Re: [RFC][PATCH] swap: send callback when swap slot is freed

[Ingo Molnar <mingo@elte.hu>]

* Nitin Gupta <ngupta@vflare.org> wrote:

> (Re-sending since I didn't get any reply last time)
> 
> Currently, we have "swap discard" mechanism which sends a discard 
> bio request when we find a free cluster during scan_swap_map(). 
> This callback can come a long time after swap slots are actually 
> freed.
> 
> This delay in callback is a great problem when (compressed) RAM 
> [1] is used as a swap device. So, this change adds a callback 
> which is called as soon as a swap slot becomes free. For above 
> mentioned case of swapping over compressed RAM device, this is 
> very useful since we can immediately free memory allocated for 
> this swap page.
> 
> This callback does not replace swap discard support. It is called 
> with swap_lock held, so it is meant to trigger action that 
> finishes quickly. However, swap discard is an I/O request and can 
> be used for taking longer actions.
> 
> This patch is simply a quick hack but it works (tested on x86 and 
> x64). Can anyone please suggest a cleaner and proper way to do 
> this? Any feedback would be really helpful. Thanks.
> 
> 
> Links:
> [1] http://code.google.com/p/compcache/
> 
> Signed-off-by: Nitin Gupta <ngupta@vflare.org>
> ---
> 
>  include/linux/swap.h |   14 ++++++++++++++
>  init/Kconfig         |    9 +++++++++
>  mm/swapfile.c        |   20 ++++++++++++++++++++
>  3 files changed, 43 insertions(+), 0 deletions(-)
> 
> diff --git a/include/linux/swap.h b/include/linux/swap.h
> index 7c15334..483d0b9 100644
> --- a/include/linux/swap.h
> +++ b/include/linux/swap.h
> @@ -8,6 +8,7 @@
>  #include <linux/memcontrol.h>
>  #include <linux/sched.h>
>  #include <linux/node.h>
> +#include <linux/blkdev.h>
>  
>  #include <asm/atomic.h>
>  #include <asm/page.h>
> @@ -20,6 +21,8 @@ struct bio;
>  #define SWAP_FLAG_PRIO_MASK	0x7fff
>  #define SWAP_FLAG_PRIO_SHIFT	0
>  
> +typedef void (swap_free_notify_fn) (struct block_device *, unsigned long);
> +
>  static inline int current_is_kswapd(void)
>  {
>  	return current->flags & PF_KSWAPD;
> @@ -155,6 +158,9 @@ struct swap_info_struct {
>  	unsigned int max;
>  	unsigned int inuse_pages;
>  	unsigned int old_block_size;
> +#ifdef CONFIG_SWAP_FREE_NOTIFY
> +	swap_free_notify_fn *swap_free_notify_fn;
> +#endif
>  };
>  
>  struct swap_list_t {
> @@ -297,6 +303,14 @@ extern int reuse_swap_page(struct page *);
>  extern int try_to_free_swap(struct page *);
>  struct backing_dev_info;
>  
> +#ifdef CONFIG_SWAP_FREE_NOTIFY
> +extern void set_swap_free_notify(unsigned, swap_free_notify_fn *);
> +#else
> +void set_swap_free_notify(unsigned, swap_free_notify_fn *)
> +{
> +}
> +#endif
> +
>  /* linux/mm/thrash.c */
>  extern struct mm_struct *swap_token_mm;
>  extern void grab_swap_token(struct mm_struct *);
> diff --git a/init/Kconfig b/init/Kconfig
> index 3f7e609..b0112c6 100644
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -176,6 +176,15 @@ config SWAP
>  	  used to provide more virtual memory than the actual RAM present
>  	  in your computer.  If unsure say Y.
>  
> +config SWAP_FREE_NOTIFY
> +	bool "Enable swap free notification"
> +	depends on SWAP
> +	default y
> +	help
> +	  This causes swap subsystem to send a callback to swap device when
> +	  a swap block becomes free. This is useful in cases such as when
> +	  compressed RAM device is used as swap disk.
> +
>  config SYSVIPC
>  	bool "System V IPC"
>  	---help---
> diff --git a/mm/swapfile.c b/mm/swapfile.c
> index 8ffdc0d..25bacfd 100644
> --- a/mm/swapfile.c
> +++ b/mm/swapfile.c
> @@ -552,6 +552,22 @@ out:
>  	return NULL;
>  }
>  
> +#ifdef CONFIG_SWAP_FREE_NOTIFY
> +/*
> + * Sets callback for event when swap_map[offset] == 0
> + * i.e. page at this swap offset is not longer used.
> + */
> +void set_swap_free_notify(unsigned type, swap_free_notify_fn *notify_fn)
> +{
> +	struct swap_info_struct *sis;
> +	sis = get_swap_info_struct(type);
> +	BUG_ON(!sis);
> +	sis->swap_free_notify_fn = notify_fn;
> +	return;
> +}
> +EXPORT_SYMBOL(set_swap_free_notify);
> +#endif
> +
>  static int swap_entry_free(struct swap_info_struct *p,
>  			   swp_entry_t ent, int cache)
>  {
> @@ -583,6 +599,10 @@ static int swap_entry_free(struct swap_info_struct *p,
>  			swap_list.next = p - swap_info;
>  		nr_swap_pages++;
>  		p->inuse_pages--;
> +#ifdef CONFIG_SWAP_FREE_NOTIFY
> +		if (p->swap_free_notify_fn)
> +			p->swap_free_notify_fn(p->bdev, offset);
> +#endif
>  	}
>  	if (!swap_count(count))
>  		mem_cgroup_uncharge_swap(ent);

Shouldnt this be all be unconditional? Sounds like a reasonable VM 
callback. The patch would become a lot smaller as well.

	Ingo