[PATCH v3] selinux: reduce locking overhead in inode_free_security()

[PATCH v3] selinux: reduce locking overhead in inode_free_security()

[Waiman Long]

The inode_free_security() function just took the superblock's isec_lock
before checking and trying to remove the inode security struct from the
linked list. In many cases, the list was empty and so the lock taking
is wasteful as no useful work is done. On multi-socket systems with
a large number of CPUs, there can also be a fair amount of spinlock
contention on the isec_lock if many tasks are exiting at the same time.

This patch changes the code to check the state of the list first before
taking the lock and attempting to dequeue it. The list_del_init()
can be called more than once on the same list with no harm as long
as they are properly serialized. It should not be possible to have
inode_free_security() called concurrently with list_add(). For better
safety, however, we use list_empty_careful() here even though it is
still not completely safe in case that happens.

Signed-off-by: Waiman Long <Waiman.Long@hp.com>
---
 security/selinux/hooks.c |   17 ++++++++++++++---
 1 files changed, 14 insertions(+), 3 deletions(-)

v1->v2:
 - Take out the second list_empty() test inside the lock.

v2->v3:
 - Fix incorrent comment and commit log message.

diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
index 7dade28..2a99804 100644
--- a/security/selinux/hooks.c
+++ b/security/selinux/hooks.c
@@ -254,10 +254,21 @@ static void inode_free_security(struct inode *inode)
 	struct inode_security_struct *isec = inode->i_security;
 	struct superblock_security_struct *sbsec = inode->i_sb->s_security;
 
-	spin_lock(&sbsec->isec_lock);
-	if (!list_empty(&isec->list))
+	/*
+	 * As not all inode security structures are in a list, we check for
+	 * empty list outside of the lock to make sure that we won't waste
+	 * time taking a lock doing nothing.
+	 *
+	 * The list_del_init() function can be safely called more than once.
+	 * It should not be possible for this function to be called with
+	 * concurrent list_add(), but for better safety against future changes
+	 * in the code, we use list_empty_careful() here.
+	 */
+	if (!list_empty_careful(&isec->list)) {
+		spin_lock(&sbsec->isec_lock);
 		list_del_init(&isec->list);
-	spin_unlock(&sbsec->isec_lock);
+		spin_unlock(&sbsec->isec_lock);
+	}
 
 	/*
 	 * The inode may still be referenced in a path walk and
-- 
1.7.1

Re: [PATCH v3] selinux: reduce locking overhead in inode_free_security()

[Stephen Smalley]

On 06/15/2015 01:13 PM, Waiman Long wrote:
> The inode_free_security() function just took the superblock's isec_lock
> before checking and trying to remove the inode security struct from the
> linked list. In many cases, the list was empty and so the lock taking
> is wasteful as no useful work is done. On multi-socket systems with
> a large number of CPUs, there can also be a fair amount of spinlock
> contention on the isec_lock if many tasks are exiting at the same time.
> 
> This patch changes the code to check the state of the list first before
> taking the lock and attempting to dequeue it. The list_del_init()
> can be called more than once on the same list with no harm as long
> as they are properly serialized. It should not be possible to have
> inode_free_security() called concurrently with list_add(). For better
> safety, however, we use list_empty_careful() here even though it is
> still not completely safe in case that happens.
> 
> Signed-off-by: Waiman Long <Waiman.Long@hp.com>

Acked-by:  Stephen Smalley <sds@tycho.nsa.gov>

> ---
>  security/selinux/hooks.c |   17 ++++++++++++++---
>  1 files changed, 14 insertions(+), 3 deletions(-)
> 
> v1->v2:
>  - Take out the second list_empty() test inside the lock.
> 
> v2->v3:
>  - Fix incorrent comment and commit log message.
> 
> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 7dade28..2a99804 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -254,10 +254,21 @@ static void inode_free_security(struct inode *inode)
>  	struct inode_security_struct *isec = inode->i_security;
>  	struct superblock_security_struct *sbsec = inode->i_sb->s_security;
>  
> -	spin_lock(&sbsec->isec_lock);
> -	if (!list_empty(&isec->list))
> +	/*
> +	 * As not all inode security structures are in a list, we check for
> +	 * empty list outside of the lock to make sure that we won't waste
> +	 * time taking a lock doing nothing.
> +	 *
> +	 * The list_del_init() function can be safely called more than once.
> +	 * It should not be possible for this function to be called with
> +	 * concurrent list_add(), but for better safety against future changes
> +	 * in the code, we use list_empty_careful() here.
> +	 */
> +	if (!list_empty_careful(&isec->list)) {
> +		spin_lock(&sbsec->isec_lock);
>  		list_del_init(&isec->list);
> -	spin_unlock(&sbsec->isec_lock);
> +		spin_unlock(&sbsec->isec_lock);
> +	}
>  
>  	/*
>  	 * The inode may still be referenced in a path walk and
> 

Re: [PATCH v3] selinux: reduce locking overhead in inode_free_security()

[Yury]

