[PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Jesper Juhl]

Hi,

Here's a small patch for kernel/cgroup.c

Removes a few pointless tests of pointer == 0 before kfree() in 
kernel/cgroup.c.
If the pointer we hand to kfree() is 0, then kfree() is a noop, so there 
is no need to test.

Reduces object file size a bit : 

Before:
$ size kernel/cgroup.o
   text    data     bss     dec     hex filename
  21593    7804    4924   34321    8611 kernel/cgroup.o
After:
$ size kernel/cgroup.o
   text    data     bss     dec     hex filename
  21537    7744    4924   34205    859d kernel/cgroup.o

and source file size shrinks a bit too - always nice.


Signed-off-by: Jesper Juhl <jj@chaosbits.net>
---

diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 9edb5c4..1c0a9b5 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -865,8 +865,7 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
 	if (opts.release_agent)
 		strcpy(root->release_agent_path, opts.release_agent);
  out_unlock:
-	if (opts.release_agent)
-		kfree(opts.release_agent);
+	kfree(opts.release_agent);
 	mutex_unlock(&cgroup_mutex);
 	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
 	return ret;
@@ -969,15 +968,13 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
 	/* First find the desired set of subsystems */
 	ret = parse_cgroupfs_options(data, &opts);
 	if (ret) {
-		if (opts.release_agent)
-			kfree(opts.release_agent);
+		kfree(opts.release_agent);
 		return ret;
 	}
 
 	root = kzalloc(sizeof(*root), GFP_KERNEL);
 	if (!root) {
-		if (opts.release_agent)
-			kfree(opts.release_agent);
+		kfree(opts.release_agent);
 		return -ENOMEM;
 	}
 


-- 
Jesper Juhl <jj@chaosbits.net>             http://www.chaosbits.net/
Plain text mails only, please      http://www.expita.com/nomime.html
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Paul Menage]

On Sun, Feb 22, 2009 at 2:05 PM, Jesper Juhl <jj@chaosbits.net> wrote:
> Hi,
>
> Here's a small patch for kernel/cgroup.c
>
> Removes a few pointless tests of pointer == 0 before kfree() in
> kernel/cgroup.c.
> If the pointer we hand to kfree() is 0, then kfree() is a noop, so there
> is no need to test.
>
> Reduces object file size a bit :
>
> Before:
> $ size kernel/cgroup.o
>   text    data     bss     dec     hex filename
>  21593    7804    4924   34321    8611 kernel/cgroup.o
> After:
> $ size kernel/cgroup.o
>   text    data     bss     dec     hex filename
>  21537    7744    4924   34205    859d kernel/cgroup.o

Thanks. I'm a bit surprised that it also shrunk the data segment size
(and in fact, by more than the text segment size). Any idea how that
came about?

>
> Signed-off-by: Jesper Juhl <jj@chaosbits.net>

Acked-by: Paul Menage <menage@google.com>

Paul

> ---
>
> diff --git a/kernel/cgroup.c b/kernel/cgroup.c
> index 9edb5c4..1c0a9b5 100644
> --- a/kernel/cgroup.c
> +++ b/kernel/cgroup.c
> @@ -865,8 +865,7 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
>        if (opts.release_agent)
>                strcpy(root->release_agent_path, opts.release_agent);
>  out_unlock:
> -       if (opts.release_agent)
> -               kfree(opts.release_agent);
> +       kfree(opts.release_agent);
>        mutex_unlock(&cgroup_mutex);
>        mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
>        return ret;
> @@ -969,15 +968,13 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
>        /* First find the desired set of subsystems */
>        ret = parse_cgroupfs_options(data, &opts);
>        if (ret) {
> -               if (opts.release_agent)
> -                       kfree(opts.release_agent);
> +               kfree(opts.release_agent);
>                return ret;
>        }
>
>        root = kzalloc(sizeof(*root), GFP_KERNEL);
>        if (!root) {
> -               if (opts.release_agent)
> -                       kfree(opts.release_agent);
> +               kfree(opts.release_agent);
>                return -ENOMEM;
>        }
>
>
>
> --
> Jesper Juhl <jj@chaosbits.net>             http://www.chaosbits.net/
> Plain text mails only, please      http://www.expita.com/nomime.html
> Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
>
>

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Cyrill Gorcunov]

[Paul Menage - Mon, Feb 23, 2009 at 12:12:51AM -0800]
| On Sun, Feb 22, 2009 at 2:05 PM, Jesper Juhl <jj@chaosbits.net> wrote:
| > Hi,
| >
| > Here's a small patch for kernel/cgroup.c
| >
| > Removes a few pointless tests of pointer == 0 before kfree() in
| > kernel/cgroup.c.
| > If the pointer we hand to kfree() is 0, then kfree() is a noop, so there
| > is no need to test.
| >
| > Reduces object file size a bit :
| >
| > Before:
| > $ size kernel/cgroup.o
| >   text    data     bss     dec     hex filename
| >  21593    7804    4924   34321    8611 kernel/cgroup.o
| > After:
| > $ size kernel/cgroup.o
| >   text    data     bss     dec     hex filename
| >  21537    7744    4924   34205    859d kernel/cgroup.o
| 
| Thanks. I'm a bit surprised that it also shrunk the data segment size
| (and in fact, by more than the text segment size). Any idea how that
| came about?

