X won't start properly, RH7.3 and SiS 645 DX

[potash]

I'm using GeForce 2 GTS card under RH Linux 7.3 kernel 2.4.18-3. My M/Board is Gigabyte GA-8STML using SiS 645 DX chipset. Under the default installed setting (ie. using the "nv" driver), X server and things works fine.

Then I installed these two rpm packages:

NVIDIA_kernel-1.0-2960.rh73up.i686.rpm
NVIDIA_GLX-1.0-2960.i386.rpm

and edit the XF86Config-4 to load the nvidia driver. The X server seems to start but the screen remains blank (as oppose to crashing out).

Checking the XFree86.0.log file, the only suspicious message is a warning that the driver fails to verify AGP usage.

It also seems that the driver didn't recognise the SiS 645 DX chipset as cat /proc/driver/nvidia/agp/host_bridge shows that a SiS (Unknown) chipset is used.

I don't know what's wrong? :confused: Can somebody help please?

[Thunderbird]

What I think is that the SIS 645 chipset is too new and because of that it isn't supported by the Nvidia drivers. That doesn't have to mean you can't get AGP. (no AGP makes the drivers a little slower in 3d upto 10% I think)

The first thing to try is the following. We'll try to disable all AGP support. To do that add the following line to the screen or device section of your XF86Config-4 file:
Option "NvAGP" "0"

If that doesn't work you might want to try agpgart. The nvidia drivers support two ways of doing AGP: builtin AGP and agpgart. Agpgart is part of the linux kernel and might have support for the Sis 645 chipset. To enable agpgart add:
Option "NvAGP" "2"

These things might help if the problem is really AGP related. (Had someone in the past using older drivers for who something like this helped. He was using an unsupported Serverworks chipset at the time)

Please post your current XF86Config-4 file if this doesn't help.

[potash]

While waiting for a reply, I made two changes:

1. Switch the card to GeForce Erazor X2, which the X configurator recognize as NVIDIA GeForce 256 (generic).

2. Compile from the NVIDIA kernel driver source using NVIDIA_kernel-1.0-2960.tar.gz

The outcome is essentially the same. The log file doesn't show any error (EE) message, only one warning message (WW):

(WW) NVIDIA(0): Failed to verify AGP usage

Please refer to the attachment for a full XFree86.0.log file.

I did try your recommendation, setting the "NvAGP" option to 0, 1, and 2, the problem persist.

On setting "NvAGP" to "0", the log file diff:

271,272d270
< (**) NVIDIA(0): Option "NvAGP" "0"
< (**) NVIDIA(0): Use of AGP disabled per request
346a345
> (WW) NVIDIA(0): Failed to verify AGP usage

On setting "NvAGP" to "2". the log file diff:

271,272d270
< (**) NVIDIA(0): Option "NvAGP" "2"
< (**) NVIDIA(0): Use of AGPGART requested

From the NVIDIA kernel driver source, it seems that the SiS 645 chipset is recognized, in fact it recognize up to SiS 740. But what bugs me is that my chipset is SiS 645 DX, the device id is 0x646 as oppose to 0x645 for SiS 645.

[Thunderbird]

If you want the drivers to detect the chipset right all you have to do is open nv.c from the kernel module sources. And add a line:
{ PCI_VENDOR_ID_SI, 0x0646, "SiS 645 DX " }, near line 1968.

Now I think the thing that "crashes" X is really something else. I see you already attached your xfree86 log file. I can't see anything strange in the part you pasted. The file also doesn't seem to be complete. Is this your full xfree86 log file?

[potash]

Hmmm, actually that's the full XFree86.0.log I have. But now that you mention it, perhaps the log file is incomplete because I "exit" the X start up by pressing ctrl-alt-del to reboot the machine? :D

Otherwise the screen would remain blank (there's a glitch, it seems like mode switch was done). It's like the X server hangs as the desktop doesn't seems to load (no disk activity at all), there's no cursor and nothing would respond. If the X server crash, I would normally get back to the text mode as I'm in run-level 3.

[Thunderbird]

Are you using any framebuffer? (rivafb)
Can you paste your full XF86Config-4 file then.

[potash]

Ok, here it is.

There's such a section:

Section "Device"
Identifier "Linux Frame Buffer"
Driver "fbdev"
BoardName "Unknown"
EndSection

[Thunderbird]

The config file looks ok. The only lines that I would change (not sure if they make a difference) is to enable the lines loading freetype and type1 since I think you aren't using xfs. Both type1 and freetype are related to font loading, perhaps that didn't work.

[potash]

Just tried your recommendation, still no good. :o

I know the GTS and Erazor X2 cards works with my older machines using older NVIDIA drivers. Maybe I'll test if the 1.0-2960 version driver works in these machines later on after upgrading them to RH 7.3. If so then probably my M/Board is the cause of the problem. A known hardware issue regarding Gigabyte GA-6BX M/Bboard mentioned in Appendix N of the README file gives me the creeps.... :(

Thx for your time, thunderbird.

[Thunderbird]

That Gigabyte mobo is an old intel 440 BX mobo for celeron/p2/p3 cpu's.

Before you said that the screen became totally black when using your card on the Sis 645 mobo. Those modeswitches occur when verifying all modes for your monitor. Something strange is going on. Since the xfree86 log file was far from complete, I suspect that the system crashed or something like it, else it would have finished writing to your xfree86 log file.

Some times it happened to me that I had black screens. (in the past using old drivers) If you can reboot using ctrl-alt-del the system isn't dead. For me doing ctrl-alt-f2, ctrl-alt-f1 (just console switching) gave some screen again. Perhaps you can see the real errors then.

[potash]

When the screen freeze, pressing ctrl-alt-f1, ctrl-alt-f2 etc don't work, at least the screen remains as blank. I tried login from another box and verified that the X server is still up. But there aren't any other X applications running such as gnome-session, sawfish, panel, naitilus, xscreensaver etc when the default desktop is Gnome. I checked the log file at this state, it is the same as the one I attached previously. Killing the X server from the remote box would leave the screen in video mode instead of switching back to text mode. The log file remains the same.

Trying to compare the modeline between NVIDIA(0) and NV(0), it seems only the clock range is different. The NVIDIA(0) one is Clock range: 12.00 to 300.00 MHz.

(II) NV(0): Mitsubishi Diamond Pro 21TX (THN9105): Using hsync range of 30.00-93.00 kHz
(II) NV(0): Mitsubishi Diamond Pro 21TX (THN9105): Using vrefresh range of 50.00-152.00 Hz
(II) NV(0): Clock range: 12.00 to 350.00 MHz
(II) NV(0): Not using default mode "1600x1200" (hsync out of range)
(II) NV(0): Not using default mode "800x600" (hsync out of range)
(II) NV(0): Not using default mode "1792x1344" (hsync out of range)
(II) NV(0): Not using default mode "896x672" (hsync out of range)
(II) NV(0): Not using default mode "1856x1392" (hsync out of range)
(II) NV(0): Not using default mode "928x696" (hsync out of range)
(II) NV(0): Not using default mode "1920x1440" (hsync out of range)
(II) NV(0): Not using default mode "960x720" (hsync out of range)
(--) NV(0): Virtual size is 1280x1024 (pitch 1280)
(**) NV(0): Default mode "1280x1024": 157.5 MHz, 91.1 kHz, 85.0 Hz
(II) NV(0): Modeline "1280x1024" 157.50 1280 1344 1504 1728 1024 1025 1028 1072 +hsync +vsync

[potash]

Thunderbird, please come in.....:)

Today I upgraded one of my old box to RH 7.3 using the same CD and straight away snap in these two rpms:

NVIDIA_kernel-1.0-2960.rh73up.i686.rpm
NVIDIA_GLX-1.0-2960.i386.rpm

Things just work! :D

Please see the attached XFree86.0.log file.

I still get this warning (which now seems harmless) and no error:

(WW) NVIDIA(0): Failed to verify AGP usage

The missing part now come out:

