Finally got around to trying the 2.2.18pre series and the agp/drm
backport and noticed something odd at bootup. Here's an extract from
dmesg:
Linux agpgart interface v0.99 (c) Jeff Hartmann
agpgart: Maximum main memory to use for agp memory: 440M
agpgart: Detected Intel 440BX chipset
agpgart: AGP aperture is 64M @ 0xe0000000
A bit later:
[drm] The mga drm module requires the agpgart module to function correctly
Please load the agpgart module before you load the mga module
Although XFree86 seems to be happy enough loading the dri and drm
modules.
Is this supposed to only work with modules? .config snippet:
CONFIG_AGP=y
CONFIG_AGP_INTEL=y
# CONFIG_AGP_I810 is not set
# CONFIG_AGP_VIA is not set
# CONFIG_AGP_AMD is not set
# CONFIG_AGP_SIS is not set
# CONFIG_AGP_ALI is not set
CONFIG_DRM=y
# CONFIG_DRM_TDFX is not set
# CONFIG_DRM_GAMMA is not set
# CONFIG_DRM_R128 is not set
# CONFIG_DRM_I810 is not set
CONFIG_DRM_MGA=y
System:
RedHat 7.0 + errata
kgcc used to build kernel, with binutils: GNU ld version 2.10.91
PIII 750
512MB RAM
Matrox g400 (OEM version)
Please cc me on any replies, as I am not subscribed. I do my best to
follow web archives...
Jamie
--
Jamie http://www.intermag.com
And I said, "This must be the place." -- Laurie Anderson, "Big Science"
On Wed, 29 Nov 2000, Jamie Manley wrote:
> Finally got around to trying the 2.2.18pre series and the agp/drm
> backport and noticed something odd at bootup. Here's an extract from
> dmesg:
>
> Linux agpgart interface v0.99 (c) Jeff Hartmann
> agpgart: Maximum main memory to use for agp memory: 440M
> agpgart: Detected Intel 440BX chipset
> agpgart: AGP aperture is 64M @ 0xe0000000
>
> A bit later:
>
> [drm] The mga drm module requires the agpgart module to function correctly
> Please load the agpgart module before you load the mga module
>
> Although XFree86 seems to be happy enough loading the dri and drm
> modules.
>
> Is this supposed to only work with modules? .config snippet:
>
> CONFIG_AGP=y
> CONFIG_AGP_INTEL=y
Probably you have modversions enabled (CONFIG_MODVERSION=y). Disable that
and try again, or build as modules. 2.4 fixed this problem in the proper
way, but I don't know what's going to happen about 2.2 ...
john
--
"Penguins are so sensitive to my needs."
- Lyle Lovett
On Fri, 1 Dec 2000 14:52:11 +0000 (GMT),
John Levon <[email protected]> wrote:
>Probably you have modversions enabled (CONFIG_MODVERSION=y). Disable that
>and try again, or build as modules. 2.4 fixed this problem in the proper
>way, but I don't know what's going to happen about 2.2 ...
I have sent a backport of inter_module_xxx to Alan Cox, he has not
included it in 2.2 kernels yet, maybe 2.2.19. When inter_module_xxx is
available in 2.2, the DRM/AGP change can be backported as well.
On Fri, Dec 01, 2000 at 02:52:11PM +0000, John Levon wrote:
> > [drm] The mga drm module requires the agpgart module to function correctly
> > Please load the agpgart module before you load the mga module
> >
> > Although XFree86 seems to be happy enough loading the dri and drm
> > modules.
> >
> > Is this supposed to only work with modules? .config snippet:
> >
> > CONFIG_AGP=y
> > CONFIG_AGP_INTEL=y
>
> Probably you have modversions enabled (CONFIG_MODVERSION=y). Disable that
> and try again, or build as modules. 2.4 fixed this problem in the proper
> way, but I don't know what's going to happen about 2.2 ...
>
> john
I knew I should have posted the whole .config :)
Yes, modversions was enabled. Should that be affecting the build of
the kernel proper?
Jamie
--
Jamie http://www.intermag.com
And I said, "This must be the place." -- Laurie Anderson, "Big Science"
[Jamie Manley]
> Yes, modversions was enabled. Should that be affecting the build of
> the kernel proper?
The bug you ran into is that MODVERSIONS messes up the
'get_module_symbol' function, which is a sort of "optional dependency"
mechanism used by a few modules such as DRI (in this case: DRI needs to
be able to use the facilities of agpgart, but should also work
*without* agpgart present, since many systems have PCI video cards).
MODVERSIONS is ugly and gross for any number of reasons, but the
get_module_symbol problem is quite localized -- AGP/DRI, MTD and maybe
one or two other subsystems. In any case it has been replaced by a
much better inter-module registration system in 2.4.
Peter
On Fri, Dec 01, 2000 at 09:22:22PM -0600, Peter Samuelson wrote:
>
> [Jamie Manley]
> > Yes, modversions was enabled. Should that be affecting the build of
> > the kernel proper?
>
> The bug you ran into is that MODVERSIONS messes up the
> 'get_module_symbol' function, which is a sort of "optional dependency"
> mechanism used by a few modules such as DRI (in this case: DRI needs to
> be able to use the facilities of agpgart, but should also work
> *without* agpgart present, since many systems have PCI video cards).
Of course, PCI and DRI shouldn't be mutually exclusive. Glad to hear
it's being worked on.
>
> MODVERSIONS is ugly and gross for any number of reasons, but the
> get_module_symbol problem is quite localized -- AGP/DRI, MTD and maybe
> one or two other subsystems. In any case it has been replaced by a
> much better inter-module registration system in 2.4.
>
> Peter
I guess that means I should start testing 2.4.0test* on this machine
:) Thanks for the background.
Jamie
--
Jamie http://www.intermag.com
And I said, "This must be the place." -- Laurie Anderson, "Big Science"