Security update carried along a broken i810 driver for Intel 830 and up

[Jim Cornette]

Alan Hourihane wrote:

>On Tue, 2005-09-20 at 18:28 -0400, Jim Cornette wrote:
>  
>
>>Kristian Høgsberg wrote:
>>
>>    
>>
>>>Jim Cornette wrote:
>>>...
>>>
>>>      
>>>
>>>>Are there any plans to seperate the driver into at least three driver 
>>>>versions to allow these video cards to find a safe zone instead of 
>>>>changes for the 830 breaking the 815 and such?
>>>>
>>>>Just an idea, since I have the low end and middle road Intel video 
>>>>cards .
>>>>        
>>>>
>>>The issue here isn't that the different Intel chipsets are supported 
>>>using one unified driver.  The problem is that we shipped a Fedora 
>>>update which contained a security fix and a number of other fixes.  
>>>One of these other fixes happened to introduce a regression on all 
>>>Intel boards, but the problem was in the general linux pci code and 
>>>would have affected all Intel chipsets even if the driver was split 
>>>into three pieces.
>>>
>>>A new update should be out soon, and in the meantime I've put up i386 
>>>RPMs here:
>>>
>>>    http://people.redhat.com/krh/xorg-x11
>>>
>>>for people to test.  See also:
>>>
>>>    https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=168750
>>>    https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=168752
>>>
>>>Thanks,
>>>Kristian
>>>      
>>>
>>The computer with the Intel 815 worked with the PCI coding problem. The 
>>security update and the fix for the Intel driver was successful with the 
>>Intel 865G. Thanks for the quick resolution to the problem.
>>
>>With the security update out of the way and the pci code cleaned up a 
>>bit. It still makes sense to split up the driver into 815, 830 and 915 
>>to allow code to be taylored for each card that has different 
>>capabilities. It also makes sense since older generations of Intel video 
>>cards won't burden the advancement of next generation cards by needing 
>>code carried on to maintain workability for the older chipsets.
>>
>>When the code was broken for the 810/815 portion of the driver, my 865G 
>>video card started working and the random lockups ceased for the 865G. 
>>Fortunately, when the code for the 810/815 was fixed, the 865G still 
>>remained lockup free.
>>
>>With the latest breakage for I830 and above video cards while the Intel 
>>815 worked in my tests, the thought of splitting off the driver for low, 
>>medium and higher capability seperate drivers came to mind again.
>>
>>Also, when the i810 driver was broken for the Intel 815, I chanced 
>>across a website that provided seperate drivers for the i810/815 and 
>>another for the 830. I believe there was a seperate driver for the Intel 
>>915 on the snapshotted drivers page.
>>
>>The driver on that site worked to bypass the problem. I can't remember 
>>if it was the refresh or the total breakage of the driver. I believe a 
>>change was made in the driver code for dual head with the 830 card at 
>>the time the driver broke for the Intel 815 video my computer has as 
>>builtin video.
>>
>>Just a though since I feel the scheme would be better.
>>    
>>
>
>Jim,
>
>Thanks for the comments, but if you've ever looked at the driver code
>it's already logically split even though it appears in the same i810
>driver package.
>
>And due to the similarities between i830 and i915 based chipsets it
>doesn't make sense to break it out at that level.
>
>Additionally, changes made for i830/i915 should not break anything with
>the i810/i815 due to the logical splitting that is already in-place. So
>I'd be keen to hear (or point me in a direction) of a bug that you think
>instigated this behaviour.
>
>Alan.
>_______________________________________________
>xorg mailing list
>xorg at lists.freedesktop.org
>http://lists.freedesktop.org/mailman/listinfo/xorg
>
>  
>
History of bugs encountered.
Inetl 865G would randomly lockup. The Intel 815 video card worked. A 
change was made for the 830 card referencing the below excerpt from Xorg 
bug 1084.

AlanH's changes for the dualhead i830 support breaks the i810 driver..

for a start it now requires that the pI810->LpRing is xalloc'ed but this doesn't
happen on the i810 driver, only in the i830_driver.c, I think also some of the
cursor mem stuff probably also needs to be changed for the i810/5...

I've cc'ed Alan I'm not sure if it is his work or Keiths.. Alan can you fixup
the i810/5 (do you have the hardware?) I've got some here but I'm not that great
with the 2D drivers yet...


When the change took place, the random lockups with the 865G ceased 
while the Intel 815 would not start.
A patch was submitted to get the Intel 815 card working again, the 
change introduced a refresh problem that was longstanding. Before the 
Intel 810/815 cards were busted, I never experienced any problems with 
any graphical server.

For the refresh problem after the 810/815 driver, xorg bug # 1817 is a 
fair bug to reference.

Both bugs were fixed, but seemed to be caused trying to make the Intel 
830 happy and massage it to keep the 810/815 chipsets working again.

I'm always afraid that making changes to the driver for one chipset will 
cause the videocard at the other end of the spectrum to stop working.

I just see personalities between the different video cards to be worthy 
of having seperate drivers.

With the recent xorg-x11 security version that Fedora released, it broke 
a lot of videocards, mostly Intel 830 or better. The Intel 815 card that 
I have was not effected by the pci related error that broke the other 
cards. I did not see any bug reports referring to 810/815 video cards 
either.

I guess if the difficulty between maintaining a "one driver fits all" or 
maintaining a "goldilocks " version for the different performance 
versions of Intel video chipsets.

Thanks,

Jim

-- 
QOTD:
	If it's too loud, you're too old.