jan 2004, Nick Piggin wrote:
>
>
>>Small big significantly better on kbuild when tested on top of the other
>>two patches (dont-rotate-active-list and my mapped-fair).
>>
>
>Where can I grab those ?
>
>
>>With this patch as well, we are now as good or better than 2.4 on
>>medium and heavy swapping kbuilds and much better than stock 2.6
>>with light swapping loads (not as good as 2.4 but close).
>>
>>http://www.kerneltrap.org/~npiggin/vm/3/
>>
>
>Neat!  Does it have any side effects to interactive
>desktop behaviour ?
>
>

Hi Rik,
I've just tried the latest patchset on my desktop system. Doing
a make -j4 bzImage, hdparm -t /dev/hda in a loop, mozilla, gnome
xterms etc.

Nothing bad is happening, although I'm only just touching swap.

--
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:"aart@kvack.org"> aart@kvack.org </a>