On 15.06.2015 20:13, Waiman Long wrote:
> The inode_free_security() function just took the superblock's isec_lock
> before checking and trying to remove the inode security struct from the
> linked list. In many cases, the list was empty and so the lock taking
> is wasteful as no useful work is done. On multi-socket systems with
> a large number of CPUs, there can also be a fair amount of spinlock
> contention on the isec_lock if many tasks are exiting at the same time.
>
> This patch changes the code to check the state of the list first before
> taking the lock and attempting to dequeue it. The list_del_init()
> can be called more than once on the same list with no harm as long
> as they are properly serialized. It should not be possible to have
> inode_free_security() called concurrently with list_add(). For better
> safety, however, we use list_empty_careful() here even though it is
> still not completely safe in case that happens.
>
> Signed-off-by: Waiman Long <Waiman.Long@hp.com>
> ---
>   security/selinux/hooks.c |   17 ++++++++++++++---
>   1 files changed, 14 insertions(+), 3 deletions(-)
>
> v1->v2:
>   - Take out the second list_empty() test inside the lock.
>
> v2->v3:
>   - Fix incorrent comment and commit log message.
>
> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 7dade28..2a99804 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -254,10 +254,21 @@ static void inode_free_security(struct inode *inode)
>   	struct inode_security_struct *isec = inode->i_security;
>   	struct superblock_security_struct *sbsec = inode->i_sb->s_security;
>   
> -	spin_lock(&sbsec->isec_lock);
> -	if (!list_empty(&isec->list))
> +	/*
> +	 * As not all inode security structures are in a list, we check for
> +	 * empty list outside of the lock to make sure that we won't waste
> +	 * time taking a lock doing nothing.
> +	 *
> +	 * The list_del_init() function can be safely called more than once.
> +	 * It should not be possible for this function to be called with
> +	 * concurrent list_add(), but for better safety against future changes
> +	 * in the code, we use list_empty_careful() here.
> +	 */
> +	if (!list_empty_careful(&isec->list)) {
> +		spin_lock(&sbsec->isec_lock);
>   		list_del_init(&isec->list);
> -	spin_unlock(&sbsec->isec_lock);
> +		spin_unlock(&sbsec->isec_lock);
> +	}
>   
>   	/*
>   	 * The inode may still be referenced in a path walk and
Hi Waiman,

If you need my Acked-by, you have it.

BR,
Yury

Re: [PATCH v3] selinux: reduce locking overhead in inode_free_security()

[Paul Moore]

On Monday, June 15, 2015 01:13:39 PM Waiman Long wrote:
> The inode_free_security() function just took the superblock's isec_lock
> before checking and trying to remove the inode security struct from the
> linked list. In many cases, the list was empty and so the lock taking
> is wasteful as no useful work is done. On multi-socket systems with
> a large number of CPUs, there can also be a fair amount of spinlock
> contention on the isec_lock if many tasks are exiting at the same time.
> 
> This patch changes the code to check the state of the list first before
> taking the lock and attempting to dequeue it. The list_del_init()
> can be called more than once on the same list with no harm as long
> as they are properly serialized. It should not be possible to have
> inode_free_security() called concurrently with list_add(). For better
> safety, however, we use list_empty_careful() here even though it is
> still not completely safe in case that happens.
> 
> Signed-off-by: Waiman Long <Waiman.Long@hp.com>
> ---
>  security/selinux/hooks.c |   17 ++++++++++++++---
>  1 files changed, 14 insertions(+), 3 deletions(-)
> 
> v1->v2:
>  - Take out the second list_empty() test inside the lock.
> 
> v2->v3:
>  - Fix incorrent comment and commit log message.

Thanks for the patch and the discussion; I've added this to the SELinux next-
queue branch and I'll push it to selinux#next as soon as the merge window 
closes.

> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 7dade28..2a99804 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -254,10 +254,21 @@ static void inode_free_security(struct inode *inode)
>  	struct inode_security_struct *isec = inode->i_security;
>  	struct superblock_security_struct *sbsec = inode->i_sb->s_security;
> 
> -	spin_lock(&sbsec->isec_lock);
> -	if (!list_empty(&isec->list))
> +	/*
> +	 * As not all inode security structures are in a list, we check for
> +	 * empty list outside of the lock to make sure that we won't waste
> +	 * time taking a lock doing nothing.
> +	 *
> +	 * The list_del_init() function can be safely called more than once.
> +	 * It should not be possible for this function to be called with
> +	 * concurrent list_add(), but for better safety against future changes
> +	 * in the code, we use list_empty_careful() here.
> +	 */
> +	if (!list_empty_careful(&isec->list)) {
> +		spin_lock(&sbsec->isec_lock);
>  		list_del_init(&isec->list);
> -	spin_unlock(&sbsec->isec_lock);
> +		spin_unlock(&sbsec->isec_lock);
> +	}
> 
>  	/*
>  	 * The inode may still be referenced in a path walk and

-- 
paul moore
www.paul-moore.com