VIA Releases Linux Driver Source Packages [LWN.net]

From:		patmeier-AT-patmeier.com
To:		lwn-AT-lwn.net
Subject:		VIA Releases Linux Driver Source Packages
Date:		Tue, 12 Apr 2005 06:03:55 EDT

The following release announces the new VIA Linux Driver Source Packages,
which are freely available and suited for tailor-made development of the
Linux distributions and can be downloaded easily from VIA Arena,
http://www.viaarena.com.

For Immediate Release 

VIA Releases Linux Driver Source Packages 

Freely available driver source code eases the development of tailor-made
applications for the latest Linux distributions

Taipei, Taiwan, 12 April 2005 - VIA Technologies, Inc, a leading innovator
and developer of silicon chip technologies and PC platform solutions, today
announced the extension of its open source support with the release of driver
source codes for specific VIA hardware.

Of particular interest to the growing number of Linux enthusiasts and
customers utilizing VIA EPIA mainboards, is the source for the S3 Graphics
UniChrome family display driver for version 2.6.x kernels. The UniChrome
family display driver supports the VIA CLE266 and the new VIA CN400 Digital
Media chipsets featured on the popular VIA EPIA M series and VIA EPIA SP
series mainboards respectively, and will provide developers with the
flexibility to autonomously incorporate support for latest applications based
on VIA hardware.

"With this approach to driver support we are broadening our contribution to
the open source community, bringing a valuable software element to our total
silicon platforms," commented Steve Charng, Head of Software Development, VIA
Technologies, Inc. "By making these source codes freely available, VIA is
expanding our software focus and enabling tailor-made solutions for the
latest Linux distributions."

On top of the UniChrome drivers, VIA also offers display driver sources for
the VIA ProSavage and ProSavage DDR integrated graphics controllers as well
as the integrated Network driver source supporting six VIA Chipset South
Bridges.

The latest source codes are available for download at VIA Arena, the premier
technical support and information resource for VIA products, at
www.viaarena.com

About VIA Technologies, Inc.

VIA Technologies, Inc. (TSE 2388) is the foremost fabless supplier of
market-leading core logic chipsets, low power x86 processors, advanced
connectivity, multimedia, networking and storage silicon, and complete
platform solutions that are driving system innovation in the PC and embedded
markets. Headquartered in Taipei, Taiwan, VIA's global network links the high
tech centers of the US, Europe and Asia, and its customer base includes the
world's top OEMs, motherboard vendors and system integrators. www.via.com.tw

VIA PR Contacts 

International:  Richard Brown   
Phone:  (886)-2-2218-5452 #6201 
Fax:    (886)-2-2218-5453       
Email:  RIBrown@via.com.tw      

North America:  Pat Meier-Johnson
Pat Meier Associates Public Relations
Phone:  (415) 389-1700
Email:  patmeier@patmeier.com

Note to reporters, editors and writers: VIA is written in ALL CAPS.

(Log in to post comments)

VIA Releases Linux Driver Source Packages

Posted Apr 12, 2005 19:44 UTC (Tue) by XERC (guest, #14626) [Link]

I don't know, how did people get the VIA chpset to work before, but now I can be certain that IF I will upgrade, it will be VIA.

Intereasting, this just sounds too good to be true, are they sic? May be it's just one small exception?

VIA Releases Linux Driver Source Packages

Posted Apr 13, 2005 0:23 UTC (Wed) by lakeland (subscriber, #1157) [Link]

Via used to work before as a standard PC, although people did experience
more odd lockups than is normal.

Why this is particularly important is that the via EPIA chip at 800MHz was
no race winner. Clock for clock, I'd bet the celeron would outperform it.
However, the via had two important chips on its MB to make up for this.
An MPEG decoder chip and a random number generator. Neither of these
chips were usable from linux, leaving the machine a little sorry.
Further, despite via being very aware that almsot all of its customers
were running linux, they weren't releasing specs for the chips.

If they are now doing so, that's a very nice bonus. Coupled with ivtv for
mpeg encoding on a hauppauge board, you should be able to build a fanless
PVR box now.

Are you sure you know what you're talking about?

Posted Apr 13, 2005 14:22 UTC (Wed) by jensend (guest, #1385) [Link]

The "more odd lockups than normal" would probably be due to running i686 binaries on processors which don't support CMOV. That's not a problem either if you use binaries compiled for i586 or c3 or if you have a Nehemiah or newer C3 processor. The hardware RNG and crypto acceleration (which are on the processor, not a separate chip) have been well supported in Linux and several BSDs for well over a year. It's true that support for the MPEG encoding/decoding has been far from perfect, but according to what I hear this source release does little or nothing to fix that.

