[Letux-kernel] OMAP3 SGX on latest kernels

H. Nikolaus Schaller hns at goldelico.com
Tue Nov 1 09:00:14 CET 2016


Hi Adam,

> Am 31.10.2016 um 23:53 schrieb Adam Ford <aford173 at gmail.com>:
> 
> On Mon, Oct 31, 2016 at 4:52 PM, H. Nikolaus Schaller <hns at goldelico.com> wrote:
>> Hi Adam,
>> I just found your attempt to get SGX working on OMAP3.
>> 
>> [RFC] ARM: OMAP3: Add device tree and pdata hooks to enable SGX on OMAP3
>> 
>> https://patchwork.kernel.org/patch/9311537/
>> 
>> I am trying for a while to achieve the same.
>> 
>> But Graphics_SDK_5_01_01_02 (AFAIK the latest for OMAP3/DM3730) seems to
>> call functions that require some reset-controller to be enabled in most
>> modern kernels or in other words they don't work without. Because they
>> directly check for "resets" and "reset-names" properties.
>> 
>> Therefore I started to work on the reset-controller which Tero Kristo
>> has proposed a while ago (Sept 2015). Officially, it is on hold because
>> they want to fix clock hwmods first. But it does not boot on v4.9-rc2
>> (it boots on v4.3).
> 
> I am working on the 4.4.y long term kernel so I'm going to try and
> experiment with that.

That would be nice to share experiences.

>> 
>> So what is your status of getting SGX running?
>> 
> 
> Unfortunately, the customer that I was working with decided not to
> pursue it any further, but I'll gladly try to help with what you've
> got.
> 
>> My latest DT patch looks like this (funnily I had the same typo of a missing
>> 0 of 50000000):
>> 
>> http://git.goldelico.com/?p=gta04-kernel.git;a=blobdiff;f=arch/arm/boot/dts/omap3.dtsi;h=9c32224f887cadce6470ffb91d600a452ea677e7;hp=353d818ce5a6dc059488db7a842bd217b4836ff6;hb=5887fe785f3f13b4f6e816c6dd0a14a9d8d1e227;hpb=27756e0e87533f1be5e2d83cd53de73c1cc8af66
>> 
>> What I wonder is why you have a different interrupt. Unfortunately I don't
>> remember where I got the <37> from. But there is some table to cross-check
>> with the TRM.
>> 
> 
> I think the different processors have different interrupts.  I am
> using the DM3730 which is 21, but I could be wrong or misinterpreting
> the datasheet.  Which OMAP3 were you targeting?

Yes, we also have the DM3730.

I now checked with spruf98k (OMAP3530) and sprugn4m (DM3730) and
both say (Table 10-4 resp. 12-4):

M_IRQ_21	SGX_IRQ	2D/3D graphics module

So the <21> appears to be correct and I will fix it.

> 
> 
>> The full patches I work on are here:
>> http://git.goldelico.com/?p=gta04-kernel.git;a=shortlog;h=refs/heads/work/hns/gpu/pvr-v4
>> http://git.goldelico.com/?p=gta04-kernel.git;a=shortlog;h=refs/heads/work/test/omap/reset-cont-v4.9
>> 
> 
> Thanks for your feedback. I'm going to pull them down and see what
> happens with the 4.4 LTS kernel

Regarding the SGX_RESET, Tero confirmed that the register address is missing
but said that RSTCTRLs are usually at offset 0x50 [1].

BR,
Nikolaus


[1] http://lists.goldelico.com/pipermail/letux-kernel/2016-October/000716.html



More information about the Letux-kernel mailing list