[PATCH] Reduce stack usage in time.c

[PATCH] Reduce stack usage in time.c

[Yum Rayan]

Attempt to reduce stack usage in time.c (linux-2.6.12-rc1-mm3). Stack
usage was noted using checkstack.pl. Specifically:

Before patch
------------
sys_adjtimex - 128

After patch
-----------
sys_adjtimex - none (register usage only)

Signed-off-by: Yum Rayan <yum.rayan@gmail.com>

--- a/kernel/time.c	2005-03-25 22:11:06.000000000 -0800
+++ b/kernel/time.c	2005-03-30 16:59:51.000000000 -0800
@@ -413,17 +413,27 @@
 
 asmlinkage long sys_adjtimex(struct timex __user *txc_p)
 {
-	struct timex txc;		/* Local copy of parameter */
-	int ret;
+	struct timex *txc;		/* Local copy of parameter */
+	int retval;
+
+	txc = kmalloc(sizeof(struct timex), GFP_KERNEL);
+	if (!txc)
+		return -ENOMEM;
 
 	/* Copy the user data space into the kernel copy
 	 * structure. But bear in mind that the structures
 	 * may change
 	 */
-	if(copy_from_user(&txc, txc_p, sizeof(struct timex)))
-		return -EFAULT;
-	ret = do_adjtimex(&txc);
-	return copy_to_user(txc_p, &txc, sizeof(struct timex)) ? -EFAULT : ret;
+	if(copy_from_user(txc, txc_p, sizeof(struct timex))) {
+		retval = -EFAULT;
+		goto free_txc;
+	}
+	retval = do_adjtimex(txc);
+	if (copy_to_user(txc_p, txc, sizeof(struct timex)))
+		retval = -EFAULT;
+free_txc:
+	kfree(txc);
+	return retval;
 }
 
 inline struct timespec current_kernel_time(void)

Re: [PATCH] Reduce stack usage in time.c

[Jeff Garzik]

Yum Rayan wrote:
> Attempt to reduce stack usage in time.c (linux-2.6.12-rc1-mm3). Stack
> usage was noted using checkstack.pl. Specifically:
> 
> Before patch
> ------------
> sys_adjtimex - 128
> 
> After patch
> -----------
> sys_adjtimex - none (register usage only)
> 
> Signed-off-by: Yum Rayan <yum.rayan@gmail.com>
> 
> --- a/kernel/time.c	2005-03-25 22:11:06.000000000 -0800
> +++ b/kernel/time.c	2005-03-30 16:59:51.000000000 -0800
> @@ -413,17 +413,27 @@
>  
>  asmlinkage long sys_adjtimex(struct timex __user *txc_p)
>  {
> -	struct timex txc;		/* Local copy of parameter */
> -	int ret;
> +	struct timex *txc;		/* Local copy of parameter */
> +	int retval;
> +
> +	txc = kmalloc(sizeof(struct timex), GFP_KERNEL);
> +	if (!txc)
> +		return -ENOMEM;
>  
>  	/* Copy the user data space into the kernel copy
>  	 * structure. But bear in mind that the structures
>  	 * may change
>  	 */
> -	if(copy_from_user(&txc, txc_p, sizeof(struct timex)))
> -		return -EFAULT;
> -	ret = do_adjtimex(&txc);
> -	return copy_to_user(txc_p, &txc, sizeof(struct timex)) ? -EFAULT : ret;
> +	if(copy_from_user(txc, txc_p, sizeof(struct timex))) {
> +		retval = -EFAULT;
> +		goto free_txc;
> +	}
> +	retval = do_adjtimex(txc);
> +	if (copy_to_user(txc_p, txc, sizeof(struct timex)))
> +		retval = -EFAULT;
> +free_txc:

It seems quite unhealthy to add a kmalloc(,GFP_KERNEL) allocation into a 
time-sensitive function.

	Jeff

Re: [PATCH] Reduce stack usage in time.c

[Denis Vlasenko]

On Thursday 31 March 2005 11:26, Jeff Garzik wrote:
> Yum Rayan wrote:
> > Attempt to reduce stack usage in time.c (linux-2.6.12-rc1-mm3). Stack
> > usage was noted using checkstack.pl. Specifically:
> > 
> > Before patch
> > ------------
> > sys_adjtimex - 128
> > 
> > After patch
> > -----------
> > sys_adjtimex - none (register usage only)
> > 
> > Signed-off-by: Yum Rayan <yum.rayan@gmail.com>
> > 
> > --- a/kernel/time.c	2005-03-25 22:11:06.000000000 -0800
> > +++ b/kernel/time.c	2005-03-30 16:59:51.000000000 -0800
> > @@ -413,17 +413,27 @@
> >  
> >  asmlinkage long sys_adjtimex(struct timex __user *txc_p)
> >  {
> > -	struct timex txc;		/* Local copy of parameter */
> > -	int ret;
> > +	struct timex *txc;		/* Local copy of parameter */
> > +	int retval;
> > +
> > +	txc = kmalloc(sizeof(struct timex), GFP_KERNEL);
> > +	if (!txc)
> > +		return -ENOMEM;
> >  
> >  	/* Copy the user data space into the kernel copy
> >  	 * structure. But bear in mind that the structures
> >  	 * may change
> >  	 */
> > -	if(copy_from_user(&txc, txc_p, sizeof(struct timex)))
> > -		return -EFAULT;
> > -	ret = do_adjtimex(&txc);
> > -	return copy_to_user(txc_p, &txc, sizeof(struct timex)) ? -EFAULT : ret;
> > +	if(copy_from_user(txc, txc_p, sizeof(struct timex))) {
> > +		retval = -EFAULT;
> > +		goto free_txc;
> > +	}
> > +	retval = do_adjtimex(txc);
> > +	if (copy_to_user(txc_p, txc, sizeof(struct timex)))
> > +		retval = -EFAULT;
> > +free_txc:

Is this a syscall?
Is it ever called from some deeply nested kernel function?
--
vda

Re: [PATCH] Reduce stack usage in time.c

[Jörn Engel]

On Thu, 31 March 2005 12:26:58 +0300, Denis Vlasenko wrote:
> 
> Is this a syscall?
> Is it ever called from some deeply nested kernel function?

Never showed up in any of my callchain-tests.  I'd leave it as is.

Jörn

-- 
When I am working on a problem I never think about beauty.  I think
only how to solve the problem.  But when I have finished, if the
solution is not beautiful, I know it is wrong.
-- R. Buckminster Fuller