Re: [PATCH 7/8] Scanner changes needed to implement per-containerscanner

[Andrew Morton]

On Wed, 30 May 2007 19:39:41 +0400
Pavel Emelianov <xemul@xxxxxxxxxx> wrote:

> The core change is that the isolate_lru_pages() call is
> replaced with struct scan_controll->isolate_pages() call.
> 
> Other changes include exporting __isolate_lru_page() for
> per-container isolator and handling variable-to-pointer
> changes in try_to_free_pages().
> 
> This makes possible to use different isolation routines
> for per-container page reclamation. This will be used by
> the following patch.
> 
> ...
>
> +struct rss_container;
> +extern unsigned long try_to_free_pages_in_container(struct rss_container *);
> +int __isolate_lru_page(struct page *page, int mode);


>  extern unsigned long shrink_all_memory(unsigned long nr_pages);
>  extern int vm_swappiness;
>  extern int remove_mapping(struct address_space *mapping, struct page *page);
> diff -upr linux-2.6.22-rc2-mm1.orig/mm/vmscan.c linux-2.6.22-rc2-mm1-0/mm/vmscan.c
> --- linux-2.6.22-rc2-mm1.orig/mm/vmscan.c	2007-05-30 12:32:36.000000000 +0400
> +++ linux-2.6.22-rc2-mm1-0/mm/vmscan.c	2007-05-30 16:13:09.000000000 +0400
> @@ -47,6 +47,8 @@
>  
>  #include "internal.h"
>  
> +#include <linux/rss_container.h>
> +
>  struct scan_control {
>  	/* Incremented by the number of inactive pages that were scanned */
>  	unsigned long nr_scanned;
> @@ -70,6 +72,13 @@ struct scan_control {
>  	int all_unreclaimable;
>  
>  	int order;
> +
> +	struct rss_container *cnt;

Can we please have a better name?  "cnt" is usually a (poorly-chosen) name
for an integer counter.  Perhaps "container", or even "rss_container".

> +		nr_reclaimed += shrink_zones(priority, zones, sc);
> +		if (sc->cnt == NULL)
> +			shrink_slab(sc->nr_scanned, gfp_mask, lru_pages);

We don't we shrink slab if called to shrink a container.

This is a fundamental design decision, and a design shortcoming.  A full
discussion of this is absolutely appropriate to the patch changelog. 
Please don't just hide stuff like this in the patch and leave people
wondering, or ignorant.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/