Bug 27757

I can confirm that with RV620 and RV635. Guess it's issue of not cleaning VRAM.

same thing on Radeon 5750

this is some kind of a security issue, since it might reveal potentially sensitive information after a restart

Or should the vram get cleaned in the kernel drm driver/memory manager?

Same thing on HD5850, any news on this?

for me, the bug disappeared some months ago.

(using git kernel, libdrm, mesa, xf86-video-ati)

Are you using ubuntu?  IIRC, ubuntu uses vesafb and then loads the radeon pretty late in the boot process.  For a smooth bootup, the ddx uses the framebuffer image from the boot manager.  vesa may use a different offset in vram so the console image may not be initialized when X starts in that case so there so it gets initialized to whatever was in vram.

Nope, this is a custom kernel without any *fb driver,
everything needed compiled inside the kernel (radeon+kms).
I normally see some kind of pixel trash (looks like a colored
space picture) for about a second.
If i ran Windows before or had a crash i can see parts of
my previous desktop session.

This is not a _real_ problem for me, but let's be honest,
it looks ugly. This is a dualhead setup, perhaps this could
be an issue?

(using kernel-2.6.39, libdrm/mesa/xf86-video-ati from git)

Should be fixed with the two commits below.

commit 673e1c7637687c74fc9bdeeeffb7ace0d04b734f
Author: Michel Dänzer <michel.daenzer@amd.com>
Date:   Thu Apr 2 17:54:33 2015 +0900

    Defer initial modeset until the first BlockHandler invocation
    
    This ensures that the screen pixmap contents have been initialized when
    the initial modes are set.

commit 80f3d727f93cb6efedd2b39338d2301035965fe2
Author: Michel Dänzer <michel.daenzer@amd.com>
Date:   Wed Apr 22 13:33:15 2015 +0900

    On screen resize, clear the new buffer before displaying it
    
    Fixes garbage being intermittently visible during a screen resize.