(II) NVIDIA(0): Using XFree86 Acceleration Architecture (XAA)
Screen to screen bit blits
Solid filled rectangles
Solid filled trapezoids
8x8 mono pattern filled rectangles
8x8 mono pattern filled trapezoids
Indirect CPU to Screen color expansion
Solid Lines
Scanline Image Writes
Offscreen Pixmaps
Driver provided FillSolidRects replacement
Driver provided FillSolidSpans replacement
Driver provided FillMono8x8PatternRects replacement
Driver provided WritePixmap replacement
Driver provided ReadPixmap replacement
Setting up tile and stipple cache:
20 128x128 slots
5 256x256 slots
(==) NVIDIA(0): Backing store disabled
(==) NVIDIA(0): Silken mouse enabled
(**) Option "dpms"
(**) NVIDIA(0): DPMS enabled
(II) Loading extension NV-GLX
(II) Loading extension NV-CONTROL
(II) Initializing built-in extension MIT-SHM
(II) Initializing built-in extension XInputExtension
(II) Initializing built-in extension XTEST
(II) Initializing built-in extension XKEYBOARD
(II) Initializing built-in extension LBX
(II) Initializing built-in extension XC-APPGROUP
(II) Initializing built-in extension SECURITY
(II) Initializing built-in extension XINERAMA
(II) Initializing built-in extension XFree86-Bigfont
(II) Initializing built-in extension RENDER
(II) [GLX]: Calling GlxExtensionInit
(**) Option "Protocol" "Microsoft"
(**) Mouse0: Protocol: "Microsoft"
(**) Option "CorePointer"
(**) Mouse0: Core Pointer
(**) Option "Device" "/dev/ttyS0"
(**) Option "BaudRate" "1200"
(**) Option "StopBits" "1"
(**) Option "DataBits" "7"
(**) Option "Parity" "None"
(**) Option "Vmin" "1"
(**) Option "Vtime" "0"
(**) Option "FlowControl" "None"
(**) Option "Emulate3Buttons" "no"
(**) Option "ZAxisMapping" "4 5"
(**) Mouse0: ZAxisMapping: buttons 4 and 5
(**) Mouse0: Buttons: 5
(**) Mouse0: BaudRate: 1200
(II) Keyboard "Keyboard0" handled by legacy driver
(II) XINPUT: Adding extended input device "Mouse0" (type: MOUSE)
(**) Option "BaudRate" "1200"
SetKbdSettings - type: 2 rate: 30 delay: 500 snumlk: 0
SetKbdSettings - Succeeded
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5
GetModeLine - scrn: 0 clock: 157500
GetModeLine - hdsp: 1280 hbeg: 1344 hend: 1504 httl: 1728
vdsp: 1024 vbeg: 1025 vend: 1028 vttl: 1072 flags: 5

So what else do you think I should try? :(

[potash]

Oppps, the attachment lost when I preview the post.

[Thunderbird]

Perhaps it really is a driver bug. The chipset doesn't seem to be supported then. (the log file still isn't complete)

[potash]

You mean the new log file is also incomplete?

Well, the log file is obtained after I logout of my X session and returns to text mode in run-level 3.

If this is really a driver bug, I hope it will get fix so I don't have to dump my Gigabyte M/Board. :)

Or should I dump it? :confused:

[Thunderbird]

So X starts after all? How do you start X? Are you switching to runtime level 5 (gdm/xdm/kdm) or do you use just startx? For testing I would use normal startx.

[potash]

I used normal startx in both cases. In the first case, the X server seems to hang or something, the X-shaped cursor never show up. In the second case the X server starts and the desktop loaded. Both tested in run-level 3.

[Thunderbird]

So X shows things after all or was that only on the other non-SIS 646 DX system?

If you still don't get an image on the SIS 645 chipset report a bug report I think to linux-bugs@nvidia.com

[potash]

No, X didn't show up on the SiS 645 DX chipset machine, but it work on another machine with older Asus M/Board (can't remember the chipset off my mind). Perhaps I'll try again later on to confirm it, and submit a bug report if it don't work.

In the mean time, do you, or anyone, know of a good MicroATX form factor M/Board that has a AGP slot, support PC-2100 SDRAM, onboard NIC and audio, to recommend?

[thebunny]

The NVidia's driver will use the loaded agpgart.o (if it's loaded) no matter what XF86Config says. Did you try unloading agpgart (oh and DRI as well) before doing startx?

Just a thought..

[Thunderbird]

It is actually better for him to use agpgart since agpgart has support for new chipsets earlier than the builtin nvagp. And normally it doesn't really matter which you use.

[thebunny]

Oh I see. I was just speaking from my experience. I have a VIA Apollo Pro 133A north bridge which is quite old and well supported in the kernel yet all the games kept crashing until I went with the NVidia AGP (I removed both agpgart & DRI). Well in any case, he can remove the DRI because NVidia isn't using that at all. Right?

[Thunderbird]

Things like DRI and agpgart are in modules so that is not a problem at all. If I were him I would send a bug report to nvidia. Normally when any agp is disabled (not using nvagp/apgart) a system should still function using the drivers in X. His chipset is too new to have been added to the drivers. (he can add it to nv.c not sure if he did it already and compile a custom driver to use)

linux-bugs@nvidia.com

[lunix]

Hey, I had the same problem with my SIS 645 Ultra. In the BIOS on the 1st or 2nd screen disable APIC. For some reason with APIC enabled it was not assigning a IRQ to VGA. I could tell cause NVDriver bitched about it when I modprobed it from CLI.