Re: [suse-xfree86] Re: ATI Radeon 9600 driver problem
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages: May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1. This comes no matter wether I try sytsem agpgart or fglrx internal agpart. I tried the agp_try_unsupported=1 parameter both with agpgart and fglrx but they compain about not knowing that parameter. BTW, shouldn't we keep these mails on suse-xfree86@suse.com? P.J. -- ---------------------------- Peter-Josef Meisch Mail: pjmeisch@web.de ----------------------------
On Fri, May 14, 2004 at 05:09:48PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages:
May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1.
Then please unload amd-k7-apg module (if possible, otherwise reboot) and try the other agp module (IIRC via-agp). Send us the output of "lspci -n" as well.
This comes no matter wether I try sytsem agpgart or fglrx internal agpart. I tried the agp_try_unsupported=1 parameter both with agpgart and fglrx but they compain about not knowing that parameter.
Yes, this parameter is not supported any longer as the agp support is splitted now in different agp modules.
BTW, shouldn't we keep these mails on suse-xfree86@suse.com?
Yes. Stefan Public Key available ---------------------------------------------------- Stefan Dirsch (Res. & Dev.) SUSE LINUX AG Tel: 0911-740 53 0 Maxfeldstrasse 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ----------------------------------------------------
Am Freitag, 14. Mai 2004 17:52 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 05:09:48PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages:
May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1.
Then please unload amd-k7-apg module (if possible, otherwise reboot) and try the other agp module (IIRC via-agp). Send us the output of "lspci -n" as well.
I can rmmod and modprobe the modules without rebooting because they are not in use. Using via_agp has the same effect: # testgart open: No such device #lspci -n 0000:00:00.0 Class 0600: 1106:3188 (rev 01) 0000:00:01.0 Class 0604: 1106:b188 0000:00:0b.0 Class 0200: 1317:0985 (rev 11) 0000:00:0d.0 Class 0c00: 11c1:5811 (rev 61) 0000:00:0f.0 Class 0101: 1106:3149 (rev 80) 0000:00:0f.1 Class 0101: 1106:0571 (rev 06) 0000:00:10.0 Class 0c03: 1106:3038 (rev 81) 0000:00:10.1 Class 0c03: 1106:3038 (rev 81) 0000:00:10.2 Class 0c03: 1106:3038 (rev 81) 0000:00:10.3 Class 0c03: 1106:3038 (rev 81) 0000:00:10.4 Class 0c03: 1106:3104 (rev 86) 0000:00:11.0 Class 0601: 1106:3227 0000:00:11.5 Class 0401: 1106:3059 (rev 60) 0000:00:18.0 Class 0600: 1022:1100 0000:00:18.1 Class 0600: 1022:1101 0000:00:18.2 Class 0600: 1022:1102 0000:00:18.3 Class 0600: 1022:1103 0000:01:00.0 Class 0300: 1002:4152 0000:01:00.1 Class 0380: 1002:4172 Greetings P.J. -- ---------------------------- Peter-Josef Meisch Mail: pjmeisch@web.de ----------------------------
On Fri, May 14, 2004 at 06:07:52PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 17:52 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 05:09:48PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages:
May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1.
Then please unload amd-k7-apg module (if possible, otherwise reboot) and try the other agp module (IIRC via-agp). Send us the output of "lspci -n" as well.
I can rmmod and modprobe the modules without rebooting because they are not in use. Using via_agp has the same effect:
# testgart open: No such device
#lspci -n 0000:00:00.0 Class 0600: 1106:3188 (rev 01)
/usr/src/linux/include/linux/pci_ids.h: #define PCI_DEVICE_ID_VIA_8385_0 0x3188 Seems to be supported by the amd64-agp driver. drivers/char/agp> grep PCI_DEVICE_ID_VIA_8385_0 * amd64-agp.c: .device = PCI_DEVICE_ID_VIA_8385_0, But also the via agp module is registered for this device, simply for every VIA host bridge. :-( /lib/modules/2.6.5-12.1-default/modules.pcimap: amd64-agp 0x00001106 0x00003188 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 via-agp 0x00001106 0xffffffff 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 ^^^^^^^^ Removing the sis-agp module completely should work for you. Then PCI hotplug loads automatically the right module (amd64-agp). I need to file a bugreport for this. We have more conflicts than this (amd64/nvidia, amd64-agp/via, amd64/amd-k7). :-( Stefan Public Key available ---------------------------------------------------- Stefan Dirsch (Res. & Dev.) SUSE LINUX AG Tel: 0911-740 53 0 Maxfeldstrasse 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ----------------------------------------------------
Am Freitag, 14. Mai 2004 18:43 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 06:07:52PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 17:52 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 05:09:48PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages:
May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1.
Then please unload amd-k7-apg module (if possible, otherwise reboot) and try the other agp module (IIRC via-agp). Send us the output of "lspci -n" as well.
