drm:i915_hangcheck_elapsed

drm:i915_hangcheck_elapsed

[Bobby Bingham]

Hi all,

On my Sandybridge desktop, I occasionally had an issue under 2.6.37
where I would get something like this in my dmesg output and the X
display would stop updating, except that the mouse cursor could still
move:

[   67.742300] [drm:i915_hangcheck_elapsed] *ERROR* Hangcheck timer elapsed... GPU hung
[   67.743736] [drm:i915_do_wait_request] *ERROR* i915_do_wait_request returns -11 (awaiting 1669 at 1372, next 1673)

This only happened occasionally, usually when a new window appeared on
the screen or when I switched virtual desktops, but most of the time it
was fine.

I upgraded to 2.6.38-rc3, and it now happens consistently just in the
process of logging in.  gdm feels laggy, and X locks before xfce
finishes loading.

It's my first time working with git bisect, but the culprit seems to be
this commit:

8fd2685911cb6c140e6d0588ac04990ce65d4537 (drm/i915: Enable RC6 autodownclocking on Sandybridge)

If you need me to run with some sort of debugging output to help fix
this, just let me know.  Please CC me as I'm not subscribed.

-- 
Bobby Bingham
このメールは再利用されたバイトでできている。

[PATCH] reverse-engineer safe snb wm0 values

[Chris Wilson]

---
Hi Bobby,

this is a patch that I am testing, but so far no else one has experienced
the same symptoms that this fixes on my machines...

It looks like you have won the privilege of being my first victim^W
tester.

Also in drm-intel-fixes
(git://git.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel.git)
is a patch to invalidate the TLB on the BLT which also may cause hangs,
but has a different pattern of failure in
/sys/kernel/debug/dri/0/i915_error_state:

commit 71a77e07d0e33b57d4a50c173e5ce4fabceddbec
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Feb 2 12:13:49 2011 +0000

    drm/i915: Invalidate TLB caches on SNB BLT/BSD rings

I hope one of these brings stability back to your machine...
-Chris

---
 drivers/gpu/drm/i915/intel_display.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index e187257..19c9395 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -4270,7 +4270,9 @@ static void sandybridge_update_wm(struct drm_device *dev)
 				 &sandybridge_cursor_wm_info, latency,
 				 &plane_wm, &cursor_wm)) {
 		I915_WRITE(WM0_PIPEA_ILK,
-			   (plane_wm << WM0_PIPE_PLANE_SHIFT) | cursor_wm);
+			   ((sandybridge_display_wm_info.fifo_size - plane_wm) << WM0_PIPE_PLANE_SHIFT) |
+			   (2 << WM0_PIPE_SPRITE_SHIFT) |
+			   cursor_wm);
 		DRM_DEBUG_KMS("FIFO watermarks For pipe A -"
 			      " plane %d, " "cursor: %d\n",
 			      plane_wm, cursor_wm);
@@ -4282,7 +4284,9 @@ static void sandybridge_update_wm(struct drm_device *dev)
 				 &sandybridge_cursor_wm_info, latency,
 				 &plane_wm, &cursor_wm)) {
 		I915_WRITE(WM0_PIPEB_ILK,
-			   (plane_wm << WM0_PIPE_PLANE_SHIFT) | cursor_wm);
+			   ((sandybridge_display_wm_info.fifo_size - plane_wm) << WM0_PIPE_PLANE_SHIFT) |
+			   (2 << WM0_PIPE_SPRITE_SHIFT) |
+			   cursor_wm);
 		DRM_DEBUG_KMS("FIFO watermarks For pipe B -"
 			      " plane %d, cursor: %d\n",
 			      plane_wm, cursor_wm);
-- 
1.7.2.3

Re: [PATCH] reverse-engineer safe snb wm0 values

[Bobby Bingham]

On Fri,  4 Feb 2011 09:12:57 +0000
Chris Wilson <chris@chris-wilson.co.uk> wrote:

> ---
> Hi Bobby,
> 
> this is a patch that I am testing, but so far no else one has experienced
> the same symptoms that this fixes on my machines...
> 
> It looks like you have won the privilege of being my first victim^W
> tester.
> 
> Also in drm-intel-fixes
> (git://git.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel.git)
> is a patch to invalidate the TLB on the BLT which also may cause hangs,
> but has a different pattern of failure in
> /sys/kernel/debug/dri/0/i915_error_state:
> 
> commit 71a77e07d0e33b57d4a50c173e5ce4fabceddbec
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Wed Feb 2 12:13:49 2011 +0000
> 
>     drm/i915: Invalidate TLB caches on SNB BLT/BSD rings
> 
> I hope one of these brings stability back to your machine...
> -Chris
> 

Alas, neither of those patches seems to help.

-- 
Bobby Bingham
このメールは再利用されたバイトでできている。