Maybe kfree was inlined by gcc for some reason?

| 
| >
| > Signed-off-by: Jesper Juhl <jj@chaosbits.net>
| 
| Acked-by: Paul Menage <menage@google.com>
| 
| Paul
| 
| > ---
| >
| > diff --git a/kernel/cgroup.c b/kernel/cgroup.c
| > index 9edb5c4..1c0a9b5 100644
| > --- a/kernel/cgroup.c
| > +++ b/kernel/cgroup.c
| > @@ -865,8 +865,7 @@ static int cgroup_remount(struct super_block *sb, int *flags, char *data)
| >        if (opts.release_agent)
| >                strcpy(root->release_agent_path, opts.release_agent);
| >  out_unlock:
| > -       if (opts.release_agent)
| > -               kfree(opts.release_agent);
| > +       kfree(opts.release_agent);
| >        mutex_unlock(&cgroup_mutex);
| >        mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
| >        return ret;
| > @@ -969,15 +968,13 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
| >        /* First find the desired set of subsystems */
| >        ret = parse_cgroupfs_options(data, &opts);
| >        if (ret) {
| > -               if (opts.release_agent)
| > -                       kfree(opts.release_agent);
| > +               kfree(opts.release_agent);
| >                return ret;
| >        }
| >
| >        root = kzalloc(sizeof(*root), GFP_KERNEL);
| >        if (!root) {
| > -               if (opts.release_agent)
| > -                       kfree(opts.release_agent);
| > +               kfree(opts.release_agent);
| >                return -ENOMEM;
| >        }
| >
| >
| >
| > --
| > Jesper Juhl <jj@chaosbits.net>             http://www.chaosbits.net/
| > Plain text mails only, please      http://www.expita.com/nomime.html
| > Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
| >
| >
| 
	- Cyrill -

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Cyrill Gorcunov]

[Cyrill Gorcunov - Mon, Feb 23, 2009 at 11:17:57AM +0300]
...
| | Thanks. I'm a bit surprised that it also shrunk the data segment size
| | (and in fact, by more than the text segment size). Any idea how that
| | came about?
| 
| Maybe kfree was inlined by gcc for some reason?

What I'm talking about? :)
Sorry for noise.

	- Cyrill -

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Li Zefan]

>> Reduces object file size a bit :
>>
>> Before:
>> $ size kernel/cgroup.o
>>   text    data     bss     dec     hex filename
>>  21593    7804    4924   34321    8611 kernel/cgroup.o
>> After:
>> $ size kernel/cgroup.o
>>   text    data     bss     dec     hex filename
>>  21537    7744    4924   34205    859d kernel/cgroup.o
> 
> Thanks. I'm a bit surprised that it also shrunk the data segment size
> (and in fact, by more than the text segment size). Any idea how that
> came about?
> 

A bit odd..

Here is the result on my x86_32:
   text    data     bss     dec     hex filename
  14118    1804    4960   20882    5192 kernel/cgroup.o.orig
   text    data     bss     dec     hex filename
  14114    1804    4960   20878    518e kernel/cgroup.o

And on IA64:
   text    data     bss     dec     hex filename
  42399    2712    5616   50727    c627 kernel/cgroup.o.orig
   text    data     bss     dec     hex filename
  42367    2712    5616   50695    c607 kernel/cgroup.o

No change in data segment.

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Li Zefan]

Jesper Juhl wrote:
> Hi,
> 
> Here's a small patch for kernel/cgroup.c
> 
> Removes a few pointless tests of pointer == 0 before kfree() in 
> kernel/cgroup.c.
> If the pointer we hand to kfree() is 0, then kfree() is a noop, so there 
> is no need to test.
> 
> Reduces object file size a bit : 
> 
> Before:
> $ size kernel/cgroup.o
>    text    data     bss     dec     hex filename
>   21593    7804    4924   34321    8611 kernel/cgroup.o
> After:
> $ size kernel/cgroup.o
>    text    data     bss     dec     hex filename
>   21537    7744    4924   34205    859d kernel/cgroup.o
> 
> and source file size shrinks a bit too - always nice.
> 
> 
> Signed-off-by: Jesper Juhl <jj@chaosbits.net>

Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>

Re: [PATCH] Remove some pointless conditionals before kfree() in kernel/cgroup.c

[Jesper Juhl]

On Mon, 23 Feb 2009, Paul Menage wrote:

> On Sun, Feb 22, 2009 at 2:05 PM, Jesper Juhl <jj@chaosbits.net> wrote:
> > Hi,
> >
> > Here's a small patch for kernel/cgroup.c
> >
> > Removes a few pointless tests of pointer == 0 before kfree() in
> > kernel/cgroup.c.
> > If the pointer we hand to kfree() is 0, then kfree() is a noop, so there
> > is no need to test.
> >
> > Reduces object file size a bit :
> >
> > Before:
> > $ size kernel/cgroup.o
> >   text    data     bss     dec     hex filename
> >  21593    7804    4924   34321    8611 kernel/cgroup.o
> > After:
> > $ size kernel/cgroup.o
> >   text    data     bss     dec     hex filename
> >  21537    7744    4924   34205    859d kernel/cgroup.o
> 
> Thanks. I'm a bit surprised that it also shrunk the data segment size
> (and in fact, by more than the text segment size). Any idea how that
> came about?
> 
Can't say I've got a good explanation for that, no.

-- 
Jesper Juhl <jj@chaosbits.net>             http://www.chaosbits.net/
Plain text mails only, please      http://www.expita.com/nomime.html
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html