Created attachment 113316 [details] [review]
[PATCH] drm/i915: Skip all the gfx clock force on CHV

Please test this patch.

(In reply to Ville Syrjala from comment #1)
> Created attachment 113316 [details] [review] [review]
[PATCH] drm/i915: Skip all the
> gfx clock force on CHV

Please test this patch.

Patch this patch on latest nightly branch commit:dbd79c545cbbf63785cbda8d13260b3e827f0f6d

S3 will not cause dmesg warn anymore, but seems S3 failed itself, when type "echo mem > /sys/power/state", it will resume immediately.

Created attachment 113440 [details]
dmesg after patch

(In reply to Ding Heng from comment #3)
> Created attachment 113440 [details]
> dmesg after patch

That dmesg starts some time after resuming, so I can't really see anything from it. Please redo the test, and also to get rid of that hotplug noise in dmesg do the following (preferably before loading i915):
intel_iosf_sb_write gpio_nc 0x5420 0x03010300
intel_iosf_sb_write gpio_nc 0x5440 0x03010300

Created attachment 113850 [details]
new dmesg with kernel20150213 and patch

This is the new dmesg with the same patch and kernel.

Can you try this patch?
http://lists.freedesktop.org/archives/intel-gfx/2015-February/060790.html

I will try and confirm if we need vlv_force_gfx_clock or not for chv

(In reply to Deepak S from comment #6)
> Can you try this patch?
> http://lists.freedesktop.org/archives/intel-gfx/2015-February/060790.html
> 
> I will try and confirm if we need vlv_force_gfx_clock or not for chv

Since latest kernel commit will cause BSW Fail to boot up, I was trying to install your patch to commit 0b2a1076c5cb4f383d6a8c940ffab1e27f241097(nightly branch 2015-02-25, latest available kernel for BSW), but your patch failed to install, here is the rej file:

[root@glody-pc drm-intel]# cat drivers/gpu/drm/i915/i915_drv.c.rej
--- drivers/gpu/drm/i915/i915_drv.c
+++ drivers/gpu/drm/i915/i915_drv.c
@@ -1193,11 +1193,13 @@
        int err;

        val = I915_READ(VLV_GTLC_SURVIVABILITY_REG);
-       WARN_ON(!!(val & VLV_GFX_CLK_FORCE_ON_BIT) == force_on);

 #define COND (I915_READ(VLV_GTLC_SURVIVABILITY_REG) & VLV_GFX_CLK_STATUS_BIT)
        /* Wait for a previous force-off to settle */
-       if (force_on) {
+       if (force_on && !IS_CHERRYVIEW(dev_priv->dev)) {
+               /* WARN_ON only for the Valleyview */
+               WARN_ON(!!(val & VLV_GFX_CLK_FORCE_ON_BIT) == force_on);
+
                err = wait_for(!COND, 20);
                if (err) {
                        DRM_ERROR("timeout waiting for GFX clock force-off (%08x)\n",

Created attachment 114068 [details]
dmesg after patch

Latest nightly branch had fixed the boot issue. I installed your patch in comment 6 on latest nightly branch. Still some warning in dmesg.

root@x-bsw01:~# dmesg -r | egrep "<[1-4]>" |grep drm
<4>[  338.174625] WARNING: CPU: 2 PID: 24 at drivers/gpu/drm/i915/i915_gem.c:4626 i915_gem_suspend+0xbb/0xd3 [i915]()
<4>[  338.174734] Modules linked in: dm_mod snd_hda_codec_hdmi iTCO_wdt iTCO_vendor_support snd_hda_codec_realtek snd_hda_codec_generic serio_raw pcspkr snd_hda_intel snd_hda_controller i2c_i801 snd_hda_codec snd_hwdep snd_pcm snd_timer lpc_ich mfd_core snd soundcore battery joydev ac acpi_cpufreq i915 button video dr _kms_helper drm
<4>[  338.175191]  [<ffffffffa0079784>] ? i915_drm_suspend+0x59/0x1a1 [i915]

Neither of the patches mentioned here-- either attached or linked-- appear to have made it into drm-intel-nightly. Was it decided to use a different approach?

Created attachment 114343 [details]
dmesg_with_drm-intel-testing-2015-03-13

This issue also exists on drm-intel-testing-2015-03-13 with V59 BIOS.

exists on drm-intel-testing-2015-03-27

commit 85250ddff7a603dfe0ec0503a9e6395f79424f61
Author: Deepak S <deepak.s@linux.intel.com>
Date:   Sat Mar 28 15:23:34 2015 +0530

    drm/i915/chv: Remove Wait for a previous gfx force-off

the other suspend warning from #8 looks like a separate issue.  Please file another bug for that unless it's a duplicate.

Tried latest nightly commit d72ff1ab1499711c941831400629c14493313b3a and this issue got fixed.

Closing old verified.