Intel G965 graphics, X.org 7.2, and FreeBSD 6.2-RELEASE

On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>
> >> Well, just judging from the Xorg log output you've posted, it appears
> >> that the driver you're trying to use requires the agpgart device.
>
> >> Try manually loading the agp driver after booting and logging in with:
>
> >> kldload agp
>
> >> Then verify that the device is created:
>
> >> ls -l /dev/agpgart
>
> >> Then try starting X again, or reconfiguring X once more.
>
> >> If your device truly is *not* an AGP device, but the driver requires
> >> one, then you've got a real problem, to say the least. :-) Are you
> >> absolutely certain you're using the right driver?
>
> > "kldload agp" returns the following error message: -----
> > kldload: can't load agp: File exists
> > -----
> > and so "ls /dev/" does not show any agp device...
>
> OK, so apparently, you have the agp device built into your kernel, but
> it's not configuring properly for some reason.
>
> Have you checked your BIOS setup? Some machines allow the graphics
> device to be configured as either a straight PCI device or as an AGP
> device.
>
> You can verify your kernel with 'kldstat|grep agp', BTW.
>
> --
> Conrad J. Sabatier


there is no mention of AGP in the BIOS on this machine. there is only
a PEG (PCI-Express Graphics, i guess) option.
and i had tried 'kldstat' before, and it does not show anything AGP
either.

thanks

On Thu, 31 May 2007 13:56:25 -0700, novembre wrote:

> On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
>> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
>> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>>
>> >> Well, just judging from the Xorg log output you've posted, it
>> >> appears that the driver you're trying to use requires the agpgart
>> >> device.
>>
>> >> Try manually loading the agp driver after booting and logging in
>> >> with:
>>
>> >> kldload agp
>>
>> >> Then verify that the device is created:
>>
>> >> ls -l /dev/agpgart
>>
>> >> Then try starting X again, or reconfiguring X once more.
>>
>> >> If your device truly is *not* an AGP device, but the driver requires
>> >> one, then you've got a real problem, to say the least. :-) Are you
>> >> absolutely certain you're using the right driver?
>>
>> > "kldload agp" returns the following error message: ----- kldload:
>> > can't load agp: File exists
>> > -----
>> > and so "ls /dev/" does not show any agp device...
>>
>> OK, so apparently, you have the agp device built into your kernel, but
>> it's not configuring properly for some reason.
>>
>> Have you checked your BIOS setup? Some machines allow the graphics
>> device to be configured as either a straight PCI device or as an AGP
>> device.
>>
>> You can verify your kernel with 'kldstat|grep agp', BTW.
>>
>> --
>> Conrad J. Sabatier
>
>
> there is no mention of AGP in the BIOS on this machine. there is only a
> PEG (PCI-Express Graphics, i guess) option. and i had tried 'kldstat'
> before, and it does not show anything AGP either.
>
> thanks

Have you seen Michel Talon's followup in this thread? He had the same
problem as you, upgraded to either 6-STABLE or 7-CURRENT, and it cured
the problem.

You may want to try a cvsup to STABLE, do a complete buildworld/
buildkernel, and see if you get lucky.

--
Conrad J. Sabatier

On May 31, 4:10 pm, "Conrad J. Sabatier" wrote:
> On Thu, 31 May 2007 13:56:25 -0700, novembre wrote:
> > On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
> >> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
> >> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>
> >> >> Well, just judging from the Xorg log output you've posted, it
> >> >> appears that the driver you're trying to use requires the agpgart
> >> >> device.
>
> >> >> Try manually loading the agp driver after booting and logging in
> >> >> with:
>
> >> >> kldload agp
>
> >> >> Then verify that the device is created:
>
> >> >> ls -l /dev/agpgart
>
> >> >> Then try starting X again, or reconfiguring X once more.
>
> >> >> If your device truly is *not* an AGP device, but the driver requires
> >> >> one, then you've got a real problem, to say the least. :-) Are you
> >> >> absolutely certain you're using the right driver?
>
> >> > "kldload agp" returns the following error message: ----- kldload:
> >> > can't load agp: File exists
> >> > -----
> >> > and so "ls /dev/" does not show any agp device...
>
> >> OK, so apparently, you have the agp device built into your kernel, but
> >> it's not configuring properly for some reason.
>
> >> Have you checked your BIOS setup? Some machines allow the graphics
> >> device to be configured as either a straight PCI device or as an AGP
> >> device.
>
> >> You can verify your kernel with 'kldstat|grep agp', BTW.
>
> >> --
> >> Conrad J. Sabatier
>
> > there is no mention of AGP in the BIOS on this machine. there is only a
> > PEG (PCI-Express Graphics, i guess) option. and i had tried 'kldstat'
> > before, and it does not show anything AGP either.
>
> > thanks
>
> Have you seen Michel Talon's followup in this thread? He had the same
> problem as you, upgraded to either 6-STABLE or 7-CURRENT, and it cured
> the problem.
>
> You may want to try a cvsup to STABLE, do a complete buildworld/
> buildkernel, and see if you get lucky.
>
> --
> Conrad J. Sabatier


