https://bugzilla.kernel.org/show_bug.cgi?id=196563

Bug ID: 196563 Summary: Garbled graphics after resume on an R100-based card Product: Drivers Version: 2.5 Kernel Version: 4.11.0-2-686-pae (Debian sid) Hardware: All OS: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: Video(DRI - non Intel) Assignee: drivers_video-dri@kernel-bugs.osdl.org Reporter: felix.von.s@posteo.de Regression: No

When I suspend and then resume my laptop with R100-based graphics, the display becomes garbled. A single scanline from the video memory is stretched over the entire height of the screen; this is usually the scanline containing the caret. Under X11, one can observe the mouse pointer sprite being corrupted as well. When switching VTs, flashes of the correct, 'intended' screen contents can be sometimes seen, but the screen instantly reverts to this single stretched scanline just described.

The issue can be readily worked around via the usual 'vbetool vbestate' hack; although even then the display contents can appear incorrect after resume, as if other programs' framebuffers were pasted onto the screen. Moreover, if the restore part doesn't run soon enough after resuming, the restored screen becomes glitched with randomly lit pixels. The glitching is more severe the more time passes between resuming and restoring VBE state; it is as if the video RAM wasn't being refreshed correctly until the VBE state is restored. Thus, best results are achieved by storing the VBE state on a tmpfs. But in any case, it's nothing a temporary VT switch can't fix.