Create DRM drivers for VESA and EFI framebuffers
Description
We already have simpledrm for firmware framebuffers. But the driver is originally for ARM boards, not PCs. It is already overloaded with code to support both use cases. At the same time it is missing possible features for VESA and EFI, such as palette modes or EDID support. We should have DRM drivers for VESA and EFI interfaces. The infrastructure exists already and initial drivers can be forked from simpledrm.
Goals
- Initially, a bare driver for VESA or EFI should be created. It can take functionality from simpledrm.
- Then we can begin to add additional features. The boot loader can provide EDID data. With VGA hardware, VESA can support paletted modes or color management. Example code exists in vesafb.
This project is part of:
Hack Week 24
Activity
Comments
-
6 months ago by tdz | Reply
Day 1: I forked of simpledrm into a new vesadrm driver. This immediately works well on non-EFI PC hardware. I further added support for EDID information. This data has to be provided by the boot loader. There's already a field in Linux's boot parameters. It is not set by grub. So a patch there seems required.
See: https://gitlab.suse.de/tzimmermann/linux/-/commits/vesadrm-day1
-
6 months ago by tdz | Reply
Day 2: With vesadrm in place and the EDID property attached, it was time to fill it with data. As mentioned, there's already such a field in the kernel's boot params on x86. It has never been used; or at least i could not find any related code in grub2, grub(-legacy) or even LiLo.
But grub2 already reads out the EDID for its graphics backends. And all we have to do is to copy is to the Linux boot parameters. We cannot transfer EDID extension headers, so we have to clear the extension field. Still, the result is EDID data for the firmware drivers. A patch for grub2 is on OBS at https://build.opensuse.org/package/show/home:tdz:branches:Base:System/grub2.
Here's a screenshot of my test system's GNOME, where it correctly names the monitor from the provided data.
This was on Wayland. I also have a patch pending for Xorg, so that it accepts vesadrm as driver.
Overall, this project is going nicely and I'm very happy with the progress being made.
See: https://gitlab.suse.de/tzimmermann/linux/-/commits/vesadrm-day2
-
6 months ago by tdz | Reply
Day 3: EDID support in vesadrm works well the patched grub2. So today I worked on support for color LUTs. On VGA hardware, it is possible to write the VGA color LUT behind the back of the VESA interface. VESA even tells us if it runs on VGA-compatible hardware. Linux' existing vesafb driver contains example code to do this. If set up, pixels in framebuffer memory will be look into the VGA color LUT during scanout to produce the actual output color.
With this in place, vesadrm now sets up a gamma-LUT property for userspace. Compositors can load gamma tables into the property and vesadrm programs them to VGA. Combined with the EDID data, this enables simple SDR color management. GNOME also implements 'night mode' via gamma luts. When I tested, this now works nicely with vesadrm.
See: https://gitlab.suse.de/tzimmermann/linux/-/commits/vesadrm-day3
-
6 months ago by tdz | Reply
Day 4: The gamma LUT also serves as the color look-up table for color-index modes. Getting DRM format C8 done was the topic of day 4. C8 is a palette mode with 8 bits per pixel. The value is the index into the color LUT. The LUT itself is again in the VGA hardware. C8 is supported by fbcon, so I could wire-up everything directly. When grub programs the mode and runs the kernel, vesadrm now displayes the console correctly. Before day 4, vesadrm would have simply refused to probe.
Compositors don't support C8 (X11 window managers did in the early 90s, but that likely don't count today.) Instead they want XRGB8888. Vesadrm also emulates this via the color LUT. The compositor's framebuffer is in XRGB8888 format, which vesadrm transforms to RGB332 during screen updates. RGB332 is 3 bits of red, 3 bits of green and 2 bits of blue; all packed into a single byte. With the hardware programmed to C8, that byte is the index into the color LUT, which vesadrm programs to RGB666 values that VGA requires. This process is called dithering and transforms the compositor's 24-bit RGB output into an 256-color output on the display.
A photo of the output of my test system is here. This is Firefox on GNOME ruining at 800x600 with 256 colors. While not exactly pretty, it is fast and usable. The desktop UI elements are fairly close to their expected colors. Graphics with higher fidelity, such as the picture on that opened website, show significant distortion.
See: https://gitlab.suse.de/tzimmermann/linux/-/commits/vesadrm-day4
-
6 months ago by tdz | Reply
Summary: I'm really happy how this project turned out. With some polish, everything up-to-and-including day 3 has the potential to go upstream as new driver. That would give a vesadrm driver similar to simpledrm, but with EDID and color management. Support for color-index modes could be added later. There, it would make sense to add additional modes, such as C4, C2, and C1. A bit more work would be required in the DRM framework work support that. Not only would that enable support for really low-end PC hardware, but it would also give an easily available (via qemu) testbed for these low-end features. Some of them are required by today's LCD display, which can be hard to obtain.
For a possible efidrm, it would be a fork of vesadrm up-to-and-including day 2. That is simpledrm plus EDID support. Seems simple enough. AFAICT grub can also provide EDID data from EFI on x86. On non-x86, the kernel would have to fetch it by itself from the EFI firmware.
-
5 months ago by Pharaoh_Atem | Reply
This is awesome! Do you know when you'll submit it to be included upstream?
-
Similar Projects
This project is one of its kind!