[Patch?] SiS 746 AGP-Support

[Patch?] SiS 746 AGP-Support

[Hemmann, Volker Armin]

Hi,

this is my first try of making and submitting a patch, so please, don't kill 
me for it. 

I don't know, if the following changes are 'clean' but they give me a working 
agpsupport for my SiS 746Fx based mobo.

This (attempt) of a patch is against 2.4.21-rc1:

*** /usr/src/linux-2.4.21-rc1/drivers/pci/pci.ids       2003-04-25 01:
52:30.000000000 +0200
--- /usr/src/linux/drivers/pci/pci.ids  2003-04-23 07:18:54.000000000
+0200
***************
*** 1119,1124 ****
--- 1119,1125 ----
        0735  735 Host
        0740  740 Host
        0745  745 Host
+       0756  746 Host
        0900  SiS900 10/100 Ethernet
                1039 0900  SiS900 10/100 Ethernet Adapter
        0961  SiS961 [MuTIOL Media IO]

*** /usr/src/linux-2.4.21-rc1/drivers/char/agp/agpgart_be.c     2003-0
4-25 01:52:29.000000000 +0200
--- /usr/src/linux/drivers/char/agp/agpgart_be.c        2003-04-23 07:
33:48.000000000 +0200
*************** static struct {
*** 4595,4600 ****
--- 4595,4606 ----
                "SiS",
                "745",
                sis_generic_setup },
+       { PCI_DEVICE_ID_SI_746,
+               PCI_VENDOR_ID_SI,
+               SIS_GENERIC,
+               "SiS",
+               "746",
+               sis_generic_setup },
        { PCI_DEVICE_ID_SI_730,
                PCI_VENDOR_ID_SI,
                SIS_GENERIC,

I am doing this since 2.4.21-pre6 without any problems.

Glück Auf,
Volker

Re: [Patch?] SiS 746 AGP-Support

[Dave Jones]

 > I don't know, if the following changes are 'clean' but they give me a working 
 > agpsupport for my SiS 746Fx based mobo.
 > 
 > This (attempt) of a patch is against 2.4.21-rc1:

Might work fine as long as you have an agp2.0 card in the slot, but the
minute you put a 3.0 (read as AGPx8) card in there, things are very
likely to break.

I've not seen the specs for this chipset, but most of the AGP3 chipsets
I've seen have a fallback mode in their register set which gets enabled
as soon as you plug in an AGP2 card. These registers don't get enabled
with an AGP3 card, instead you need to read from different registers,
and in most cases, act completly different to decode aperture sizes etc.

The generic routines in 2.5 *might* work, but are untested on this chipset.
2.4 currently has no AGP3 support at all. Some folks did backport what
I've done in 2.5 a while ago, but I would advise against merging it at
this stage, as there is still work to be done there, including stability
fixes. Right there are a number of possible problems which may include
random memory corruption.

		Dave

Re: [Patch?] SiS 746 AGP-Support

[Hemmann, Volker Armin]

On Friday 25 April 2003 02:40, Dave Jones wrote:
>  > I don't know, if the following changes are 'clean' but they give me a
>  > working agpsupport for my SiS 746Fx based mobo.
>  >
>  > This (attempt) of a patch is against 2.4.21-rc1:
>
> Might work fine as long as you have an agp2.0 card in the slot, but the
> minute you put a 3.0 (read as AGPx8) card in there, things are very
> likely to break.
>
> I've not seen the specs for this chipset, but most of the AGP3 chipsets
> I've seen have a fallback mode in their register set which gets enabled
> as soon as you plug in an AGP2 card. These registers don't get enabled
> with an AGP3 card, instead you need to read from different registers,
> and in most cases, act completly different to decode aperture sizes etc.
>
> The generic routines in 2.5 *might* work, but are untested on this chipset.
> 2.4 currently has no AGP3 support at all. Some folks did backport what
> I've done in 2.5 a while ago, but I would advise against merging it at
> this stage, as there is still work to be done there, including stability
> fixes. Right there are a number of possible problems which may include
> random memory corruption.
>
> 		Dave

I have only a AGP 2 (geforce 4-mx) card, so I missed that(and with one I would 
only to be able to say 'it doesn't work' so thanks for your explanation). But 
without this changes I won't even able to use dga, because the first 
dga-enabled app completely locks up my box. 
And to have working AGP2 and non working APG3 looks a lot better for me than 
no AGP-support at all.  

2.5.* is able to 'mostly' boot on my mobo so I can't say anything about that..


Glück Auf,
Volker

Re: [Patch?] SiS 746 AGP-Support

[Dave Jones]

On Fri, Apr 25, 2003 at 03:02:26AM +0200, Hemmann, Volker Armin wrote:

 > I have only a AGP 2 (geforce 4-mx) card, so I missed that(and with one I would 
 > only to be able to say 'it doesn't work' so thanks for your explanation). But 
 > without this changes I won't even able to use dga, because the first 
 > dga-enabled app completely locks up my box. 
 > And to have working AGP2 and non working APG3 looks a lot better for me than 
 > no AGP-support at all.  

Sure, I wasn't objecting per'se to the patch, but people should be made
aware it's not going to help them a tiny bit if they have an AGP3 card.
It may just abort nicely, it may take down the machine in horrible ways
depending on how well SiS handles reads/writes to disabled registers.

		Dave

Re: [Patch?] SiS 746 AGP-Support

[Hemmann, Volker Armin]

On Friday 25 April 2003 04:05, Dave Jones wrote:
> On Fri, Apr 25, 2003 at 03:02:26AM +0200, Hemmann, Volker Armin wrote:
>  > I have only a AGP 2 (geforce 4-mx) card, so I missed that(and with one I
>  > would only to be able to say 'it doesn't work' so thanks for your
>  > explanation). But without this changes I won't even able to use dga,
>  > because the first dga-enabled app completely locks up my box.
>  > And to have working AGP2 and non working APG3 looks a lot better for me
>  > than no AGP-support at all.
>
> Sure, I wasn't objecting per'se to the patch, but people should be made
> aware it's not going to help them a tiny bit if they have an AGP3 card.
> It may just abort nicely, it may take down the machine in horrible ways
> depending on how well SiS handles reads/writes to disabled registers.
>
> 		Dave

Enabling FastWrites with the nvidia-module kills X and the box instantly at 
boot, but is ok when X was running without FastWirties, is shut down, 
reloading the modules with FastWrites and starting X. So there are even some 
odd things with APG2

I am sorry that I can't help very much, despite of testing,  I am not afraid 
of building a lot of kernels, because my knowledge of C tends to zero. I came 
to the first mentioned changes by grepping for SiS stuff in the kernel 
sources, meditating about the stuff, and than simple copied what seems to be 
ok for me.

If you or somebody else wants me to test SiS related stuff, I am quite happy 
to help, except of that I am just an user with too little knowledge about C 
or kernel stuff. 

Glück Auf,
Volker