Linux framebuffer graphics and VSync

Question

I\'m making a game in C that works on the linux framebuffer. So far I have a red 100x100 square that moves in tandem with the mouse, just like a pointer. I have implemented

Answer 1

This isn't the correct way to do double-buffering. You're right to do all the painting on a back-buffer, but then you do a memcpy to transfer the data to the front. A screen refresh could easily happen during the copy.

To do this properly, you should only have to switch a pointer to the data; not copy the data itself. With the Linux framebuffer device, this is done by having a "virtual" screen that is twice as large as the physical screen, and using an offset variable to set whether you're showing the top or bottom half. You can query the size and set offset using the FBIOGET_VSCREENINFO, FBIOPUT_VSCREENINFO, and FBIOPAN_DISPLAY ioctl calls.

This page briefly gives some details about this: http://www.ummon.eu/Linux/API/Devices/framebuffer.html

All the relevant data structures are in the linux/fb.h header file.

Answer 2

Unfortunately I've learned the hard way that @Steven Bell's answer isn't quite right. Though he is right that a screen refresh could easily happen during the memcpy the proper resolution is not to create a virtual framebuffer of double the size of the screen and pan between them as so many threads suggest. Anyone who attempts this solution will receive an error described here: invalid argument error when setting yres_virtual in fb_var_screeninfo.

Per this thread: https://forum.odroid.com/viewtopic.php?f=55&t=8741 it is not quite possible to truly double buffer the framebuffer (/dev/fb0, although I have heard that the raspberry PI might be an exception to this rule). But that doesn't mean there isn't a way to double buffer in linux using low level graphics.

The real solution here is to use libdrm (/dev/dri/card0) to draw low level graphics to the screen. There is a really nice example of how to do this here: https://github.com/dvdhrm/docs/blob/master/drm-howto/modeset-vsync.c that I used myself when I was trying to solve this problem.

Anyways I hope I've saved someone the amazing headache that I had to go through to figure this out in the future.