yes, i have seen it...didn't have time to try it though one other
thing though: this graphics chip is PCI-Express and not AGP.

but i had installed -STABLE branch before, and that did not solve my
problem. so maybe i need to use the -CURRENT branch this time, but i'm
a bit reluctant, since i'm going to use this machine for numerical
calculations at the university, and i need something very stable (i
know that FreeBSD is the definition of stable, but the stability of -
CURRENT and -RELEASE branches are different IMO).
i don't like the idea of reinstalling the whole operating system, but
i also don't know how to replace only the agp module in the kernel.
how do i take the current agp module out from the kernel? do i need to
recompile the kernel? also, if i get the new agp source from, say, -
CURRENT, compile it myself, how can i load it? in loader.conf?

i apologize for all these questions, but i'm not very experience and
haven't done anything like this before.

novembre-at-gmail.com wrote:
> how do i take the current agp module out from the kernel? do i need to
> recompile the kernel? also, if i get the new agp source from, say, -
> CURRENT, compile it myself, how can i load it? in loader.conf?
>
> i apologize for all these questions, but i'm not very experience and
> haven't done anything like this before.
>

Make a backup of /usr/src/sys/pci/agp_i810.c and copy the one from
CURRENT here. Then
cd /usr/src/sys/modules/agp
make all install clean
kldxref /boot/kernel

Load agp from loader.conf and you are fine, if it works...

--

Michel TALON

On May 31, 5:24 pm, t...@lpthe.jussieu.fr (Michel Talon) wrote:
> novem...@gmail.com wrote:
> > how do i take the current agp module out from the kernel? do i need to
> > recompile the kernel? also, if i get the new agp source from, say, -
> > CURRENT, compile it myself, how can i load it? in loader.conf?
>
> > i apologize for all these questions, but i'm not very experience and
> > haven't done anything like this before.
>
> Make a backup of /usr/src/sys/pci/agp_i810.c and copy the one from
> CURRENT here. Then
> cd /usr/src/sys/modules/agp
> make all install clean
> kldxref /boot/kernel
>
> Load agp from loader.conf and you are fine, if it works...
>
> --
>
> Michel TALON


okay, i got the latest agp_i810.c (and also the latest agp_intel.c,
but using that specific file would make 'make' give errors, so i
reverted to using the old agp_intel.c), copied it to /usr/src/sys/
pci/, and did a 'make all install clean' in /usr/src/sys/modules/agp/.
then i did 'kldxref /boot/kernel'.
however, still 'kldstat | grep agp' does not return anything, just as
'dmesg | grep agp'. and X.org would still give me the same error
message when i use i810 as the driver in xorg.conf, but using vesa it
can load...

this is the results of 'dmesg | grep VGA'
pci0: at device 2.0 (no driver attached)
sc0: VGA <16 virtual consoles, flags=0x300>
vga0: at port 0x3c0-0x3df iomem 0xa0000-0xbffff on
isa0

i think the problem is the chip being PCI Express and not AGP.

thanks

novembre-at-gmail.com wrote:

> however, still 'kldstat | grep agp' does not return anything, just as
> 'dmesg | grep agp'. and X.org would still give me the same error
> message when i use i810 as the driver in xorg.conf, but using vesa it
> can load...
>
> this is the results of 'dmesg | grep VGA'
> pci0: at device 2.0 (no driver attached)
> sc0: VGA <16 virtual consoles, flags=0x300>
> vga0: at port 0x3c0-0x3df iomem 0xa0000-0xbffff on
> isa0
>
> i think the problem is the chip being PCI Express and not AGP.

The chip in my laptop is also pci express and it is recognized, so i
think the problem is perhaps that the chip identifier is not listed
in agp_i810.c . Perhaps you should try to identify this number and shove
it in an appropriate part of agp_i810.c. In my case the difference
between the working and the non working agp_i810.c was exactly that, the
addition of a few more identifiers.
To get the identifier, you run pciconf -l, which gives something like
agp0-at-pci0:0:0: class=0x060000 card=0x80b21043 chip=0x25608086 rev=0x02
hdr=0x00
Here the numer is from chip=0x25608086. For the present case, this is
identified in agp_intel.c as being
case 0x25608086:
return ("Intel 82845G host to AGP bridge");

>
> thanks
>

--

Michel TALON

On Jun 1, 4:26 am, t...@lpthe.jussieu.fr (Michel Talon) wrote:
> novem...@gmail.com wrote:
> > however, still 'kldstat | grep agp' does not return anything, just as
> > 'dmesg | grep agp'. and X.org would still give me the same error
> > message when i use i810 as the driver in xorg.conf, but using vesa it
> > can load...
>
> > this is the results of 'dmesg | grep VGA'
> > pci0: at device 2.0 (no driver attached)
> > sc0: VGA <16 virtual consoles, flags=0x300>
> > vga0: at port 0x3c0-0x3df iomem 0xa0000-0xbffff on
> > isa0
>
> > i think the problem is the chip being PCI Express and not AGP.
>
> The chip in my laptop is also pci express and it is recognized, so i
> think the problem is perhaps that the chip identifier is not listed
> in agp_i810.c . Perhaps you should try to identify this number and shove
> it in an appropriate part of agp_i810.c. In my case the difference
> between the working and the non working agp_i810.c was exactly that, the
> addition of a few more identifiers.
> To get the identifier, you run pciconf -l, which gives something like
> agp0-at-pci0:0:0: class=0x060000 card=0x80b21043 chip=0x25608086 rev=0x02
> hdr=0x00
> Here the numer is from chip=0x25608086. For the present case, this is
> identified in agp_intel.c as being
> case 0x25608086:
> return ("Intel 82845G host to AGP bridge");
>
>
>
> > thanks
>
> --
>
> Michel TALON



'pciconf -lv | less' returns the following

hostb0ci0:0:0: class=0x060000 card=0x01dd1028 chip=0x29a08086
rev=0x02 hdr=0x00
vendor = 'Intel Corporation'
class = bridge
subclass = HOST-PCI
pcib1-at-pci0:1:0: class=0x060400 card=0x00000088 chip=0x29a18086
rev=0x02 hdr=0x01
vendor = 'Intel Corporation'
class = bridge
subclass = PCI-PCI
none0-at-pci0:2:0: class=0x030000 card=0x01dd1028 chip=0x29a28086
rev=0x02 hdr=0x00
vendor = 'Intel Corporation'
class = display
subclass = VGA
none1-at-pci0:2:1: class=0x038000 card=0x01dd1028 chip=0x29a38086
rev=0x02 hdr=0x00
vendor = 'Intel Corporation'
class = display

The description in agp_i810.c for adding new devices goes as:
"...devid is the id of the graphics controller (pci:0:2:0, for
example)...The bridge_offset is the offset to subtract from devid to
get the id of the hostb that the device is on." So in my case,
bridge_offset should be 0x00020000 (not 0x00010000), right? It should
point to hostb0 (not to pcib1), am I right?
So in agp_i810.c, I added the following line to agp_i810_matches

{0x29A28086, CHIP_I915, 0x00020000, "Intel 82965G (965G GMCH) SVGA
conrtoller"}

I also used the -CURRENT branch's agp_intel.c (which needed the -
CURRENT branch's agpreg.h to compile) just to be up-to-date. I also
noticed that 'make all install clean' in /usr/src/sys/modules/agp/
executes the command 'kldxref /boot/kernel' by itself, so I don't
think I need to do it myself again, do I?
Anyhow, I rebooted the machine, and again, there's no AGP device
detected, and 'pciconf -lv' returns the same message as before, i.e.
"none0" instead of, say, "agp0".