I can rmmod and modprobe the modules without rebooting because they are not in use. Using via_agp has the same effect:
# testgart open: No such device
#lspci -n 0000:00:00.0 Class 0600: 1106:3188 (rev 01)
/usr/src/linux/include/linux/pci_ids.h: #define PCI_DEVICE_ID_VIA_8385_0 0x3188
Seems to be supported by the amd64-agp driver.
drivers/char/agp> grep PCI_DEVICE_ID_VIA_8385_0 * amd64-agp.c: .device = PCI_DEVICE_ID_VIA_8385_0,
But also the via agp module is registered for this device, simply for every VIA host bridge. :-(
/lib/modules/2.6.5-12.1-default/modules.pcimap: amd64-agp 0x00001106 0x00003188 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 via-agp 0x00001106 0xffffffff 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 ^^^^^^^^
Removing the sis-agp module completely should work for you. Then PCI hotplug loads automatically the right module (amd64-agp).
THANK YOU!!!! What I did now: I rebooted into runlevel 3, #lsmod|grep agp showed amd_k7_agp, via_agp and agpgart as usual. I then modprobed amd64_agp without unloading any of the other modules, after that testgart worked. I now have in my XF86Config : Option "UseInternalAGPGART" "yes" Starting X gives me #fglrxinfo display: :0.0 screen: 0 OpenGL vendor string: ATI Technologies Inc. OpenGL renderer string: RADEON 9600 XT Generic OpenGL version string: 1.3 (X4.3.0-3.7.6) and glxgears: 11019 frames in 5.0 seconds = 2203.800 FPS 11462 frames in 5.0 seconds = 2292.400 FPS 11554 frames in 5.0 seconds = 2310.800 FPS So it should be enough modprobing amd64_agp in boot.local? Or can I put it in modprobe.conf (I'm not too familiar with that one) as a dependancy for fglrx? Anyway, thanks again for your help! P.J.
I need to file a bugreport for this. We have more conflicts than this (amd64/nvidia, amd64-agp/via, amd64/amd-k7). :-(
Stefan
Public Key available ---------------------------------------------------- Stefan Dirsch (Res. & Dev.) SUSE LINUX AG Tel: 0911-740 53 0 Maxfeldstrasse 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ----------------------------------------------------
-- ---------------------------- Peter-Josef Meisch Mail: pjmeisch@web.de ----------------------------
On Fri, May 14, 2004 at 07:04:36PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 18:43 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 06:07:52PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 17:52 schrieb Stefan Dirsch:
On Fri, May 14, 2004 at 05:09:48PM +0200, Peter-Josef Meisch wrote:
Am Freitag, 14. Mai 2004 16:41 schrieben Sie:
Sounds strange. IMHO this is one chipset specific agp module to much. You might try to disable PCI hotplug (remove "06" in COLDPLUG_PCI_CLASSES_WHITELIST in /etc/sysconfig/hotplug) , reboot and load one of the agp modules manually and see if that helps to run testgart. Most likely the amd-k7-agp module is the right one.
I tried that, after booting in runlevel 3 there were no agp moduls loaded. I modprobe amd_k7_agp but testgart and X failed. What I noticed is an error message in /var/log/messages:
May 14 17:02:04 gruyere kernel: agpgart: Unsupported Via chipset (device id: 3188), you might want to try agp_try_unsupported=1.
Then please unload amd-k7-apg module (if possible, otherwise reboot) and try the other agp module (IIRC via-agp). Send us the output of "lspci -n" as well.
I can rmmod and modprobe the modules without rebooting because they are not in use. Using via_agp has the same effect:
# testgart open: No such device
#lspci -n 0000:00:00.0 Class 0600: 1106:3188 (rev 01)
/usr/src/linux/include/linux/pci_ids.h: #define PCI_DEVICE_ID_VIA_8385_0 0x3188
Seems to be supported by the amd64-agp driver.
drivers/char/agp> grep PCI_DEVICE_ID_VIA_8385_0 * amd64-agp.c: .device = PCI_DEVICE_ID_VIA_8385_0,
But also the via agp module is registered for this device, simply for every VIA host bridge. :-(
/lib/modules/2.6.5-12.1-default/modules.pcimap: amd64-agp 0x00001106 0x00003188 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 via-agp 0x00001106 0xffffffff 0xffffffff 0xffffffff 0x00060000 0xffffffff 0x0 ^^^^^^^^
Removing the sis-agp module completely should work for you. Then PCI hotplug loads automatically the right module (amd64-agp).
THANK YOU!!!!
What I did now:
I rebooted into runlevel 3, #lsmod|grep agp showed amd_k7_agp, via_agp and agpgart as usual.
I then modprobed amd64_agp without unloading any of the other modules, after that testgart worked.
I now have in my XF86Config : Option "UseInternalAGPGART" "yes"
Starting X gives me #fglrxinfo display: :0.0 screen: 0 OpenGL vendor string: ATI Technologies Inc. OpenGL renderer string: RADEON 9600 XT Generic OpenGL version string: 1.3 (X4.3.0-3.7.6)
and glxgears: 11019 frames in 5.0 seconds = 2203.800 FPS 11462 frames in 5.0 seconds = 2292.400 FPS 11554 frames in 5.0 seconds = 2310.800 FPS
So it should be enough modprobing amd64_agp in boot.local?
Yes, this should work. Stefan Public Key available ---------------------------------------------------- Stefan Dirsch (Res. & Dev.) SUSE LINUX AG Tel: 0911-740 53 0 Maxfeldstrasse 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ----------------------------------------------------
participants (2)
-
Peter-Josef Meisch -
Stefan Dirsch