2.6.30 and amd catalyst :)

Message

kernelOfTruth · Post by **kernelOfTruth** » Wed Jun 17, 2009 2:52 pm

Hi guys,

I've just stumbled over a post at archlinux' AUR and there seems to be a patch for amd/ati catalyst supporting 2.6.30:

http://aur.archlinux.org/packages.php?ID=22899

at the ubuntuforums in from China:

kudos to those / the creator of the patch:

http://forum.ubuntu.org.cn/viewtopic.php?f=42&p=1329973

i92guboj · Post by **i92guboj** » Wed Jun 17, 2009 3:12 pm

kernelOfTruth wrote:Hi guys,

I've just stumbled over a post at archlinux' AUR and there seems to be a patch for amd/ati catalyst supporting 2.6.30:

http://aur.archlinux.org/packages.php?ID=22899

at the ubuntuforums in from China:

kudos to those / the creator of the patch:

http://forum.ubuntu.org.cn/viewtopic.php?f=42&p=1329973

There's an ebuild at bugs.gentoo.org which seems to work ok. Also with xorg-server 1.6

http://bugs.gentoo.org/show_bug.cgi?id=266819

kernelOfTruth · Post by **kernelOfTruth** » Thu Jun 18, 2009 6:07 am

works as expected,

thanks i92guboj !

i92guboj · Post by **i92guboj** » Thu Jun 18, 2009 9:13 am

By the way, this driver spams my syslog and kern.log files to the point that today some services could start normally. After booting I had to delete it and reboot because they were consuming 1gb on my var partition (it was completely full).

I had heared about this problem somewhere in the mailing lists, but I didn't know it was soooo aggressive. So, if you reboot and some services fail to start or you find that some GBs disappeared, look at /var/log

kernelOfTruth · Post by **kernelOfTruth** » Thu Jun 18, 2009 9:20 am

it almost completely disappears if you're not using compositing or direct rendering / opengl apps,

I know this isn't an option for most people (including me) but just in case you mostly want to get rid of it

Link31 · Post by **Link31** » Thu Jun 18, 2009 9:16 pm

Since the driver appears to work correctly on 2.6.30, I'm going to try writing a small patch to remove the dmesg spam... As with everything else on my Gentoo system, when I don't like something I just patch it out, and the kernel or drivers are no exception.

Before I start digging through the drm code, do anyone have an idea on where this message could be caught ? Do you think that a simple "if (strcmp()...)" in the printk() function will do the job, or is there any other way that could be more efficient ?

kernelOfTruth · Post by **kernelOfTruth** » Thu Jun 18, 2009 10:14 pm

Link31 wrote:Since the driver appears to work correctly on 2.6.30, I'm going to try writing a small patch to remove the dmesg spam... As with everything else on my Gentoo system, when I don't like something I just patch it out, and the kernel or drivers are no exception.

Before I start digging through the drm code, do anyone have an idea on where this message could be caught ? Do you think that a simple "if (strcmp()...)" in the printk() function will do the job, or is there any other way that could be more efficient ?

see: http://debianforum.de/forum/viewtopic.php?f=2&p=707443 (you can translate via google)

Code: Select all

filter f_glrx { match("fglrx:"); };
destination ofglrx { file("/var/log/fglrx_crap.log" owner("root") group("adm") perm(0644)); };
log {
        source(src);
        filter(f_glrx);
        destination(ofglrx);
};

maybe you can adjust it to put stuff to /dev/null or anywhere else ...

Link31 · Post by **Link31** » Fri Jun 19, 2009 7:09 am

Yes, sure, but I don't want it to clutter up dmesg. Your config file does only prevent the message from reaching the hard drive, but the kernel dmesg buffer remains cluttered.

kernelOfTruth · Post by **kernelOfTruth** » Fri Jun 19, 2009 8:16 am

Link31 wrote:Yes, sure, but I don't want it to clutter up dmesg. Your config file does only prevent the message from reaching the hard drive, but the kernel dmesg buffer remains cluttered.

it's not from me - it's the only solution I could google up so far

and yes - it's still cluttering dmesg

Link31 · Post by **Link31** » Sat Jun 20, 2009 12:08 am

OK, as promised here is a quick and (very) dirty hack to remove the useless crap that this crappy blob puts in dmesg:

fglrx-dmesg-hack.patch

Code: Select all

--- a/kernel/printk.c	2009-06-20 01:33:36.000000000 +0200
+++ b/kernel/printk.c	2009-06-20 01:57:16.000000000 +0200
@@ -578,6 +578,16 @@ asmlinkage int printk(const char *fmt, .
 	va_list args;
 	int r;
 
+	/* if (strcmp(fmt, "fmt=%s\n") != 0)
+	 	printk("fmt=%s\n", fmt); */
+
+	if (strcmp(fmt, "<3>[fglrx:firegl_find_any_map] *ERROR* Invalid map handle!") == 0
+		|| strstr(fmt, "<3>[fglrx:drm_vm_open] *ERROR* map not found -> inconsistent kernel data!!!") == fmt
+		|| strstr(fmt, "<3>[fglrx:drm_vm_close] *ERROR* map not found -> inconsistent kernel data!!!") == fmt)
+	{
+		return 0;
+	}
+
 	va_start(args, fmt);
 	r = vprintk(fmt, args);
 	va_end(args);

You can use "if (strstr(fmt, "<3>[fglrx:") == fmt)" to get the same effect as the syslog-ng config, but I prefer keeping the other messages for now.

If you can't get some message to hide, just uncomment the block above and reboot to find out what is going on with the format string.

DISCLAIMER: I applied this patch on my own kernel and it works as expected, but I don't know yet if there are any side effects. Use it at your own risk

amissus · Post by **amissus** » Sat Jul 04, 2009 1:01 pm

Driver works, but I have no mouse and keyboard after startx : \

DaggyStyle · Post by **DaggyStyle** » Sat Jul 04, 2009 1:16 pm

kernelOfTruth wrote:it almost completely disappears if you're not using compositing or direct rendering / opengl apps,

I know this isn't an option for most people (including me) but just in case you mostly want to get rid of it

in that case, what is the diff between the precompiled drivers and the open source one? the compositing or direct rendering / opengl apps are the only reason for users not to switch to the open source driver

altrent · Post by **altrent** » Sat Jul 04, 2009 1:24 pm

amissus wrote:Driver works, but I have no mouse and keyboard after startx : \

Re-emerge the ebuilds given by below command:

Code: Select all

qlist -I -C x11-drivers

i92guboj · Post by **i92guboj** » Sat Jul 04, 2009 2:10 pm

DaggyStyle wrote:
kernelOfTruth wrote:it almost completely disappears if you're not using compositing or direct rendering / opengl apps,

I know this isn't an option for most people (including me) but just in case you mostly want to get rid of it
in that case, what is the diff between the precompiled drivers and the open source one? the compositing or direct rendering / opengl apps are the only reason for users not to switch to the open source driver

Nope. In my case neither of radeon nor radeonhd will work with my card.

Well, to be fair they work, until I enable DRI, then my screen is completely garbled to the point of being unusable. That means no 2d acceleration and no xv support, which in turn means they perform worse than vesa.

amissus wrote:Driver works, but I have no mouse and keyboard after startx : \

That's a separate issue. You have hit the upgrade from kbd/mouse to hal/evdev. I can't give you a link right now because I have to go, maybe someone else can guide you, but that's nothing to do with the graphics driver. At least it shouldn't have to do.

amissus · Post by **amissus** » Mon Jul 06, 2009 6:03 pm

I recompiled xorg-server with hal support and run hal daemon and my mouse and keyboard working now

)))
Thanks for clue i92guboj.

Enverex · Post by **Enverex** » Sat Jul 18, 2009 4:14 pm

I'm running into some serious issues.

Versions:
x11-drivers/ati-drivers-9.6
x11-base/xorg-server-1.6.2-r1
Linux defiant 2.6.29-gentoo-r5 #1 SMP Tue Jul 14 20:22:41 BST 2009 x86_64 Intel(R) Core(TM)2 Quad CPU Q9450 @ 2.66GHz GenuineIntel GNU/Linux (tried .30 too but no change)

Basically whenever the kernel/driver spams out that message (it doesn't do it too many times) the machine seems to lock up for about 4 seconds. For example if I open something in Wine then the machine will freeze up quite a few times before the app finally loads. This applies to other apps too i.e. UT2004. Is there a fix for this actual issue? (crap in dmesg itself doesn't really bother me, it's only about 40 lines in total, not hundreds). Plus this lockup is happening whenever I change screenmodes or do anything else of the sort either (and some other desktop operations).

It's making the machine unusable and is -really- annoying.

kernelOfTruth · Post by **kernelOfTruth** » Sat Jul 18, 2009 4:25 pm

http://ati.cchtml.com/

go bug them !

sorry, I've no fix for this problem right now and have stopped using compositing for this very reason (it's not locking up for me but slowing down noticably ...)

amd / ati - you guys make me sad

Enverex · Post by **Enverex** » Sat Jul 18, 2009 4:34 pm

kernelOfTruth wrote:http://ati.cchtml.com/

go bug them !

sorry, I've no fix for this problem right now and have stopped using compositing for this very reason (it's not locking up for me but slowing down noticably ...)

amd / ati - you guys make me sad

I'm not using compositing or anything of that sort (no Compiz or anything like that either). I regret getting an ATi card so much, it was hit and miss in the past but now I seem stuck. Does anyone know of a downgrade route I could follow to get the system back to normal? I'd downgrade the kernel and ATi drivers but I'm not sure if the X ABI change will make that impossible.

DaggyStyle · Post by **DaggyStyle** » Sat Jul 18, 2009 5:30 pm

Enverex wrote:
kernelOfTruth wrote:http://ati.cchtml.com/

go bug them !

sorry, I've no fix for this problem right now and have stopped using compositing for this very reason (it's not locking up for me but slowing down noticably ...)

amd / ati - you guys make me sad
I'm not using compositing or anything of that sort (no Compiz or anything like that either). I regret getting an ATi card so much, it was hit and miss in the past but now I seem stuck. Does anyone know of a downgrade route I could follow to get the system back to normal? I'd downgrade the kernel and ATi drivers but I'm not sure if the X ABI change will make that impossible.

actually, I would stick with ati because there problem is software based which will be solved soon, I won't get nvidia because past experience which included faulty hardware that left their production line knowingly...

2.6.30 and amd catalyst :)

2.6.30 and amd catalyst :)

Re: 2.6.30 and amd catalyst :)