Proprietary drivers

[Lennart Sorensen]

On Mon, Jun 29, 2009 at 11:29:20PM -0400, Steve Harvey wrote:
>   I have a lenovo ThinkPad with ATI X1300 video.  The latest
> proprietary driver that I can download from ATI's support site 
> for it is the Catalyst 9.3 driver.  Under Debian Lenny (2.6.26
> kernel), it often generates kernel tracebacks when I suspend
> to RAM.  Most of the time it is noticable only in the logfiles
> but occasionally the system is practically unusable until
> Xorg is killed and the driver reloaded, which defeats the 
> purpose of using s2ram.

Do the 8-12 drivers that Lenny comes with not support the X1300?

>   Now, if I use the open source driver instead, it wants to
> set the brightness at the medium setting (which for ThinkPads
> is too often too dim).  The only quick way to adjust the brightness
> that I've found so far is to switch to a text console, do it 
> there, and switch back.  However, about 1 in 20 times, the box 
> just locks up when trying to switch back, necessitating a reboot.  
> Applications such as Google Earth are also much slower.
> 
>   ATI has effectively orphaned my video card, despite the 
> laptop being still under warranty.  The Catalyst driver is now
> at version 9.6, but a number of hardware models including my
> X1300 "have been moved to the legacy software support 
> structure". 
>   
>   I've also got openSuSE and an ancient Ubuntu on the laptop
> and I'd really like to stick with Debian for now but these
> driver problems are too disruptive.

Well nvidia at least maintains legacy drivers for old cards.  No idea if
ATI does the same or not.  If my past experience with them is anything
to go by, then they won't, because they really don't care much about
anything but current customers.  I hope they change that at some point.

>   Another proprietary driver that has recently caused me
> problems is a Linux driver for a partial network stack from a 
> well-known networking company.  It used the MMX or SSE
> instructions to do it's crypto, but under sufficiently new kernels
> didn't ensure at interrupt time that it was using the proper 
> context, ergo corrupting MMX or floating point operations
> in some random process and probably causing random
> performance stalls in it's communications as well.  Again
> at the mercy of a cathedral.

MMX and floating point is not permitted in kernel space on linux.
Any attempt to do so is bound to get into trouble.

>   Licenses for these drivers prohibit reverse engineering
> and while I can appreciate these companies wanting to
> protect their investments in IP, there is a problem when
> this interferes with interoperability.

AMD/ATI claims to be releasing specs, but I don't know for which chips
they are going to do it.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists