linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Mel Gorman <mel@csn.ul.ie>
To: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Andrea Arcangeli <aarcange@redhat.com>,
	Minchan Kim <minchan.kim@gmail.com>,
	Linux-MM <linux-mm@kvack.org>,
	KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [PATCH 5/5] extend KSM refcounts to the anon_vma root
Date: Wed, 12 May 2010 22:07:07 +0100	[thread overview]
Message-ID: <20100512210706.GQ24989@csn.ul.ie> (raw)
In-Reply-To: <20100512134111.467fb6c2@annuminas.surriel.com>

On Wed, May 12, 2010 at 01:41:11PM -0400, Rik van Riel wrote:
> Subject: extend KSM refcounts to the anon_vma root
> 
> KSM reference counts can cause an anon_vma to exist after the processe
> it belongs to have already exited.  Because the anon_vma lock now lives
> in the root anon_vma, we need to ensure that the root anon_vma stays
> around until after all the "child" anon_vmas have been freed.
> 
> The obvious way to do this is to have a "child" anon_vma take a
> reference to the root in anon_vma_fork.  When the anon_vma is freed
> at munmap or process exit, we drop the refcount in anon_vma_unlink
> and possibly free the root anon_vma.
> 
> The KSM anon_vma reference count function also needs to be modified
> to deal with the possibility of freeing 2 levels of anon_vma.  The
> easiest way to do this is to break out the KSM magic and make it
> generic.
> 
> When compiling without CONFIG_KSM, this code is compiled out.
> 
> Signed-off-by: Rik van Riel <riel@redhat.com>
> ---
>  include/linux/rmap.h |   12 ++++++++++++
>  mm/ksm.c             |   17 ++++++-----------
>  mm/rmap.c            |   45 ++++++++++++++++++++++++++++++++++++++++++++-
>  3 files changed, 62 insertions(+), 12 deletions(-)
> 
> diff --git a/include/linux/rmap.h b/include/linux/rmap.h
> index 33ffe14..387d40c 100644
> --- a/include/linux/rmap.h
> +++ b/include/linux/rmap.h
> @@ -126,6 +126,18 @@ int anon_vma_fork(struct vm_area_struct *, struct vm_area_struct *);
>  void __anon_vma_link(struct vm_area_struct *);
>  void anon_vma_free(struct anon_vma *);
>  
> +#ifdef CONFIG_KSM
> +static inline void get_anon_vma(struct anon_vma *anon_vma)
> +{
> +	atomic_inc(&anon_vma->ksm_refcount);
> +}
> +
> +void drop_anon_vma(struct anon_vma *);
> +#else
> +#define get_anon_vma(x)		do {} while(0)
> +#define drop_anon_vma(x)	do {} while(0)
> +#endif
> +
>  static inline void anon_vma_merge(struct vm_area_struct *vma,
>  				  struct vm_area_struct *next)
>  {
> diff --git a/mm/ksm.c b/mm/ksm.c
> index 7ca0dd7..9f2acc9 100644
> --- a/mm/ksm.c
> +++ b/mm/ksm.c
> @@ -318,19 +318,14 @@ static void hold_anon_vma(struct rmap_item *rmap_item,
>  			  struct anon_vma *anon_vma)
>  {
>  	rmap_item->anon_vma = anon_vma;
> -	atomic_inc(&anon_vma->ksm_refcount);
> +	get_anon_vma(anon_vma);
>  }

I'm not quite getting this. Here, we get the local anon_vma so we
increment its reference count and later we drop it but without a
refcount taken on the root anon_vma, why is it guaranteed to stay
around?

>  
> -static void drop_anon_vma(struct rmap_item *rmap_item)
> +static void ksm_drop_anon_vma(struct rmap_item *rmap_item)
>  {
>  	struct anon_vma *anon_vma = rmap_item->anon_vma;
>  
> -	if (atomic_dec_and_lock(&anon_vma->ksm_refcount, &anon_vma->root->lock)) {
> -		int empty = list_empty(&anon_vma->head);
> -		anon_vma_unlock(anon_vma);
> -		if (empty)
> -			anon_vma_free(anon_vma);
> -	}
> +	drop_anon_vma(anon_vma);
>  }
>  
>  /*
> @@ -415,7 +410,7 @@ static void break_cow(struct rmap_item *rmap_item)
>  	 * It is not an accident that whenever we want to break COW
>  	 * to undo, we also need to drop a reference to the anon_vma.
>  	 */
> -	drop_anon_vma(rmap_item);
> +	ksm_drop_anon_vma(rmap_item);
>  
>  	down_read(&mm->mmap_sem);
>  	if (ksm_test_exit(mm))
> @@ -470,7 +465,7 @@ static void remove_node_from_stable_tree(struct stable_node *stable_node)
>  			ksm_pages_sharing--;
>  		else
>  			ksm_pages_shared--;
> -		drop_anon_vma(rmap_item);
> +		ksm_drop_anon_vma(rmap_item);
>  		rmap_item->address &= PAGE_MASK;
>  		cond_resched();
>  	}
> @@ -558,7 +553,7 @@ static void remove_rmap_item_from_tree(struct rmap_item *rmap_item)
>  		else
>  			ksm_pages_shared--;
>  
> -		drop_anon_vma(rmap_item);
> +		ksm_drop_anon_vma(rmap_item);
>  		rmap_item->address &= PAGE_MASK;
>  
>  	} else if (rmap_item->address & UNSTABLE_FLAG) {
> diff --git a/mm/rmap.c b/mm/rmap.c
> index f0ba648..d63cd91 100644
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -238,6 +238,12 @@ int anon_vma_fork(struct vm_area_struct *vma, struct vm_area_struct *pvma)
>  	 */
>  	root_avc = list_entry(vma->anon_vma_chain.prev, struct anon_vma_chain, same_vma);
>  	anon_vma->root = root_avc->anon_vma;
> +	/*
> +	 * With KSM refcounts, an anon_vma can stay around longer than the
> +	 * process it belongs to.  The root anon_vma needs to be pinned,
> +	 * because that is where the lock lives.
> +	 */
> +	get_anon_vma(anon_vma->root);
>  	/* Mark this anon_vma as the one where our new (COWed) pages go. */
>  	vma->anon_vma = anon_vma;
>  	anon_vma_chain_link(vma, avc, anon_vma);
> @@ -267,8 +273,11 @@ static void anon_vma_unlink(struct anon_vma_chain *anon_vma_chain)
>  	empty = list_empty(&anon_vma->head) && !ksm_refcount(anon_vma);
>  	anon_vma_unlock(anon_vma);
>  
> -	if (empty)
> +	if (empty) {
> +		/* We no longer need the root anon_vma */
> +		drop_anon_vma(anon_vma->root);
>  		anon_vma_free(anon_vma);
> +	}
>  }
>  
>  void unlink_anon_vmas(struct vm_area_struct *vma)
> @@ -1355,6 +1364,40 @@ int try_to_munlock(struct page *page)
>  		return try_to_unmap_file(page, TTU_MUNLOCK);
>  }
>  
> +#ifdef CONFIG_KSM
> +/*
> + * Drop an anon_vma refcount, freeing the anon_vma and anon_vma->root
> + * if necessary.  Be careful to do all the tests under the lock.  Once
> + * we know we are the last user, nobody else can get a reference and we
> + * can do the freeing without the lock.
> + */
> +void drop_anon_vma(struct anon_vma *anon_vma)
> +{
> +	if (atomic_dec_and_lock(&anon_vma->ksm_refcount, &anon_vma->root->lock)) {
> +		struct anon_vma *root = anon_vma->root;
> +		int empty list_empty(&anon_vma->head);
> +		int last_root_user = 0;
> +		int root_empty = 0;
> +
> +		/*
> +		 * The refcount on a non-root anon_vma got dropped.  Drop
> +		 * the refcount on the root and check if we need to free it.
> +		 */
> +		if (empty && anon_vma != root) {
> +			last_root_user = atomic_dec_and_test(&root->ksm_refcount);
> +			root_empty = list_empty(&root->head);
> +		}
> +		anon_vma_unlock(anon_vma);
> +
> +		if (empty) {
> +			anon_vma_free(anon_vma);
> +			if (root_empty && last_root_user)
> +				anon_vma_free(root);
> +		}
> +	}
> +}
> +#endif
> +
>  #ifdef CONFIG_MIGRATION
>  /*
>   * rmap_walk() and its helpers rmap_walk_anon() and rmap_walk_file():
> 

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
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/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2010-05-12 21:07 UTC|newest]

