| Summary: | [BDW] Desktop (font/texts) is corrupted after running SynMark test(s) | ||||||
|---|---|---|---|---|---|---|---|
| Product: | xorg | Reporter: | Eero Tamminen <eero.t.tamminen> | ||||
| Component: | Driver/intel | Assignee: | Chris Wilson <chris> | ||||
| Status: | VERIFIED DUPLICATE | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||
| Severity: | normal | ||||||
| Priority: | medium | CC: | valtteri.rantala, wendy.wang | ||||
| Version: | git | ||||||
| Hardware: | Other | ||||||
| OS: | All | ||||||
| Whiteboard: | |||||||
| i915 platform: | i915 features: | ||||||
| Bug Depends on: | 84958 | ||||||
| Bug Blocks: | |||||||
| Attachments: |
|
||||||
Try
diff --git a/src/mesa/drivers/dri/i965/intel_blit.c b/src/mesa/drivers/dri/i965/intel_blit.c
index 73ab488..e5206df 100644
--- a/src/mesa/drivers/dri/i965/intel_blit.c
+++ b/src/mesa/drivers/dri/i965/intel_blit.c
@@ -316,7 +316,7 @@ intelEmitCopyBlit(struct brw_context *brw,
if (src_offset & 4095)
return false;
}
- if ((dst_y_tiled || src_y_tiled) && brw->gen < 6)
+ if (dst_y_tiled || src_y_tiled)
return false;
/* do space check before going any further */
This issue is triggered by running a program suffering from bug 84958 and test-cases that don't trigger bug 84958 don't seem to cause this. -> adding dependency (In reply to Chris Wilson from comment #1) > - if ((dst_y_tiled || src_y_tiled) && brw->gen < 6) > + if (dst_y_tiled || src_y_tiled) > return false; This didn't affect the issue. Eero, do you have a moment to see if the recent couple of BDW fixes resolve this corruption as well? (In reply to Chris Wilson from comment #3) > Eero, do you have a moment to see if the recent couple of BDW fixes resolve > this corruption as well? Unfortunately, when trying this, I still got the frame update stop issue on F0. Not sure whether it's due to test setup. We verified bug 84958 with Ubuntu 14.04 slim + unity setup (with no panels), on two different BDW machines, this I tried to verify with X, icewm, gnome-terminal combo with which I had originally encountered it. These setups are on different disks, but I've run them on same machine, so machine shouldn't be a problem. I'm not able to run Unity on the failing setup with latest X etc (X crashes when Unity comes up) to check whether (Unity/compiz) compositor would there hide bug 84958. On the slim+unity setup, stopping slim [1] freezes all keyboard input on the machine for some reason (haven't had time to debug that more yet), so I couldn't try there things without compositor to see whether that reveals the bug again. I'll try to get both setups working tomorrow. [1] haven't been able to get our own X running with Ubuntu lightdm, that's why slim session manager is used instead. lightdm removes LD_LIBRARY_PATH etc from environment, but fixing that and few other issues still doesn't produce working desktop, unlike with slim. Created attachment 109051 [details]
Screenshot of broken gnome-terminal
|
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
Everything seems to work OK in GL composited Ubuntu Unity desktop, but desktop is messed up (all text is invisible/garbage) after running SynMark2 tests (e.g. OglDrv* ones) if one uses non-compositing window manager or non-GL composting window manager. If one triggers GPU reset from sysfs: echo 1 > /sys/kernel/debug/dri/0/i915_wedged The problem goes away. Comment from Chris: --------------------------- The DRI clients operate in an isolated hardware context(s), but X uses the global context. This means that either mesa is corrupting surface outside of its GTT e.g. through the global GTT, of which there are still reports in bugzilla) or that X is not setting up all the state that it uses when rendering. The escription of how the GPU reset fixes things suggests the latter. --------------------------- Not sure whether this is related to bug 84958, both seem to have been on BDW "always".