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

[PATCH] 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. We still need
to do the empty list test inside the lock for safety reason, but it
minimizes the chance of unnecessary spinlock contention.

Signed-off-by: Waiman Long <Waiman.Long@hp.com>
---
 security/selinux/hooks.c |   17 +++++++++++++----
 1 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
index 7dade28..cd736c3 100644
--- a/security/selinux/hooks.c
+++ b/security/selinux/hooks.c
@@ -254,10 +254,19 @@ 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))
-		list_del_init(&isec->list);
-	spin_unlock(&sbsec->isec_lock);
+	/*
+	 * 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. Lock taking can be slow
+	 * especially if the lock is being contended. We do, however, need
+	 * to recheck the list again before deleting it for safety.
+	 */
+	if (!list_empty(&isec->list)) {
+		spin_lock(&sbsec->isec_lock);
+		if (!list_empty(&isec->list))
+			list_del_init(&isec->list);
+		spin_unlock(&sbsec->isec_lock);
+	}
 
 	/*
 	 * The inode may still be referenced in a path walk and
-- 
1.7.1

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

[Stephen Smalley]

On 06/10/2015 04:17 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. We still need
> to do the empty list test inside the lock for safety reason, but it
> minimizes the chance of unnecessary spinlock contention.
> 
> Signed-off-by: Waiman Long <Waiman.Long@hp.com>
> ---
>  security/selinux/hooks.c |   17 +++++++++++++----
>  1 files changed, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 7dade28..cd736c3 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -254,10 +254,19 @@ 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))
> -		list_del_init(&isec->list);
> -	spin_unlock(&sbsec->isec_lock);
> +	/*
> +	 * 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. Lock taking can be slow
> +	 * especially if the lock is being contended. We do, however, need
> +	 * to recheck the list again before deleting it for safety.
> +	 */
> +	if (!list_empty(&isec->list)) {
> +		spin_lock(&sbsec->isec_lock);
> +		if (!list_empty(&isec->list))
> +			list_del_init(&isec->list);
> +		spin_unlock(&sbsec->isec_lock);
> +	}
>  
>  	/*
>  	 * The inode may still be referenced in a path walk and
> 

Do we really need the second list_empty() test at all?
Once removed, inode security structures are never re-added to the list.
For comparison, inode_sb_list_del() only tests list_empty() outside the
lock.

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

[Waiman Long]

On 06/11/2015 08:38 AM, Stephen Smalley wrote:
> On 06/10/2015 04:17 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. We still need
>> to do the empty list test inside the lock for safety reason, but it
>> minimizes the chance of unnecessary spinlock contention.
>>
>> Signed-off-by: Waiman Long<Waiman.Long@hp.com>
>> ---
>>   security/selinux/hooks.c |   17 +++++++++++++----
>>   1 files changed, 13 insertions(+), 4 deletions(-)
>>
>> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
>> index 7dade28..cd736c3 100644
>> --- a/security/selinux/hooks.c
>> +++ b/security/selinux/hooks.c
>> @@ -254,10 +254,19 @@ 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))
>> -		list_del_init(&isec->list);
>> -	spin_unlock(&sbsec->isec_lock);
>> +	/*
>> +	 * 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. Lock taking can be slow
>> +	 * especially if the lock is being contended. We do, however, need
>> +	 * to recheck the list again before deleting it for safety.
>> +	 */
>> +	if (!list_empty(&isec->list)) {
>> +		spin_lock(&sbsec->isec_lock);
>> +		if (!list_empty(&isec->list))
>> +			list_del_init(&isec->list);
>> +		spin_unlock(&sbsec->isec_lock);
>> +	}
>>
>>   	/*
>>   	 * The inode may still be referenced in a path walk and
>>
> Do we really need the second list_empty() test at all?
> Once removed, inode security structures are never re-added to the list.
> For comparison, inode_sb_list_del() only tests list_empty() outside the
> lock.
>

Yes, I think we can remove the second list_empty() test. I will update 
the patch to do that.

Cheers,
Longman

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

[Stephen Smalley]

On 06/10/2015 04:17 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. We still need
> to do the empty list test inside the lock for safety reason, but it
> minimizes the chance of unnecessary spinlock contention.
> 
> 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, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 7dade28..cd736c3 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -254,10 +254,19 @@ 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))
> -		list_del_init(&isec->list);
> -	spin_unlock(&sbsec->isec_lock);
> +	/*
> +	 * 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. Lock taking can be slow
> +	 * especially if the lock is being contended. We do, however, need
> +	 * to recheck the list again before deleting it for safety.
> +	 */
> +	if (!list_empty(&isec->list)) {
> +		spin_lock(&sbsec->isec_lock);
> +		if (!list_empty(&isec->list))
> +			list_del_init(&isec->list);
> +		spin_unlock(&sbsec->isec_lock);
> +	}
>  
>  	/*
>  	 * The inode may still be referenced in a path walk and
>