A weird thing happend while compiling: at first, I had the wrong value
of 0x00010000 for the bridge_offset in the agp_i810.c (with the -
CURRENT files agpreg.h and agp_intel.c), and the compiling went
smoothly. Then I realized that 0x00010000 is the wrong value, so I
changed it to 0x00020000, but then it could not compile! There was a
couple of error while compiling agp.c, but I had NOT touched that file
at all. The first thing that came to my mind was to revert
bridge_offset back to 0x0001000 again, but it was not compiling at all
(the same errors about agp.c). I have no idea why it was compiling
smoothly one moment ago while it could not compile the same file after
that small change. So I downloaded the -CURRENT branch's agp.c and,
with that it compiled with only two warnings...but the results are not
good anyway.

I assume that the chipset is not supported by FreeBSD yet, although
I've read that Intel has recently released the drivers for 965G. I
could not find the chip identifier of this machine's HOST-PCI, i.e.
0x29a08086, even in the -CURRENT branch's agp_intel.c. So I assume
there's no support for thic chipset yet...

Thanks a lot

In article <1180649649.131164.137650@g4g2000hsf.googlegroups.c om>,
wrote:
>On May 31, 4:10 pm, "Conrad J. Sabatier" wrote:
>> On Thu, 31 May 2007 13:56:25 -0700, novembre wrote:
>> > On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
>> >> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
>> >> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>>
>> >> >> Well, just judging from the Xorg log output you've posted, it
>> >> >> appears that the driver you're trying to use requires the agpgart
>> >> >> device.
>>
>> >> >> Try manually loading the agp driver after booting and logging in
>> >> >> with:
>>
>> >> >> kldload agp
>>
>> >> >> Then verify that the device is created:
>>
>> >> >> ls -l /dev/agpgart
>>
>> >> >> Then try starting X again, or reconfiguring X once more.
>>
>> >> >> If your device truly is *not* an AGP device, but the driver requires
>> >> >> one, then you've got a real problem, to say the least. :-) Are you
>> >> >> absolutely certain you're using the right driver?
>>
>> >> > "kldload agp" returns the following error message: ----- kldload:
>> >> > can't load agp: File exists
>> >> > -----
>> >> > and so "ls /dev/" does not show any agp device...
>>
>> >> OK, so apparently, you have the agp device built into your kernel, but
>> >> it's not configuring properly for some reason.
>>
>> >> Have you checked your BIOS setup? Some machines allow the graphics
>> >> device to be configured as either a straight PCI device or as an AGP
>> >> device.
>>
>> >> You can verify your kernel with 'kldstat|grep agp', BTW.
>>
>> >> --
>> >> Conrad J. Sabatier
>>
>> > there is no mention of AGP in the BIOS on this machine. there is only a
>> > PEG (PCI-Express Graphics, i guess) option. and i had tried 'kldstat'
>> > before, and it does not show anything AGP either.
>>
>> > thanks
>>
>> Have you seen Michel Talon's followup in this thread? He had the same
>> problem as you, upgraded to either 6-STABLE or 7-CURRENT, and it cured
>> the problem.
>>
>> You may want to try a cvsup to STABLE, do a complete buildworld/
>> buildkernel, and see if you get lucky.
>>
>> --
>> Conrad J. Sabatier
>
>
>yes, i have seen it...didn't have time to try it though one other
>thing though: this graphics chip is PCI-Express and not AGP.
>
>but i had installed -STABLE branch before, and that did not solve my
>problem. so maybe i need to use the -CURRENT branch this time, but i'm
>a bit reluctant, since i'm going to use this machine for numerical
>calculations at the university, and i need something very stable (i
>know that FreeBSD is the definition of stable, but the stability of -
>CURRENT and -RELEASE branches are different IMO).

FWIW, I've been running -CURRENT for quite some time now, and it's really
pretty solid. Stability, I would say, is every bit the equal of -STABLE,
or even of the latest -RELEASE.

Right now, in fact, would be a good time to upgrade to -CURRENT, if you're
so inclined. No earth-shaking changes in progress at the moment, so
things are currently pretty well "settled" for the time being.

