Bug#454218: xserver-xorg-video-ati: xserver crashes (freezes the whole system) with DRI enabled on ATI 9250

Brice Goglin Brice.Goglin at ens-lyon.org
Tue Dec 4 01:47:12 PST 2007

Rui Barreiros wrote:
> Look i was already expecting this, why ? because debian ppl tend to 
> just discard the stuff if it's not precisely the way they want with 
> all dot's and commas just because it's more 'comfortable' that way and 
> doesn't give you any work to solve your problems.

How do you want us to debug something if you don't even send the config 
and log corresponding to the problem? I am sorry but we know better than 
you what we need to debug a problem and forward it to the upstream dev 
if needed.

> I'm going to repeat myself in a very detailed explanation about 
> everything i did in 3 days of useless time trying to get this working, 
> i'm not going to go through another 3 days of useless

I don't really need all this. I just need the config and log from ati 
6.7.196 currently in experimental. If I need more, I will ask you.

> (i think this is pretty informative enough without the need to paste a 
> xorg.conf, but if required i can make several xorg.conf used in each 
> test i did)

No, we want to see the full xorg.conf anyway because:
* it's easier for us to read a full xorg.conf, we know where to look, 
while reading 10 lines of text takes much more time
* even if you are sure of your xorg.conf, you could still have placed 
one option in the wrong section, we want to see it

> startx -> crash
> System rendered useless, i always have 1 SSH from a remote PC to this 
> PC, all system lockdown only thing i could do was a reset.

Note that this is a lockup then, not a crash, it is highly different.

> The only difference i had about system crashing and X using 100% cpu 
> (but couldn't be killed in my SSH session) was when using BusType PCI

Try to attach a gdb and see where X is stuck.

And we need the exact config and log for this setup anyway.

> Options i tried, all option was tested alone, with all other values 
> being the ones default (i.e. either commented or not present in 
> xorg.conf)
> Example, when tested AGPMode "2" there was no other option in 
> xorg.conf thus assuming it's default value.

Option BusType PCI is kind of the biggest solution. If it doesn't work, 
there are very few chances that AGPMode would help.

> AccelMethod (both EXA and XAA)
> GARTSize (always lower or equal to the value specified in BIOS, from 8 
> to 256)
> AGPFastWrite - tried the 4 examples
>  - BIOS on, xorg.conf on
>  - BIOS off, xorg.conf on
>  - BIOS on, xorg.conf off
>  - BIOS off, xorg.conf off
> EnablePageFlip (0|1)
> ColorTiling (0|1)

You should just drop ALL of these options for now, all they can do is 
break things even more.

> Brice, if i was a guy that just threw in the bugg without any 
> explanation of what/why/when/how i would be the first guy kicking this 
> bug report to the garbage bin, but i explain, in the bug report if 
> your had read it ALL, you would see my remarks about ABI in the log, i 
> don't think my english is so terrible that you couldn't understand.
> One thing i hate in bug reports specially in 2 ML's (lkml and some 
> Debian ppl) is the attitude, that attitude is not one that should be 
> from a person who wishes to contribute and help users of free software.

Please understand that we have hundreds of bug reports to deal with. If 
you don't send us the required information, we just can't do anything 
but ask for the info that we need! And if the bug report is messy like 
yours, we spend a lot of time gathering the important information. 
That's why closing a messy bug report and opening a new clean one is 
MUCH better for us. And also better for you since you want the bug to be 
fixed faster. So please do what I said before (reinstall a clean 
6.7.196, clean your xorg.conf and send it with the corresponding log).

By the way, there is 6.7.196-2 in experimental with the latest upstream 
git master included as of yesterday.

Also, bug #444049 might be related.

Thank you,

More information about the xorg-driver-ati mailing list