Are you sure you know what you're talking about?

Posted Apr 14, 2005 4:30 UTC (Thu) by lakeland (subscriber, #1157) [Link]

Well, I thought so, but if I'm wrong then I'm wrong :-)

I'd looked into it about a year ago when I was trying to decide what
hardware would be best to base a PVR on. The via had appealed because the
only hauppauge card you could semi-affortably buy here is the 250 which
does not support hardware decoding but at that time forum posts said it
wasn't supported.

Since then a work colleague bought one for controlling an aeroplane and
has had a few hassles with it, but he doesn't do any crypto or MPEG.

VIA Releases Linux Driver Source Packages

Posted Apr 14, 2005 12:04 UTC (Thu) by hmh (subscriber, #3838) [Link]

The random number generator was not useable?! All the specs and programming information were out in the open from day one, AFAIK. And it is accessible from userland, so it is not like anyone needed to wait for a kernel hacker to use it...

Although VIA does document that the HRNG in some older CPUs (the first three steppings, I think) is unuseable. Maybe that is what you are refering to?

Try the development branch of rng-tools at http://arch.debian.org/arch/private/hmh/rng-tools-2005/ (tla archive) for VIA PadLock HRNG speeds on the megabit per second range. If you like it, send word by the way of the nice guys at mekensleep.com which provided me with root access to a Nehemiah CPU to debug it...

VIA Releases Linux Driver Source Packages

Posted Apr 12, 2005 19:10 UTC (Tue) by ber (subscriber, #2142) [Link]

Examining CN-CLE-PMXF40056-kernel-src_20050124.tgz from the website, it even seems to be Free Software. A toplevel statement seems to be missing, but I looked at a few files and many say:

* Copyright 1998-2005 VIA Technologies, Inc. All Rights Reserved.
 * Copyright 2001-2005 S3 Graphics, Inc. All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sub license,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
 * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.

VIA Releases Linux Driver Source Packages

Posted Apr 13, 2005 11:11 UTC (Wed) by ber (subscriber, #2142) [Link]

There is CN-CLE-PMXF40056-kernel-src_20050124/src/via/libddmpeg.so though,
so I guess this needs more eximination before it can be considered as Free Software.

VIA Releases Linux Driver Source Packages

Posted Apr 12, 2005 19:35 UTC (Tue) by lacostej (guest, #2760) [Link]

This is a great news. As a VIA user, quite frankly fed up with their way of working, I had been in contact with VIA for a little bit more than a couple of months and tried to evangelize them with regard to a better interaction with the FOSS community. I received an early news some weeks ago that there was going to be some change and that my arguments had helped. I should have blogged about it :)
Unfortunately I don't know exactly the impact my emails had. I will try to get some feedback on that. If they didn't start this initiative before I contacted them, or if my arguments helped a lot in the balance, that's great news: that means anyone with a little bit of time, the right person on the other side of the communication channel and the right arguments can make things change.

Note: that's not the only thing that changed. They also allocated more people to directly communicate with the community on the forums (viaarena.com). And from what I understood in my communications with them, there could be more to come. But I prefer to wait and see :)

VIA Releases Linux Driver Source Packages

Posted Apr 13, 2005 2:15 UTC (Wed) by beoba (guest, #16942) [Link]

What sorts of arguments did you use, for those of us who plan on pestering other manufacturers?

Is this an actual open source 3D driver?

Posted Apr 12, 2005 22:15 UTC (Tue) by mattdm (subscriber, #18) [Link]

From reviews and comments I've seen, I'd expect it to be on par with Intel's integrated graphics chipset, which already has an open source 3D driver. This is still a really good thing, but it's not going to shake up Nvidia and ATI.

Re: Is this an actual open source 3D driver?

Posted Apr 13, 2005 10:09 UTC (Wed) by tialaramex (subscriber, #21167) [Link]

DRI supports Intel's onboard chipsets, with AFAIK support from Intel engineers

DRI has in the past supported some VIA chipsets (presumably not these ones)

DRI also has support for various ATI chipsets still used in laptops (but admittedly not the latest M10 and above, not yet)

So the discerning Free Software buyer already has options for 3D acceleration in a brand new laptop, it's just that the two most reputable 3D hardware companies (ATI and nVidia) don't offer us any support. If you need very high performance (e.g. want to play Doom 3 on a laptop) then that's a serious problem, but many people's 3D use is within the capability of the r200 or even i8xx drivers.