>i don't like the idea of reinstalling the whole operating system, but
>i also don't know how to replace only the agp module in the kernel.
>how do i take the current agp module out from the kernel? do i need to
>recompile the kernel? also, if i get the new agp source from, say, -
>CURRENT, compile it myself, how can i load it? in loader.conf?

Actually, since you mentioned before that kldstat showed no agp module, I
have to wonder if you even have agp compiled into your kernel. Even if
it's not working, it *should* still show up in kldstat.

As far as using the agp module from -CURRENT, you could either build it
into the kernel or load it as a module, it's up to you, and yes, it does
require doing a "make buildkernel" either way (unless you have
MODULES_WITH_WORLD defined in /etc/make.conf, in which case, of course,
you'd need to do a buildworld as well).

I would think, though, that it would be a lot simpler in the long run just
to cvsup your source tree to -CURRENT, and do a full
buildworld/buildkernel.

>i apologize for all these questions, but i'm not very experience and
>haven't done anything like this before.

No problem, that's what this group is here for. :-)

--
Conrad J. Sabatier

On Jun 2, 8:01 pm, conr...@cox.net (Conrad J. Sabatier) wrote:
> In article <1180649649.131164.137...@g4g2000hsf.googlegroups.c om>,
>
>
>
> wrote:
> >On May 31, 4:10 pm, "Conrad J. Sabatier" wrote:
> >> On Thu, 31 May 2007 13:56:25 -0700, novembre wrote:
> >> > On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
> >> >> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
> >> >> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>
> >> >> >> Well, just judging from the Xorg log output you've posted, it
> >> >> >> appears that the driver you're trying to use requires the agpgart
> >> >> >> device.
>
> >> >> >> Try manually loading the agp driver after booting and logging in
> >> >> >> with:
>
> >> >> >> kldload agp
>
> >> >> >> Then verify that the device is created:
>
> >> >> >> ls -l /dev/agpgart
>
> >> >> >> Then try starting X again, or reconfiguring X once more.
>
> >> >> >> If your device truly is *not* an AGP device, but the driver requires
> >> >> >> one, then you've got a real problem, to say the least. :-) Are you
> >> >> >> absolutely certain you're using the right driver?
>
> >> >> > "kldload agp" returns the following error message: ----- kldload:
> >> >> > can't load agp: File exists
> >> >> > -----
> >> >> > and so "ls /dev/" does not show any agp device...
>
> >> >> OK, so apparently, you have the agp device built into your kernel, but
> >> >> it's not configuring properly for some reason.
>
> >> >> Have you checked your BIOS setup? Some machines allow the graphics
> >> >> device to be configured as either a straight PCI device or as an AGP
> >> >> device.
>
> >> >> You can verify your kernel with 'kldstat|grep agp', BTW.
>
> >> >> --
> >> >> Conrad J. Sabatier
>
> >> > there is no mention of AGP in the BIOS on this machine. there is only a
> >> > PEG (PCI-Express Graphics, i guess) option. and i had tried 'kldstat'
> >> > before, and it does not show anything AGP either.
>
> >> > thanks
>
> >> Have you seen Michel Talon's followup in this thread? He had the same
> >> problem as you, upgraded to either 6-STABLE or 7-CURRENT, and it cured
> >> the problem.
>
> >> You may want to try a cvsup to STABLE, do a complete buildworld/
> >> buildkernel, and see if you get lucky.
>
> >> --
> >> Conrad J. Sabatier
>
> >yes, i have seen it...didn't have time to try it though one other
> >thing though: this graphics chip is PCI-Express and not AGP.
>
> >but i had installed -STABLE branch before, and that did not solve my
> >problem. so maybe i need to use the -CURRENT branch this time, but i'm
> >a bit reluctant, since i'm going to use this machine for numerical
> >calculations at the university, and i need something very stable (i
> >know that FreeBSD is the definition of stable, but the stability of -
> >CURRENT and -RELEASE branches are different IMO).
>
> FWIW, I've been running -CURRENT for quite some time now, and it's really
> pretty solid. Stability, I would say, is every bit the equal of -STABLE,
> or even of the latest -RELEASE.
>
> Right now, in fact, would be a good time to upgrade to -CURRENT, if you're
> so inclined. No earth-shaking changes in progress at the moment, so
> things are currently pretty well "settled" for the time being.
>
> >i don't like the idea of reinstalling the whole operating system, but
> >i also don't know how to replace only the agp module in the kernel.
> >how do i take the current agp module out from the kernel? do i need to
> >recompile the kernel? also, if i get the new agp source from, say, -
> >CURRENT, compile it myself, how can i load it? in loader.conf?
>
> Actually, since you mentioned before that kldstat showed no agp module, I
> have to wonder if you even have agp compiled into your kernel. Even if
> it's not working, it *should* still show up in kldstat.
>

