Linux Tiny Patch Ideas - Revision history

Peter Huewe: ac

2011-10-27T10:29:06Z

Tim Bird: add PROC_FS_FULL idea and link to printk reduction ideas

2007-10-04T18:08:24Z

add PROC_FS_FULL idea and link to printk reduction ideas

Tim Bird: /* proposal to compress printk messages */

2007-09-21T22:49:46Z

‎proposal to compress printk messages

Tim Bird at 22:49, 21 September 2007

2007-09-21T22:49:38Z

Tim Bird: create page

2007-09-19T21:26:36Z

create page

New page

This is an unorganized list of items that may be worth looking at:

== Mark Rustad patch to remove init printk messages ==

http://lkml.org/lkml/2004/12/9/146

This change allows format strings in printks to be directed to an object section where they can be discarded at run time. By using "pkinit" instead of "printk" inside of __init functions, the format strings associated with the __init functions can be reclaimed.

← Older revision		Revision as of 10:29, 27 October 2011
Line 32:		Line 32:
	and maybe 'free' need, and find a way to make the others conditional on		and maybe 'free' need, and find a way to make the others conditional on
	CONFIG_PROC_FS_FULL in the least intrusive way.		CONFIG_PROC_FS_FULL in the least intrusive way.
		+
		+	[[Category:Linux Tiny]]

← Older revision		Revision as of 18:08, 4 October 2007
Line 13:		Line 13:

	http://selenic.com/pipermail/linux-tiny/2005-June/000207.html		http://selenic.com/pipermail/linux-tiny/2005-June/000207.html
		+
		+	=== compile-time selection of printk messages, by message level ===
		+	There was a big discussion about this on LKML in September, 2007.
		+
		+	See http://lkml.org/lkml/2007/9/20/352 (by Rob Landley) for the start of the thread.
		+
		+	== Other Areas of reduction ==
		+	=== make a PROC_FS_FULL ===
		+	There's a lot of junk in /proc file system. If you turn off CONFIG_PROC_FS it saves
		+	about 130K, but it makes the system pretty difficult to use (e.g. mount has problems
		+	on init that you have to work around, there's no 'ps' is available, etc.)
		+
		+	Similar to how we have BASE_FULL and BASE_SMALL, maybe we can have a PROC_FS_FULL
		+	and a PROC_FS_SMALL. Only the essential items would be configured on with
		+	PROC_FS_SMALL, and everything would be on for PROC_FS_FULL.
		+
		+	The research needed here is to find out the list of essential items that 'ps'
		+	and maybe 'free' need, and find a way to make the others conditional on
		+	CONFIG_PROC_FS_FULL in the least intrusive way.

@@ Line 8: / Line 8: @@
 This change allows format strings in printks to be directed to an object section where they can be discarded at run time. By using "pkinit" instead of "printk" inside of __init functions, the format strings associated with the __init functions can be reclaimed.
-=== proposal to compress printk messages ===
+=== proposals to compress printk messages ===
 http://uwsg.indiana.edu/hypermail/linux/kernel/0306.0/1542.html - Timothy Miller actual results
 (using a compile time string replacer)
 http://selenic.com/pipermail/linux-tiny/2005-June/000207.html

@@ Line 1: / Line 1: @@
 This is an unorganized list of items that may be worth looking at:
-== Mark Rustad patch to remove init printk messages ==
+== stuff about printk messages ==
 http://lkml.org/lkml/2004/12/9/146
 This change allows format strings in printks to be directed to an object section where they can be discarded at run time. By using "pkinit" instead of "printk" inside of __init functions, the format strings associated with the __init functions can be reclaimed.