ATI Radeon M6 LY and Others.

[Mike A. Harris]

On Mon, 25 Aug 2003, Jonathan C. Sitte wrote:

>I have been talking with allot of other linux users including people 
>runnning the latest redhat beta and it seems that there is not much 
>support or developement going for these chipsets. Here is the mailing 
>list that most of these talks have been happening so far:
>
>http://staticnull.org/mailman/listinfo/ati-mobililty-linux_staticnull.org
>
>Feel free to put input too on such hardware. We are all working hard to 
>get the word out that getting these cards supported are one of our 
>number one priorities. Anyone at redhat working on this?

The XFree86 project's xfree86 at xfree86.org mailing list is where
you should ask about hardware support questions such as that.  
For 3D related questions, you should use the dri-devel mailing 
list on sourceforge.  Those two lists will contact everyone 
involved in developing and maintaining XFree86 video drivers.

Just to be clear, Red Hat ships XFree86 and thus the hardware 
supported by Red Hat Linux, is whatever is supported by the 
version of XFree86 included in Red Hat Linux.  Some of the 
hardware supported directly by Red Hat to a certain extent, 
which means that problems that occur on certain hardware will 
be investigated directly by me if I have that hardware, and the 
problem that has been reported is both reproduceable and climbs 
up the priority ladder high enough.

Other drivers are merely provided as-is in hopes that they 
happen to work for someone out there.  An example of a "provided" 
driver, is the "apm" driver.  I have no APM hardware, no APM 
specifications, the hardware is very ancient legacy hardware, 
nobody upstream maintains the driver, and nobody has touched it 
in years.  Should someone file a bug report about the APM driver 
not working, there isn't anything I can really do about it.  If 
the APM driver doesn't work at all period, then my solution would 
be to remove the driver from future releases, and if the "vesa" 
driver happens to work on that hardware, I'd point the hardware 
database to use the vesa driver instead.

In general, support for new hardware usually is done by
XFree86.org or a contributing developer in the community.  I have
added support to various drivers such as the ati, r128, radeon,
savage, and some other drivers on occasion and contributed it
back to XFree86 when I've had the information and the time to do
so, but the majority of that has been adding new chip IDs to
existing codepaths, and fine tuning any problems that arose when
people used it.

For hardware that is already supported, but has a lot of problems 
for people, no developer can really do anything about that unless 
they physically own the hardware or have long term access to the 
hardware, and have the complete specifications for that hardware.  
In addition to that, the problems encountered must be considered 
of high enough importance compared to all other tasks that it 
warrants allocating official developer hours to perform the work 
on.

Since there are only 2 X developers present at Red Hat (myself 
and John Dennis), and all XFree86 work being done is done by one 
of us, devoting 4 weeks to fix a single bug on a laptop is more 
or less out of the question I'm afraid.

I have no laptop hardware and no access to any laptop hardware.  
What's more, is that laptop related video problems are often 
model specific, which means that you basically need one of every 
laptop model ever made by every manufacturer.  We're hardly in a 
position to purchase one of every laptop, and it's highly 
unlikely that every laptop manufacturer is going to send us one 
of every laptop model they make.

Even if every manufacturer were to send me one of every laptop
they make, and the video manufacturer was to send me the complete
specifications to every chip they make, I could spend 40 hours a 
week for a year straight and not be able to fix all of the laptop 
related bugs that occur for people.  Laptops are generally very 
unique as far as hardware problems go, which is unfortunate.

In general, laptop specific XFree86 problems are very difficult 
to deal with, and server and desktop systems get priority because 
that is the hardware that is available.  Laptop manufacturers 
generally do not make Linux specific laptops, and they do not 
usually certify their hardware for use on laptops.  In order for 
laptops to truely be properly supported, laptop vendors need to 
support Linux and require certification, and then to pay 
XFree86.org or whoever to make sure their hardware is supported 
well.  Until that happens, the majority of support for laptops is 
basically whatever the community can contribute to XFree86.org.  
Unfortunately, no XFree86 developer out there has access to the 
specific model of laptop that one might file a bug against, and 
no OS vendor does either.

Laptops are more or less a "hope it works" thing as far as Linux 
is concerned, and that wont really change until the manufacturers 
of those machines want full official Linux support, and are 
willing to pay for it.

I realize that that isn't the answer people like to hear, but I'm 
just being realistic.  The best thing people can do, is 
telephone or email their hardware vendors and tell them politely 
that Linux support is very important to them.

But to answer your question:

>number one priorities. Anyone at redhat working on this?

Nobody at Red Hat is working on XFree86 laptop support in any
official capacity, no.  If a laptop related problem gets fixed in
upstream XFree86 CVS, and I'm aware of it (I monitor X CVS
closely), and someone has reported a bug about that to us (and to
XFree86.org at http://bugs.xfree86.org) then I usually try to
backport whatever fixes are available to maximize our hardware
support.  I also will examine source code when possible depending 
on the nature of a bug that gets reported.  Sometimes even if I 
don't have the hardware, but the problem is clear enough, I might 
be able to spot a problem in the code or brainstorm something.  

The overwhelming majority of problems however require direct 
physical access to the specific hardware having the problem, the 
specifications for that hardware, and the ability to easily 
reproduce it, along with the problem being large enough to have 
it prioritized for time allotment.

In general, people should report bugs directly to XFree86.org, as
there are about 30 or more X developers reading their bugzilla
(including me), and 2 reading our bugzilla.  It doesn't hurt to 
report things in both places either.

The truth of the matter is that we are not a video driver
development house, although we certainly do contribute as much as 
possible to XFree86 development, but that can't of course solve 
all problems for all people.

If anyone is interested, the XFree86 and DRI projects are always 
looking for more volunteers.



-- 
Mike A. Harris     ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat