[Fake patch] Make sysfs_dirent.s_type an unsigned short

[Fake patch] Make sysfs_dirent.s_type an unsigned short

[Adam J. Richter]

	Here is a fake patch against my heavily hacked sysfs tree
to change sysfs_dirent.s_type from an int to an unsigned short.
It appears next to another unsigned short (s_mode), so it should
save 4 bytes per sysfs node.

	Note that this patch will not apply to a pristine 2.6.10-rc2-bk15
tree, because I've moved the declaration of struct sysfs_dirent
from include/linux/sysfs.h to fs/sysfs/sysfs.h in a previous patch.

	By the way, I have to sheepishly admit that somehow I previously
underestimated the size of struct sysfs_dirent.  Only now
with s_children and s_count removed and s_type shortened to 16 bits
does sysfs_dirent occupy 32 bytes, according to /proc/slabinfo.
This does not effect my previous statements about how much memory
is saved by each of the patches that I've posted.  It just means
the original amount of memory being used was more.

                    __     ______________
Adam J. Richter        \ /
adam@yggdrasil.com      | g g d r a s i l

--- linux.prev/fs/sysfs/sysfs.h	2004-12-03 11:51:19.000000000 +0800
+++ linux/fs/sysfs/sysfs.h	2004-12-03 00:51:44.000000000 +0800
@@ -13,7 +13,7 @@
 struct sysfs_dirent {
 	struct list_head	s_sibling;
 	void 			* s_element;
-	int			s_type;
+	unsigned short		s_type;
 	umode_t			s_mode;
 	struct dentry		* s_dentry;
 };

Re: [Fake patch] Make sysfs_dirent.s_type an unsigned short

[Greg KH]

On Thu, Dec 02, 2004 at 07:56:28PM -0800, Adam J. Richter wrote:
> 	Here is a fake patch against my heavily hacked sysfs tree
> to change sysfs_dirent.s_type from an int to an unsigned short.
> It appears next to another unsigned short (s_mode), so it should
> save 4 bytes per sysfs node.

Note, Maneesh is currently moving across continents so he's going to be
away from email for a while.  I'm queuing up most of these changes to go
in after 2.6.10 is out, as most of them look good.

thanks,

greg k-h

Re: [Fake patch] Make sysfs_dirent.s_type an unsigned short

[Maneesh Soni]

On Fri, Dec 03, 2004 at 04:08:58AM +0000, Adam J. Richter wrote:
> 	Here is a fake patch against my heavily hacked sysfs tree
> to change sysfs_dirent.s_type from an int to an unsigned short.
> It appears next to another unsigned short (s_mode), so it should
> save 4 bytes per sysfs node.
> 
> 	Note that this patch will not apply to a pristine 2.6.10-rc2-bk15
> tree, because I've moved the declaration of struct sysfs_dirent
> from include/linux/sysfs.h to fs/sysfs/sysfs.h in a previous patch.
> 
> 	By the way, I have to sheepishly admit that somehow I previously
> underestimated the size of struct sysfs_dirent.  Only now
> with s_children and s_count removed and s_type shortened to 16 bits
> does sysfs_dirent occupy 32 bytes, according to /proc/slabinfo.
> This does not effect my previous statements about how much memory
> is saved by each of the patches that I've posted.  It just means
> the original amount of memory being used was more.
> 

Looks sane.

Thanks
Maneesh

> --- linux.prev/fs/sysfs/sysfs.h	2004-12-03 11:51:19.000000000 +0800
> +++ linux/fs/sysfs/sysfs.h	2004-12-03 00:51:44.000000000 +0800
> @@ -13,7 +13,7 @@
>  struct sysfs_dirent {
>  	struct list_head	s_sibling;
>  	void 			* s_element;
> -	int			s_type;
> +	unsigned short		s_type;
>  	umode_t			s_mode;
>  	struct dentry		* s_dentry;
>  };
> -

-- 
Maneesh Soni
Linux Technology Center, 
IBM India Software Labs,
Bangalore, India
email: maneesh@in.ibm.com
Phone: 91-80-25044990