I'm not quite sure what is wrong here! When booting, I can see the
loaded modules with 'lsmod', and agp is among them. Also, if I try to
load agp by 'load agp' there, it says that the module already exists.
Moreover, I do have load_agp="YES" in my loader.conf. And 'kldload
agp' returns
-----
kldload: can't load agp: File exists
-----
so I don't know what else to do to make 'kldstat' show the loaded agp
module!

>
> As far as using the agp module from -CURRENT, you could either build it
> into the kernel or load it as a module, it's up to you, and yes, it does
> require doing a "make buildkernel" either way (unless you have
> MODULES_WITH_WORLD defined in /etc/make.conf, in which case, of course,
> you'd need to do a buildworld as well).
>
> I would think, though, that it would be a lot simpler in the long run just
> to cvsup your source tree to -CURRENT, and do a full
> buildworld/buildkernel.
>

Thanks...I'll read the handbook whenever I get some free time to see
how I can upgrade the source tree and compile the kernel. Then I'll
report the results...
But still, I'm currently using agp.c, agp_intel.c, and agp_i810.c
(which I modified to include the devid of Intel G965) from the -
CURRENT branch, and the graphics chipset is not recognized. Do you
think that if I do the upgrade, it will help?

>
> >i apologize for all these questions, but i'm not very experience and
> >haven't done anything like this before.
>
> No problem, that's what this group is here for. :-)
>

Thanks so much. This is what I really like about FreeBSD groups...

>
> --
> Conrad J. Sabatier

On Jun 4, 3:16 pm, novem...@gmail.com wrote:
> On Jun 2, 8:01 pm, conr...@cox.net (Conrad J. Sabatier) wrote:
>
>
>
> > In article <1180649649.131164.137...@g4g2000hsf.googlegroups.c om>,
>
> > wrote:
> > >On May 31, 4:10 pm, "Conrad J. Sabatier" wrote:
> > >> On Thu, 31 May 2007 13:56:25 -0700, novembre wrote:
> > >> > On May 30, 3:30 pm, "Conrad J. Sabatier" wrote:
> > >> >> On Wed, 30 May 2007 13:04:51 -0700, novembre wrote:
> > >> >> > On May 28, 10:30 pm, "Conrad J. Sabatier" wrote:
>
> > >> >> >> Well, just judging from the Xorg log output you've posted, it
> > >> >> >> appears that the driver you're trying to use requires the agpgart
> > >> >> >> device.
>
> > >> >> >> Try manually loading the agp driver after booting and logging in
> > >> >> >> with:
>
> > >> >> >> kldload agp
>
> > >> >> >> Then verify that the device is created:
>
> > >> >> >> ls -l /dev/agpgart
>
> > >> >> >> Then try starting X again, or reconfiguring X once more.
>
> > >> >> >> If your device truly is *not* an AGP device, but the driver requires
> > >> >> >> one, then you've got a real problem, to say the least. :-) Are you
> > >> >> >> absolutely certain you're using the right driver?
>
> > >> >> > "kldload agp" returns the following error message: ----- kldload:
> > >> >> > can't load agp: File exists
> > >> >> > -----
> > >> >> > and so "ls /dev/" does not show any agp device...
>
> > >> >> OK, so apparently, you have the agp device built into your kernel, but
> > >> >> it's not configuring properly for some reason.
>
> > >> >> Have you checked your BIOS setup? Some machines allow the graphics
> > >> >> device to be configured as either a straight PCI device or as an AGP
> > >> >> device.
>
> > >> >> You can verify your kernel with 'kldstat|grep agp', BTW.
>
> > >> >> --
> > >> >> Conrad J. Sabatier
>
> > >> > there is no mention of AGP in the BIOS on this machine. there is only a
> > >> > PEG (PCI-Express Graphics, i guess) option. and i had tried 'kldstat'
> > >> > before, and it does not show anything AGP either.
>
> > >> > thanks
>
> > >> Have you seen Michel Talon's followup in this thread? He had the same
> > >> problem as you, upgraded to either 6-STABLE or 7-CURRENT, and it cured
> > >> the problem.
>
> > >> You may want to try a cvsup to STABLE, do a complete buildworld/
> > >> buildkernel, and see if you get lucky.
>
> > >> --
> > >> Conrad J. Sabatier
>
> > >yes, i have seen it...didn't have time to try it though one other
> > >thing though: this graphics chip is PCI-Express and not AGP.
>
> > >but i had installed -STABLE branch before, and that did not solve my
> > >problem. so maybe i need to use the -CURRENT branch this time, but i'm
> > >a bit reluctant, since i'm going to use this machine for numerical
> > >calculations at the university, and i need something very stable (i
> > >know that FreeBSD is the definition of stable, but the stability of -
> > >CURRENT and -RELEASE branches are different IMO).
>
> > FWIW, I've been running -CURRENT for quite some time now, and it's really
> > pretty solid. Stability, I would say, is every bit the equal of -STABLE,
> > or even of the latest -RELEASE.
>
> > Right now, in fact, would be a good time to upgrade to -CURRENT, if you're
> > so inclined. No earth-shaking changes in progress at the moment, so
> > things are currently pretty well "settled" for the time being.
>
> > >i don't like the idea of reinstalling the whole operating system, but
> > >i also don't know how to replace only the agp module in the kernel.
> > >how do i take the current agp module out from the kernel? do i need to
> > >recompile the kernel? also, if i get the new agp source from, say, -
> > >CURRENT, compile it myself, how can i load it? in loader.conf?
>
> > Actually, since you mentioned before that kldstat showed no agp module, I
> > have to wonder if you even have agp compiled into your kernel. Even if
> > it's not working, it *should* still show up in kldstat.
>
> I'm not quite sure what is wrong here! When booting, I can see the
> loaded modules with 'lsmod', and agp is among them. Also, if I try to
> load agp by 'load agp' there, it says that the module already exists.
> Moreover, I do have load_agp="YES" in my loader.conf. And 'kldload
> agp' returns
> -----
> kldload: can't load agp: File exists
> -----
> so I don't know what else to do to make 'kldstat' show the loaded agp
> module!
>
>
>
> > As far as using the agp module from -CURRENT, you could either build it
> > into the kernel or load it as a module, it's up to you, and yes, it does
> > require doing a "make buildkernel" either way (unless you have
> > MODULES_WITH_WORLD defined in /etc/make.conf, in which case, of course,
> > you'd need to do a buildworld as well).
>
> > I would think, though, that it would be a lot simpler in the long run just
> > to cvsup your source tree to -CURRENT, and do a full
> > buildworld/buildkernel.
>
> Thanks...I'll read the handbook whenever I get some free time to see
> how I can upgrade the source tree and compile the kernel. Then I'll
> report the results...
> But still, I'm currently using agp.c, agp_intel.c, and agp_i810.c
> (which I modified to include the devid of Intel G965) from the -
> CURRENT branch, and the graphics chipset is not recognized. Do you
> think that if I do the upgrade, it will help?
>
>
>
> > >i apologize for all these questions, but i'm not very experience and
> > >haven't done anything like this before.
>
> > No problem, that's what this group is here for. :-)
>
> Thanks so much. This is what I really like about FreeBSD groups...
>
>
>
> > --
> > Conrad J. Sabatier


today, i again did some search on the net for Intel 965 graphics
chipset and came up with the following post by Eric Anholt on the
FreeBSD -CURRENT mailing list:
http://docs.freebsd.org/cgi/getmsg.c...reebsd-current

Considering that the mailing list post was made on April 20, 2007 and
that the diff is made on top of agp_i810.c revision 1.39 (which was
submitted 8 months ago to -CURRENT), I guess I was right in saying
that this chip is still not supported even in the -CURRENT branch (I
have compiled and am using 1.39 right now, and the chip is not
detected). I'll try his patch and report back...