Re: [PATCH v2 4/4] mm/zswap: use GFP_NOIO instead of GFP_KERNEL

[Bob Liu]

On 09/06/2013 01:16 PM, Weijie Yang wrote:
> To avoid zswap store and reclaim functions called recursively,
> use GFP_NOIO instead of GFP_KERNEL
> 

The reason of using GFP_KERNEL in write back path is we want to try our
best to move those pages from zswap to real swap device.

I think it would be better to keep GFP_KERNEL flag but find some other
ways to skip zswap/zswap_frontswap_store() if zswap write back is in
progress.

What I can think of currently is adding a mutex to zswap, take that
mutex when zswap write back happens and check the mutex in
zswap_frontswap_store().


> Signed-off-by: Weijie Yang <weijie.yang@xxxxxxxxxxx>
> ---
>  mm/zswap.c |    6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/mm/zswap.c b/mm/zswap.c
> index cc40e6a..3d05ed8 100644
> --- a/mm/zswap.c
> +++ b/mm/zswap.c
> @@ -427,7 +427,7 @@ static int zswap_get_swap_cache_page(swp_entry_t entry,
>  		 * Get a new page to read into from swap.
>  		 */
>  		if (!new_page) {
> -			new_page = alloc_page(GFP_KERNEL);
> +			new_page = alloc_page(GFP_NOIO);
>  			if (!new_page)
>  				break; /* Out of memory */
>  		}
> @@ -435,7 +435,7 @@ static int zswap_get_swap_cache_page(swp_entry_t entry,
>  		/*
>  		 * call radix_tree_preload() while we can wait.
>  		 */
> -		err = radix_tree_preload(GFP_KERNEL);
> +		err = radix_tree_preload(GFP_NOIO);
>  		if (err)
>  			break;
>  
> @@ -636,7 +636,7 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
>  	}
>  
>  	/* allocate entry */
> -	entry = zswap_entry_cache_alloc(GFP_KERNEL);
> +	entry = zswap_entry_cache_alloc(GFP_NOIO);
>  	if (!entry) {
>  		zswap_reject_kmemcache_fail++;
>  		ret = -ENOMEM;
> 

-- 
Regards,
-Bob
--
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/