Re: [PATCH 1/2] frontswap: allow backends to register after frontswap initilization

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Wed, Mar 14, 2012 at 06:13:27PM +0100, Andor Daam wrote:
> This patch allows backends to register to frontswap even after
> swapon was run. Before a backend registers all calls to init are
> recorded and the creation of tmem_pools delayed until a backend
> registers.

Couple of questions..
> 
> Sigend-off-by: Stefan Hengelein <ilendir@googlemail.com>
> Signed-off-by: Florian Schmaus <fschmaus@gmail.com>
> Sigend-off-by: Andor Daam <andor.daam@googlemail.com>
> ---
>  mm/frontswap.c |   72 +++++++++++++++++++++++++++++++++++++++++++++++++-------
>  1 files changed, 63 insertions(+), 9 deletions(-)
> 
> diff --git a/mm/frontswap.c b/mm/frontswap.c
> index 2b80c5a..209487b 100644
> --- a/mm/frontswap.c
> +++ b/mm/frontswap.c
> @@ -49,15 +49,33 @@ static u64 frontswap_failed_puts;
>  static u64 frontswap_invalidates;
>  
>  /*
> + * When no backend is registered all calls to init are registered and
> + * remembered but fail to create tmem_pools. When a backend registers with
> + * frontswap the previous calls to init are executed to create tmem_pools
> + * and set the respective poolids.
> + * While no backend is registered all "puts", "gets" and "flushes" are
> + * ignored or fail.
> + */
> +#define MAX_INITIALIZABLE_SD 32

Why 32? Would 42 work better :-) What does 'SD' stand for?


> +static int sds[MAX_INITIALIZABLE_SD];
> +static int backend_registered;

bool.
> +
> +/*
>   * Register operations for frontswap, returning previous thus allowing
>   * detection of multiple backends and possible nesting
>   */
>  struct frontswap_ops frontswap_register_ops(struct frontswap_ops *ops)
>  {
>  	struct frontswap_ops old = frontswap_ops;
> +	int i;

unsigned int.

>  
>  	frontswap_ops = *ops;
> -	frontswap_enabled = 1;
> +
> +	backend_registered = 1;
> +	for (i = 0; i < MAX_INITIALIZABLE_SD; i++) {
> +		if (sds[i] != -1)

Hm, why not have sds be an array of bool's instead?

> +			(*frontswap_ops.init)(sds[i]);
> +	}
>  	return old;
>  }
>  EXPORT_SYMBOL(frontswap_register_ops);
> @@ -66,12 +84,21 @@ EXPORT_SYMBOL(frontswap_register_ops);
>  void __frontswap_init(unsigned type)
>  {
>  	struct swap_info_struct *sis = swap_info[type];
> +	int i;
>  
>  	BUG_ON(sis == NULL);
>  	if (sis->frontswap_map == NULL)
>  		return;
> -	if (frontswap_enabled)
> -		(*frontswap_ops.init)(type);
> +	if (frontswap_enabled) {
> +		if (backend_registered)
> +			(*frontswap_ops.init)(type);
> +		for (i = 0; i < MAX_INITIALIZABLE_SD; i++) {
> +			if (sds[i] == -1) {
> +				sds[i] = type;

Ok, so you save 'unsigned' in a 'int' array. If the unsigned ends up being
32769, that translates to -1 in int. So we would wind up thinking it is unregisted.

> +				break;
> +			}
> +		}
> +	}
>  }
>  EXPORT_SYMBOL(__frontswap_init);
>  
> @@ -90,6 +117,11 @@ int __frontswap_put_page(struct page *page)
>  	struct swap_info_struct *sis = swap_info[type];
>  	pgoff_t offset = swp_offset(entry);
>  
> +	if (!backend_registered) {
> +		frontswap_failed_puts++;
> +		return ret;
> +	}
> +
>  	BUG_ON(!PageLocked(page));
>  	BUG_ON(sis == NULL);
>  	if (frontswap_test(sis, offset))
> @@ -127,12 +159,16 @@ int __frontswap_get_page(struct page *page)
>  	struct swap_info_struct *sis = swap_info[type];
>  	pgoff_t offset = swp_offset(entry);
>  
> +	if (!backend_registered)
> +		return ret;
> +
>  	BUG_ON(!PageLocked(page));
>  	BUG_ON(sis == NULL);
>  	if (frontswap_test(sis, offset))
>  		ret = (*frontswap_ops.get_page)(type, offset, page);
>  	if (ret == 0)
>  		frontswap_gets++;
> +
>  	return ret;
>  }
>  EXPORT_SYMBOL(__frontswap_get_page);
> @@ -145,6 +181,9 @@ void __frontswap_invalidate_page(unsigned type, pgoff_t offset)
>  {
>  	struct swap_info_struct *sis = swap_info[type];
>  
> +	if (!backend_registered)
> +		return;
> +
>  	BUG_ON(sis == NULL);
>  	if (frontswap_test(sis, offset)) {
>  		(*frontswap_ops.invalidate_page)(type, offset);
> @@ -162,13 +201,23 @@ EXPORT_SYMBOL(__frontswap_invalidate_page);
>  void __frontswap_invalidate_area(unsigned type)
>  {
>  	struct swap_info_struct *sis = swap_info[type];
> +	int i;

unsigned int

>  
> -	BUG_ON(sis == NULL);
> -	if (sis->frontswap_map == NULL)
> -		return;
> -	(*frontswap_ops.invalidate_area)(type);
> -	atomic_set(&sis->frontswap_pages, 0);
> -	memset(sis->frontswap_map, 0, sis->max / sizeof(long));
> +	if (backend_registered) {
> +		BUG_ON(sis == NULL);
> +		if (sis->frontswap_map == NULL)
> +			return;
> +		(*frontswap_ops.invalidate_area)(type);
> +		atomic_set(&sis->frontswap_pages, 0);
> +		memset(sis->frontswap_map, 0, sis->max / sizeof(long));
> +	} else {
> +		for (i = 0; i < MAX_INITIALIZABLE_SD; i++) {
> +			if (sds[i] == type) {
> +				sds[i] = -1;
> +				break;
> +			}
> +		}
> +	}
>  }
>  EXPORT_SYMBOL(__frontswap_invalidate_area);
>  
> @@ -255,6 +304,7 @@ EXPORT_SYMBOL(frontswap_curr_pages);
>  static int __init init_frontswap(void)
>  {
>  	int err = 0;
> +	int i;
>  
>  #ifdef CONFIG_DEBUG_FS
>  	struct dentry *root = debugfs_create_dir("frontswap", NULL);
> @@ -267,6 +317,10 @@ static int __init init_frontswap(void)
>  	debugfs_create_u64("invalidates", S_IRUGO,
>  				root, &frontswap_invalidates);
>  #endif
> +	for (i = 0; i < MAX_INITIALIZABLE_SD; i++)
> +		sds[i] = -1;
> +
> +	frontswap_enabled = 1;
>  	return err;
>  }
>  
> -- 
> 1.7.5.4

--
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>