Linux: 2.4.21-pre3-ac1; rmap removed

Submitted by Jeremy

on January 9, 2003 - 3:16am

Linux news

Alan Cox [interview] released a patch against Marcelo's latest -pre kernel today saying, "Its back, its smaller and it wants a lot of testing. Much of the older stuff has been merged with Marcelo, and some stuff has been dropped out." Alan's last patch against a 2.4 kernel was 2.4.20-ac2 nearly a month ago, weighing in at just under seven megabytes. Alan's current patch is 2.4.21-pre3-ac1, weighing in at just under three megabytes, thus significantly smaller.

One of the items removed in this version was Rik van Riel's [interview] 15a rmap VM, which is now part of the 2.5 kernel [story]. Alan explains, "I took out the 15a rmap because it has problems and is obsolete. I've left the 15b one for the moment as the merge is a pain to do right and it also makes separating pending stuff for Marcelo much trickier." Alan went on to explain that after further merging with Marcelo, he'll take another look at the newer rmap code. Changelog and full details follow.

Update: 2.4.21-pre3-ac2 quickly followed -ac1 with Alan explaining, "The skb_padto bug is quite ugly so people really want to be using ac2 not ac1".

From: Alan Cox
To: linux-kernel mailing list
Subject: Linux 2.4.21-pre3-ac1
Date: 	Wed, 8 Jan 2003 13:24:27 -0500 (EST)

Its back, its smaller and it wants a lot of testing. Much of the older
stuff has been merged with Marcelo, and some stuff has been dropped out.

+	Handle battery quirk on the Vaio Z600-RE	(Paul Mitcheson)
*	EHCI USB updates				(David Brownell)
+	IDE Raid support for AMI/SI 'Medley' IDE Raid	(Arjan van de Ven)
+	NVIDIA nForce2 IDE PCI identifiers		(Johannes Deisenhofer,
							 Tim Krieglstein)
*	CPU bitmask truncation fix			(Bjorn Helgaas)
o	HP100 cleanup					(Pavel Machek)
o	Fix initial capslock handling on USB keyboard	(Pete Zaitcev)
+	Update dscc4 driver for new wan			(Francois Romieu)
+	Fix boot on Chaintech 4BEA/4BEA-R and		(Alexander Achenbach)
	Gigabyte 9EJL by handing wacky E820 memory
	reporting
o	SysKonnect driver updates			(Mirko Lindner)
o	Fix memory leak in n_hdlc			(Paul Fulghum)
o	Fix missing mtd dependancy			(Herbert Xu)
+	Clean up ide-tape printk stuff			(Pete Zaitcev)
+	IDE tape fixes					(Pete Zaitcev)
o	Fix size reporting of large disks in scsi	(Andries Brouwer)
+	Fix excessive stack usage in NMI handlers	(Mikael Pettersson)
+	Add support for Epson 785EPX USB printer pcmcia	(Khalid Aziz)
*	Quirk handler to sort out IDE compatibility	(Ivan Kokshaysky)
	mishandling
+	Model 1 is valid for PIV in MP table		(Egenera)
+	Ethernet padding fixes for various drivers	(me)
o	Allow trident codec setup to time out		(Ian Soboroff)
	This can happen with non PM codecs
o	Fix broken documentation link			(Henning Meier-Geinitz)
o	Update video4linux docbook			(William Stimson)
o	Correct kmalloc check in dpt_i2o		(Pablo Menichini)
o	Shrink kmap area to required space only		(Manfred Spraul)
o	Fix irq balancing				(Ben LaHaise)
o	CPUfreq updates					(Dominik Brodowski)
o	Fix typo in pmagb fb				(John Bradford)
o	EDD backport					(Matt Domsch)
o	First pass at fixing all the ethernet padding	(me)

REMOVED FOR NOW

-	RMAP

REMOVED FOR GOOD

-	LLC 	(See 2.5)
-	VaryIO  (Never accepted mainstream)
-

I wrote Alan to ask why he had removed the rmap VM, to which he replied:

From: Alan Cox
To: Jeremy Andrews
Subject: Re: Linux 2.4.21-pre3-ac1
Date: Wed, 8 Jan 2003 19:49:33 -0500 (EST)

I took out the 15a rmap because it has problems and is obsolete. I've left
the 15b one for the moment as the merge is a pain to do right and it also
makes separating pending stuff for Marcelo much trickier.

Once I've established if Marcelo wants things like vm overcommit [story] I'll look
at the rmap stuff again.

In another thread on the lkml, Alan further explained:

From: Alan Cox
Subject: Re: Question for Marcelo
Date: 	08 Jan 2003 14:38:17 +0000

On Wed, 2003-01-08 at 13:19, Tomas Szepe wrote:
> > [Alan Cox]
> > 
> > I've also dropped rmap out for now.
> 
> Hmm, what for?

15a wasnt working very well, the base VM isn't too bad now and its
a _lot_ easier to do merging with Marcelo without rmap. The other
related bits are seperated out but present (vm overcommit handling,
fixed shmem, removepage callback)

2.4.21-pre3-ac1 was quickly followed by -ac2:

From: Alan Cox
Subject: Linux 2.4.21pre3-ac2
Date: 	Wed, 8 Jan 2003 20:39:09 -0500 (EST)

The skb_padto bug is quite ugly so people really want to be using ac2 not
ac1. 

Linux 2.4.21pre3-ac2
o	Fix the dumb bug in skb_pad			(Dave Miller)
o	Confirm some sparc bits are wrong and drop them	(Dave Miller)
o	Remove a wrong additional copyright comment	(Dave Miller)
o	Upgrade IPMI driver to v16			(Corey Minyard)
o	Fix 3c523 compile				(Francois Romieu)
o	Handle newer rpm where -ta is rpmbuild not rpm	(me)
o	Driver for Aurora Sio16 PCI adapter series	(Joachim Martillo)
	(SIO8000P, 16000P, and CPCI)
	| Initial merge
o	Backport Hammer 32bit mtrr/nmi changes		(Andi Kleen)
o	Add the fast IRQ path to via 8233/5 audio	(me)

Related Links:

Google Archive Of Above Threads [thread 1] [thread 2]

NVidia driver

Con Kolivas

January 9, 2003 - 7:34am

It also broke the nvidia binary only driver, and the linking module needs to be patched for it to work. Whether you have an objection to using the driver or not, there are those who do want to use it and it adds to the confusion. One of the reasons I dont have rmap by default in -ck*, but I do have it available and the patch for nvidia available if you really really want that combination.

"rmap15a is obsolete"

Anonymous

January 9, 2003 - 4:09pm

I'm not a VM guru.

Anyway I find astonishing he says that... maybe also will be removed from 2.5.x?

"rmap15a is obsolete"

Cuboci

January 9, 2003 - 4:51pm

No, I think he meant it is obsolete, because there is 15b out there already.

anybody can confirm this?

Anonymous

January 9, 2003 - 6:23pm

If the reason what you mention, Why remove it?

idea

Anonymous

January 9, 2003 - 7:28pm

if i read it correctly, my guess is that...

he removed this from his 2.4 tree because his tree is for debugging and testing new features before they go into the mail line...

now this rmap is in 2.5 allready and seems to work there, so ne need to test it in his 2.4 any longer... there is of course no chance to get it ever into 2.4... it is a very big patch as i read it, so it made much problems merging this patch into the 2.4 and for the same reason it will not be appended to 2.4-stable, becasue its no development branch...

so if u want rmap and want testing, go to 2.5 now (for one of my desktops 2.5 is working fine for month now... (sometimes kernel doesnt compile brcause broken stuff, but i can stay at a version until i get some next version to compile...)... else stay were u are or what u like ...

Eugene

rmap in 2.5 != rmap patch for 2.4

Anonymous

January 9, 2003 - 9:43pm

There is an important distinction between the rmap in 2.5 and Rik's full rmap patches for 2.4. The rmap that is in kernel 2.5 contains the minimal rmap feature, atop the "classzone" page replacement routines that Andrea established in the base kernel starting with 2.4.10.

The 2.4 rmap patch not only has Rik's basic rmap feature, but his own version of the page replacement routines that are more similar to that of kernels before 2.4.10 and the notions in FreeBSD than Andrea's classzone approach. The latest iteration (rmap15*) includes Arjan's O(1) page laundering scheme which still needs much testing and profiling.

If rmap is removed from 2.5, it will probably be Andrew Morton's call, since he has overhauled and improved many of the VM internals in 2.5. Thus far, he has at least tolerated it. I'd expect it to stay.

Also

arashi no kaze

January 9, 2003 - 10:52pm

There are a lot of things that have built on top of "rmap" the reverse-mapping bits of the VM. Removing "rmap" from 2.5 would be a really very painful process. I don't think anyone expects it to possibly be ripped out, the expectation has been that it would go in since OLS in 2001.