[PATCH] alloc uid cleanup

[PATCH] alloc uid cleanup

[Prasanna Meda]

Cleanup: Release the lock before key_put methods. They call 
schedule_work etc. They block interrupts now, so it is not bug fix.

Signed-off-by: Prasanna Meda

--- a/kernel/user.c	2006-04-17 23:02:54.000000000 +0530
+++ b/kernel/user.c	2006-04-17 23:06:01.000000000 +0530
@@ -160,15 +160,15 @@ struct user_struct * alloc_uid(uid_t uid
 		spin_lock_irq(&uidhash_lock);
 		up = uid_hash_find(uid, hashent);
 		if (up) {
+			spin_unlock_irq(&uidhash_lock);
 			key_put(new->uid_keyring);
 			key_put(new->session_keyring);
 			kmem_cache_free(uid_cachep, new);
 		} else {
 			uid_hash_insert(new, hashent);
 			up = new;
+			spin_unlock_irq (&uidhash_lock);
 		}
-		spin_unlock_irq(&uidhash_lock);
-
 	}
 	return up;
 }

--
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>

Re: [PATCH] alloc uid cleanup

[Andrew Morton]

Prasanna Meda <mlp@google.com> wrote:
>
> Cleanup: Release the lock before key_put methods. They call 
>  schedule_work etc. They block interrupts now, so it is not bug fix.
> 
>  Signed-off-by: Prasanna Meda
> 
>  --- a/kernel/user.c	2006-04-17 23:02:54.000000000 +0530
>  +++ b/kernel/user.c	2006-04-17 23:06:01.000000000 +0530
>  @@ -160,15 +160,15 @@ struct user_struct * alloc_uid(uid_t uid
>   		spin_lock_irq(&uidhash_lock);
>   		up = uid_hash_find(uid, hashent);
>   		if (up) {
>  +			spin_unlock_irq(&uidhash_lock);
>   			key_put(new->uid_keyring);
>   			key_put(new->session_keyring);
>   			kmem_cache_free(uid_cachep, new);
>   		} else {
>   			uid_hash_insert(new, hashent);
>   			up = new;
>  +			spin_unlock_irq (&uidhash_lock);
>   		}
>  -		spin_unlock_irq(&uidhash_lock);
>  -
>   	}
>   	return up;
>   }

The path you're optimising is the wildly-improbable
we-raced-with-someone-else one.  The benefits from this patch will most
likely be outweighed by the loss of increased text size.

--
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>