Totally unofficial reverse engineered r300 drivers are under development for DRI, these seem to be capable of working with ATI's latest released hardware (probably not the new chips being released this quarter though). It would be nice to see more people helping out on these drivers, because R300 era hardware is very powerful, the M10 laptop chipset should be supported for example.

http://r300.sf.net/

It's probably also possible to port the Utah GLX code to the DRI design, allowing modern systems to use nVidia hardware up to the GF4 MX class. It seems likely that mid-range laptops might still have these chips. Utah-GLX had enough of this hardware working to play Quake 2 at least.

What...? - consider the Open Graphics Project

Posted Apr 13, 2005 3:49 UTC (Wed) by amarsh04 (guest, #29240) [Link]

You should probably check out the Open Graphics Project mailing list.

Techsource are working on a graphics card with 3D support, fully open specifications and drivers initially available using a FPGA, then when the hardware is debugged, release a faster and cheaper ASIC-based model.

Where are they?

Posted Apr 13, 2005 3:49 UTC (Wed) by jwb (guest, #15467) [Link]

I just pulled the UniChrome driver for 2.6 from the viaarena website*, and it's dated Feb 3, 2005.
So it seems this is recent and supports recent kernels. Unfortunately the source if very
disintegrated from both Xorg and the kernel. It has its own directory tree and build system, with
only the most bare documentation. I suppose the enthusiasts of this hardware will integrate the
code into an actual project someday.

*How do you make a website this poor? Does it take actual concentrated effort, or do websites run
amok if you don't pay enough attention? Good luck finding anything.

Where are they?

Posted Apr 29, 2005 22:21 UTC (Fri) by smoku (guest, #29649) [Link]

It does not even compile with the current kernel.
I've had a real hard time to get it to work (it's my first time with kernel sources).

If anyone is interested you can get the 2.6.11 ported driver from http://staff.xiaoka.com/smoku/stuff/VIA/CN-CLE-PMXF40056-...

This is news?!

Posted Apr 13, 2005 7:33 UTC (Wed) by fooker (guest, #14834) [Link]

Well, off the top of my head, Unichrome website says that "K8M800 and PM800/CN400 support is still limited". The CLE266 driver was the only open source driver VIA released until now. This of course means that Unichrome doesn't support most of the features of the newer chipsets (MPEG-4 acceleration in CN400, for example) very well or at all.

But after the driver release I'd expect that Unichrome will support all the features and chipsets very soon :). Now if only VIA would adopt the Unichrome project instead of producing their own source releases.

On a related sidenote, how long will it take for companies to realise that the best way to support their hardware on all the Linux distributions is to get their open source driver into the main kernel tree?

Nothing really new under the sun...

Posted Apr 13, 2005 12:10 UTC (Wed) by totte (guest, #29252) [Link]

I have some trouble understanding all the fuzz.

The now released source has more or less been available through VIA's developer's portal for a year or so.

Also, as stated in one of the previous posts, VIA's drivers have some features that the truly open source unichrome / Xorg / Mesa drivers don't have, but here are some facts.

1. VIA's mpeg support remains closed source.
2. VIA's HW mpeg4 info is not even available under NDA, and is
definitively not open source.
3. VIA's 3D drivers are slow and buggy compared to the open source version in Mesa. They also contain a severe security hole.
4. VIA's mpeg drivers are runnable only as root. This combined with known security holes in the applications that can use them makes it able for an attacker to create an url that would make your system unusable in an instant at best. Worse gain a lot of secret information.
5. The community open source drivers outperforms VIA's driver in raw speed both in terms of Xv, 2D accel, 3D accel and mpeg2 accel. Unverified statements say that they work faster with sw mpeg4 decoding than VIA's drivers do with hw mpeg4 acceleration.
6. The community drivers work with standard versions of xine, mplayer and mythtv. VIA has forked xine and mplayer to create support for their drivers and do not keep the forks up to date, not even with security fixes.
7. VIA has had every opportunity to backport the OS development.

What's all the fuzz about? Get the facts!

http://unichrome.sourceforge.net
http://www.mesa3d.org
http://www.x.org

/Thomas