lkml.org 
[lkml]   [2009]   [May]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC][PATCH 5/5] (experimental) chase and free cache only swap
On Wed, 27 May 2009 14:14:42 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Tue, 26 May 2009 12:18:34 +0900
> KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>
> >
> > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> >
> This is a replacement for this. (just an idea, not testd.)
>
> I think this works well. Does anyone has concerns ?
I think so too, except some trivial build errors ;)

I'll test it, but it will take a long time to see the effect of this patch
even if setting the swap space to reasonable size.


Thanks,
Daisuke Nishimura.

> Do I have to modify swap-cluster code to do this in sane way ?
>
> ---
> mm/swapfile.c | 40 ++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 40 insertions(+)
>
> Index: new-trial-swapcount/mm/swapfile.c
> ===================================================================
> --- new-trial-swapcount.orig/mm/swapfile.c
> +++ new-trial-swapcount/mm/swapfile.c
> @@ -74,6 +74,26 @@ static inline unsigned short make_swap_c
> return ret;
> }
>
> +static int try_to_reuse_swap(struct swap_info_struct *si, unsigned long offset)
> +{
> + int type = si - swap_info;
> + swp_entry_t entry = swp_entry(type, offset);
> + struct page *page;
> +
> + page = find_get_page(page);
> + if (!page)
> + return 0;
> + if (!trylock_page(page)) {
> + page_cache_release(page);
> + return 0;
> + }
> + try_to_free_swap(page);
> + unlock_page(page);
> + page_cache_release(page);
> + return 1;
> +}
> +
> +
> /*
> * We need this because the bdev->unplug_fn can sleep and we cannot
> * hold swap_lock while calling the unplug_fn. And swap_lock
> @@ -295,6 +315,18 @@ checks:
> goto no_page;
> if (offset > si->highest_bit)
> scan_base = offset = si->lowest_bit;
> +
> + /* reuse swap entry of cache-only swap if not busy. */
> + if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
> + int ret;
> + spin_unlock(&swap_lock);
> + ret = try_to_reuse_swap(si, offset));
> + spin_lock(&swap_lock);
> + if (ret)
> + goto checks; /* we released swap_lock */
> + goto scan;
> + }
> +
> if (si->swap_map[offset])
> goto scan;
>
> @@ -378,6 +410,10 @@ scan:
> spin_lock(&swap_lock);
> goto checks;
> }
> + if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
> + spin_lock(&swap_lock);
> + goto checks;
> + }
> if (unlikely(--latency_ration < 0)) {
> cond_resched();
> latency_ration = LATENCY_LIMIT;
> @@ -389,6 +425,10 @@ scan:
> spin_lock(&swap_lock);
> goto checks;
> }
> + if (vm_swap_full() && si->swap_map[offset] == SWAP_HAS_CACHE) {
> + spin_lock(&swap_lock);
> + goto checks;
> + }
> if (unlikely(--latency_ration < 0)) {
> cond_resched();
> latency_ration = LATENCY_LIMIT;
>


\
 
 \ /
  Last update: 2009-05-27 08:49    [W:0.073 / U:0.204 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site