ASPEED's AST2500 Display To Be Supported By Linux 4.11's DRM

[phoronix]

Phoronix: ASPEED's AST2500 Display To Be Supported By Linux 4.11's DRM

David Airlie sent in another pull request of DRM material for Linux 4.11, which follows last week's main DRM feature update for Linux 4.11...

http://www.phoronix.com/scan.php?page=news_item&px=ASPEED-AST2500-DRM-4.11

[Ardje]

So on which side of the display is this driver living? Don't forget that we are talking host-cpu vs motherboard controller cpu, both running linux, the latter emulating a VGA card to the host.

[starshipeleven]

So on which side of the display is this driver living?

By looking at the commit descriptions here, it seems a driver for the Linux running on the x86 processor(s), not for the BMC itself.

https://github.com/ozbenh/linux-ast/commits/master

Although it seems that one of the commits allows the host system (Linux on x86 processor) to fully initialize the GPU on its own, in case the BMC firmware or the vBIOS didn't initialize it.

https://github.com/ozbenh/linux-ast/commit/974ff65f293c243641c2f9d2c69a12ebb75bd0f9

Same for older gens, there is a commit that fixes the initialization of the GPU in case the BMC firmware didn't initialize it https://github.com/ozbenh/linux-ast/...3bd4d11450ee06

I'm a bit puzzled about how in the heck this thing works. It seems the GPU is actually shared and the host system can actually access it to pretty low level if it can initialize it from a cold state.

[AndyChow]

It has video-over-ip, maybe that's why you can initialize it remotely. Sort of like wake on lan? Anyway, just guessing.

[starshipeleven]

It has video-over-ip, maybe that's why you can initialize it remotely. Sort of like wake on lan? Anyway, just guessing.

The code in those patches isn't doing anything remotely. It's acting on a local GPU, as if both host CPU and BMC CPU have same access to that GPU.

[Ardje]

I'm a bit puzzled about how in the heck this thing works. It seems the GPU is actually shared and the host system can actually access it to pretty low level if it can initialize it from a cold state.

Afaik, the host thing is legit, and the BMC just can read the config and the VRAM is just shared memory. I really love those BMC's, especially when I have ssh root access to them because then I can put my own serial console clients (picocom) on them instead of the crap IPMI implementations...
Anyway: having reviewed the binaries of the "remote KVM" (10 years ago ;-) )it is just a mangled version of VNC. It would have been nice if the BMC side had a H.264 encoder so it could do some more interesting stuff.
These days the BMC's are unfortunately locked down and license riddled.

[nevion]

Finally, these chips drive me nuts with conflicting error messages in kernel vs xorg module...