[Openpvrsgx-devgroup] New version letux-pvrsrvkm-v5.6.0-rc1 available

H. Nikolaus Schaller hns at goldelico.com
Mon Feb 17 09:05:48 CET 2020

Hi Tony,

> Am 17.02.2020 um 01:20 schrieb Tony Lindgren <tony at atomide.com>:
> * Tony Lindgren <tony at atomide.com> [200216 19:38]:
>> * H. Nikolaus Schaller <hns at goldelico.com> [200216 18:47]:
>>> Hi Tony,
>>>> Am 16.02.2020 um 19:29 schrieb Tony Lindgren <tony at atomide.com>:
>>>> Hi,
>>>> * H. Nikolaus Schaller <hns at goldelico.com> [200211 13:28]:
>>>>> A new version of our holistic PVR/SGX driver is available
>>>>> Kernel: v5.6.0-rc1
>>>>> Result: letux-pvrsrvkm-v5.6.0-rc1 (rebased on v5.6.0-rc1)
>>>>> Result: letux-pvrsrvkm (linear history)
>>>>> Github: https://github.com/openpvrsgx-devgroup/linux_openpvrsgx/branches
>>>>> Components:
>>>> ...
>>>>> letux/pvrsrvkm-1.17.4948957
>>>> FYI, I heard from ruleh on #maemo-leste today that this branch has been
>>>> updated for kernel modules and the omap5-sgx-ddk-um-linux branch for
>>>> related blobs. And the blobs now at least init with pvrsrvinit also on
>>>> omap4 and presumably also on omap34xx.
>>>> So presumably now oma34xx (n900) and am3517 should work the same as
>>>> in the beaglebone example but with 1.17:
>>>> http://blog.0xpebbles.org/PowerVR-SGX-on-the-beaglebone-black-in-2019
>>>> Nikolaus, care to check that letux/pvrsrvkm-1.17.4948957 is
>>>> in sync? Stuff has just recently gotten updated it seems.
>>> Wow, cool!
>>> https://git.ti.com/cgit/graphics/omap5-sgx-ddk-linux/log/?h=ti-img-sgx/1.17.4948957/k5.4
>>> shows e.g.:
>>> km: Migrate to kernel version 5.4
>>> km: Add build targets for sgx530-121 and sgx540-120 devices
> I think I got this done and pushed out to a testing branch.
> I first applied the missing patches from omap5-sgx-ddk-um-linux:
> 50c1ec2308b9 ("km: Use CONFIG_PREEMPTION")
> 76f305295daf ("km: pvr_fence: Replace seqcount with seqlock_t")
> 1cc2eded30f8 ("km: Migrate to kernel version 5.4")
> 2a777b8fb72a ("km: Add build targets for sgx530-121 and sgx540-120 devices")

This really sounds cool :) Although I think the key problem that
we have with 1.14 is not building a kernel module but user-space libs
matching the 1.14 kernel module.

Anyways, having the latest DDK on as many platforms as possible is

> 5e1217b70a13 ("km: Fix null pointer dereference when dma_ops is null")
> 4dc02a9be903 ("km: Add support for compiling with the ARM GCC 9.2+ toolchain")
> a45dd03f1e3e ("km: Fix check against mixed-bit DDK and kernel builds")
> cf7f48cb30ab ("km: Remove Xorg and Linux from version string")
> 2e83a1d081c5 ("km: Drop unused platform definitions for jz4780")

Well, I am not sure if we should drop this :)

I am currently working on getting the HDMI driver of the CI20 board working.
Then it is at least possible to test the jz4780 variant.

And, the openpvrsgx goal is to support all platforms where we can get code

> c85f932329f8 ("km: Use CONFIG_ZONE_DMA for LPAE")
> 290c3f3c9a7c ("build: Make eurasia_km makefile match to UM")
> 14f7ec76136d ("build: Move target selection to top and use elseif logic")
> 6dac08386fe5 ("build: Remove unused config variable SYS_OMAP5_UEVM")
> b1c0f63c128e ("build: Do not build without TARGET_PRODUCT set")
> 40eade347566 ("build: Remove deprecated no warning flags")
> Then fixed up things to compile and use pvr-drv. I've only tested that
> the kernel module loads so far,


> I need to update my test system to
> test the pvrsrvctl as it needs GLIBC_2.28.

Yes, I think one needs to have Debian Buster (10). Stretch (9) is too old.

> I've pushed it out to [0][1] below, please test.

Find. I'll try to integrate into letux-pvrsrvkm-v5.6.0-rc2.

BR and thanks,

> Droid4 users need to apply the led-backlight and related dts changes
> to test on droid4.
> Regards,
> Tony
> [0] git://github.com/tmlind/linux_openpvrsgx letux-pvrsrvkm-v5.6.0-rc1-pvr-drv
> [1] https://github.com/tmlind/linux_openpvrsgx/tree/letux-pvrsrvkm-v5.6.0-rc1-pvr-drv

More information about the openpvrsgx-devgroup mailing list