Back in the late 90s to early 2000s, SiS graphics chips were fairly common and found in many low-end devices. Today, the chips are still capable enough for simple graphics needs, but the graphics cards were on PCI and AGP buses. They are not usable in modern computers.

However, there exist USB-based graphics cards with a SiS 315 graphics chip. Those are around on Ebay et al [1] and easily usable with current computers. I already do have a driver for the old PCI-based SiS drivers and have long been struggling to find something useful to do with it. Converting it to serve USB devices would finally make it useful.

The goal for Hackweek 20 is to dedust my SiS driver and make it work with the 315 chip. That might take a bit or not; I'm not sure yet. Afterwards, the PCI I/O needs to be replaced with corresponding USB operations.

There are quite a few resources. My PCI driver is at [2]. For the USB devices, there exists an old userspace driver at [3] and a kernel stub at [4]. A general description of the device can be found in the Wayback Machine. [5]

If successful, the driver is supposed to be included in the upstream kernel.






Looking for hackers with the skills:

kernel graphics drm usb

This project is part of:

Hack Week 20


  • about 1 month ago: fos liked this project.
  • about 1 month ago: ptesarik liked this project.
  • about 2 months ago: tdz liked this project.
  • about 2 months ago: tdz started this project.
  • about 2 months ago: tdz added keyword "kernel" to this project.
  • about 2 months ago: tdz added keyword "graphics" to this project.
  • about 2 months ago: tdz added keyword "drm" to this project.
  • about 2 months ago: tdz added keyword "usb" to this project.
  • about 2 months ago: tdz originated this project.

  • Comments

    • tdz
      about 1 month ago by tdz | Reply

      Day 1: Today I returned to my old driver for SiS PCI devices, which used to work 3yrs ago. Admittedly, the driver has bit-rotted quite a bit. I since updated it and added atomic modesetting. But I never really tested these changes. I can get the modes from the monitor via EDID functionality and the driver now detects the VRAM size correctly. My monitor reports a signal error, which indicates a bug when programming display resolution or timing. Fixing the driver is what I currently do. I'll also have to update it for the 315 chip. My hope is that I can then replace the PCI-bus functionality with USB and have a semi-working USB driver later this week.

    • tdz
      29 days ago by tdz | Reply

      Day 2: I'm still working on getting the old PCI-based SiS driver to work correctly within the current kernel. I managed to get the display mode set and improved color-format settings. The display still looks wrong, but it's getting better. I give it one more day. No matter what the state is tomorrow, I'll push for USB support on Thursday.

    • tdz
      28 days ago by tdz | Reply

      Day 3: The old driver uses the device VRAM directly. Today I converted it to DRM's SHMEM helpers. This is necessary to work with the USB device. SHMEM buffers will serve as shadow framebuffers that the driver can copy over USB into the device's memory.

    • tdz
      26 days ago by tdz | Reply

      Day 4: I began to turn the PCI driver into a USB driver. Each PCI I/O operation has to be converted to a USB bulk message. The rsp code is already in the old kernel stub driver somewhere. The PCI device is a VGA card and therefore POSTed by the BIOS. For the USB device, I expect that there might be an additional POST process necessary. The old userspace driver should have the rsp code somewhere. No way I'll be able to finish all of this by the end of the week.

    • tdz
      26 days ago by tdz | Reply

      Day 5: After some back and forth, I'm able to communicate with the USB device. No mode setting yet, just basic communication. I extracted this form the old sisusb driver in the kernel. I'll probably continue this project during the next hackweek.

    Similar Projects

    Test mainline kernel on an older Qualcomm SOC (msm89xx), explore current Qualcomm mainlining kernel effort by pvorel

    Project Description

    Qualcomm concentrate ...

    Separate kernel module signature to a independent file by joeyli

    [comment]: # (Please use the project descriptio...

    work on sunxi a64 cpufreq driver (for teres-1, pine64) by mbrugger

    With the teres-1 [1] laptop we have a first arm...

    Design the 2021 Open Build Service sticker by hennevogel


    ![2019 sticker](

    Learn DaVinci Resolve by psimons

    [comment]: # (Please use the project descriptio...