Note, that

commit 51a0559b120dd0a57a938b300bccefbf1142e495
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 18 20:55:02 2014 +0100

    sna: Do not treat both src/dst having no clones as being the same
    
    Fixes regression from
    commit 62aaf2ff4f8597067cf387865707baa00ed9a123
    Author: Chris Wilson <chris@chris-wilson.co.uk>
    Date:   Wed Jun 18 18:41:15 2014 +0100
    
        sna: Skip redundant copies when already cloned
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

was a stupendously stupid commit. Please double check that the failure is still there this morning.

Nevermind:

"It was before both of them and it still present in
 c257c936b42b92827b784cee0b7caa54e6040364."

Interesting thing. when i started http://ie.microsoft.com/testdrive/Performance/ParticleAcceleration/ in Firefox garbage became orange :)

What is the name of the widget? And how do I install it?

System Load Viewer.
http://kde-look.org/content/show.php/System+Status?content=74891
But i already had it installed with kde.
Oh it happened right now - instead of garbade it drew exact copy of clock that i have on panel below - without distortions )

Sorry, I haven't succeeded in getting that widget up and running yet. Is that the only corruption you've spotted?

Sometimes it happened in dropdown menus. But easily to notice in combination with Kwin vsync bug - it makes menu drawing kind of slower

I have been focusing on other bugs, and haven't reproduced this. Would it be possible for you to bisect this?

Yep. Bug was introduced in:
322e51a3a552d89e5e5ecbb82a9243f134c8d36f sna: Relax PREFER_GPU so that we don't needless create small GPU bo

Oh boy. Just to confirm the same bug is still on master:

diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c
index 78f5fed..1feea14 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -3504,7 +3504,7 @@ sna_drawable_use_bo(DrawablePtr drawable, unsigned flags, const BoxRec *box,
        }
        if ((flags & FORCE_GPU) == 0 && priv->gpu_bo == NULL && sna_pixmap_choose_tiling(pixmap, DEFAULT_TILING) == I915_TILING_NONE) {
                DBG(("%s: no gpu bo and linear, discard PREFER_GPU\n", __FUNCTION__));
-               flags &= ~PREFER_GPU;
+               //flags &= ~PREFER_GPU;
        }
 
        if ((flags & (PREFER_GPU | IGNORE_CPU)) == IGNORE_CPU) {

Bug is still on master. But this patch fixes it.

How about

diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c
index f481c19..409ae70 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -3824,6 +3824,8 @@ cpu_fail:
 
                if (!sna->kgem.can_blt_cpu)
                        goto move_to_gpu;
+
+               goto move_to_gpu;
        }
 
        if (!sna->kgem.can_blt_cpu)

?

No. This patch doesn't fix.

But if fixes scrolling bug. Didn't test 1st patch for scrolling though.

The actual line in question was removed in

commit 416c223861aaeb61c6408e6537315b78e908df43
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Jun 23 09:20:55 2014 +0100

    sna: Do not set CAN_CREATE_GPU flag for untiled allocations
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

but in theory we should still take the same path. There are two possible causes here:

1. We discard damage that we shouldn't

2. The client assumes that Pixmaps are zeroed on initialisation (they aren't).

Hmm, I think I'll try concocting a debug patch to initialise all Pixmaps to magenta, but first I have to rule out (1).

(In reply to comment #14)
> But if fixes scrolling bug. Didn't test 1st patch for scrolling though.

Interesting. Please try

commit bb4c057620fb0558a77a9b7db538786ffda7cf13
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Jun 23 09:22:13 2014 +0100

    sna: Prefer to render over damage into the GPU bo
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

for the scrolling bug.

garbage bug is fixed.
scrolling bug is not.

I believed that i made a comment that:
"But if fixes scrolling bug. Didn't test 1st patch for scrolling though."
was not correct. It seemed so for a quite while. But then it's gone

Ok, thanks for the update on the scrolling bug. Every time I think I have a lead it evapourates! :)

Let's mark this as resolved, even though I am sure I haven't fixed the underlying cause...