Thread overview: 54+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-05-12 17:38 [PATCH 0/5] always lock the root anon_vma Rik van Riel
2010-05-12 17:39 ` [PATCH 1/5] rename anon_vma_lock to vma_lock_anon_vma Rik van Riel
2010-05-12 20:57   ` Mel Gorman
2010-05-13  0:30   ` KAMEZAWA Hiroyuki
2010-05-12 17:39 ` [PATCH 3/5] track the root (oldest) anon_vma Rik van Riel
2010-05-12 20:59   ` Mel Gorman
2010-05-12 21:01     ` Rik van Riel
2010-05-13  0:38   ` KAMEZAWA Hiroyuki
2010-05-13  2:25     ` Rik van Riel
2010-05-14  0:04       ` KAMEZAWA Hiroyuki
2010-05-12 17:40 ` [PATCH 4/5] always lock " Rik van Riel
2010-05-12 21:02   ` Mel Gorman
2010-05-12 21:08     ` Rik van Riel
2010-05-13  9:54       ` Mel Gorman
2010-05-13 14:33         ` [PATCH -v2 " Rik van Riel
2010-05-13 21:09           ` Andrew Morton
2010-05-13 22:50             ` Rik van Riel
2010-05-14  9:33               ` Mel Gorman
2010-05-26  4:00             ` Rik van Riel
2010-05-26  4:15               ` Andrew Morton
2010-05-26  5:46                 ` james toy
2010-06-01  0:57                   ` james toy
2010-05-26 15:24               ` [PATCH -v2 0/5] always lock the root anon_vma Rik van Riel
2010-05-26 15:25                 ` [PATCH 1/5] rename anon_vma_lock to vma_lock_anon_vma Rik van Riel
2010-05-26 17:25                   ` Linus Torvalds
2010-05-26 19:01                     ` Rik van Riel
2010-05-26 19:25                       ` Linus Torvalds
2010-05-26 19:35                         ` Rik van Riel
2010-05-26 15:25                 ` [PATCH 2/5] change direct call of spin_lock(anon_vma->lock) to inline function Rik van Riel
2010-05-26 15:26                 ` [PATCH 3/5] track the root (oldest) anon_vma Rik van Riel
2010-05-26 15:27                 ` [PATCH 4/5] always lock " Rik van Riel
2010-05-26 15:27                 ` [PATCH 5/5] extend KSM refcounts to the anon_vma root Rik van Riel
2010-05-12 21:55   ` [PATCH 4/5] always lock the root (oldest) anon_vma Linus Torvalds
2010-05-12 22:18     ` Rik van Riel
2010-05-12 22:26       ` Linus Torvalds
2010-05-12 17:41 ` [PATCH 5/5] extend KSM refcounts to the anon_vma root Rik van Riel
2010-05-12 21:07   ` Mel Gorman [this message]
2010-05-12 21:09     ` Rik van Riel
2010-05-13 11:26       ` Mel Gorman
2010-05-13 13:11         ` Rik van Riel
2010-05-13 13:24           ` Mel Gorman
2010-05-13 14:34             ` [PATCH -v2 " Rik van Riel
2010-05-19  1:05               ` Andrea Arcangeli
2010-05-12 17:41 ` [PATCH 2/5] change direct call of spin_lock(anon_vma->lock) to inline function Rik van Riel
2010-05-12 20:58   ` Mel Gorman
2010-05-13  0:32   ` KAMEZAWA Hiroyuki
2010-05-20 22:42 ` [PATCH 6/5] adjust mm_take_all_locks to anon-vma-root locking Andrea Arcangeli
2010-05-20 23:07   ` Rik van Riel
2010-05-26 19:38 [PATCH -v3 0/5] always lock the root anon_vma Rik van Riel
2010-05-26 19:41 ` [PATCH 5/5] extend KSM refcounts to the anon_vma root Rik van Riel
2010-05-26 20:47   ` Larry Woodman
2010-05-27 14:02   ` Minchan Kim
2010-05-27 14:09     ` Rik van Riel
2010-05-27 14:31   ` Minchan Kim
2010-05-27 17:50   ` Mel Gorman

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20100512210706.GQ24989@csn.ul.ie \
    --to=mel@csn.ul.ie \
    --cc=aarcange@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=kamezawa.hiroyu@jp.fujitsu.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=minchan.kim@gmail.com \
    --cc=riel@redhat.com \
    --cc=torvalds@linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox