Skip to contentSkip to conversation

drm/amd/amdgpu: S3 resumed failed after 4-5 times loop

PATCH In review jimqu 3 participants 4 messages Aug 18, 2016 – Aug 19, 2016 3 messages not in the archive

Subsystems (2)
lore
Share & tools
lore

jimqu

author
#0 lore raw

PATCH 1/1 drm/amd/amdgpu: S3 resumed failed after 4-5 times loop

phenomenon: software hang when device resume back, read UVD fence is 0xffffffff and read pcie pid is 0xffff.
the issue is caused by VCE reset when update cg setting. according to HW programming guide, adjust update VCE cg sequence.

Change-Id I18b12eea21c045908cdd23f93a0b196b87bfed6c
---
 drivers/gpu/drm/amd/amdgpu/vce_v2_0.c | 123 +++++++++++++++++++++++-----------
 1 file changed, 84 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +84 -39
index 9e70df9..6ce7c07 100644
@@ -40,10 +40,12 @@
 #define VCE_V2_0_FW_SIZE	(256 * 1024)
 #define VCE_V2_0_STACK_SIZE	(64 * 1024)
 #define VCE_V2_0_DATA_SIZE	(23552 * AMDGPU_MAX_VCE_HANDLES)
+#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK	0x02
 
 static void vce_v2_0_mc_resume(struct amdgpu_device *adev);
 static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev);
 static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev);
+static int vce_v2_0_wait_for_idle(void *handle);
 
 /**
  * vce_v2_0_ring_get_rptr - get read pointer
@@ -96,6 +98,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
 		WREG32(mmVCE_RB_WPTR2, ring->wptr);
 }
 
+static int vce_v2_0_lmi_clean(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			uint32_t status = RREG32(mmVCE_LMI_STATUS);
+
+			if (status & 0x337f)
+				return 0;
+			mdelay(10);
+		}
+	}
+
+	return -ETIMEDOUT;
+}
+
+static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			uint32_t status = RREG32(mmVCE_STATUS);
+
+			if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK)
+				return 0;
+			mdelay(10);
+		}
+
+		DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
+		WREG32_P(mmVCE_SOFT_RESET,
+			VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+		WREG32_P(mmVCE_SOFT_RESET, 0,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+	}
+
+	return -ETIMEDOUT;
+}
+
 /**
  * vce_v2_0_start - start VCE block
  *
@@ -106,7 +151,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
 static int vce_v2_0_start(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	int i, j, r;
+	int r;
 
 	vce_v2_0_mc_resume(adev);
 
@@ -137,26 +182,7 @@ static int vce_v2_0_start(struct amdgpu_device *adev)
 
 	WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
 
-	for (i = 0; i < 10; ++i) {
-		uint32_t status;
-		for (j = 0; j < 100; ++j) {
-			status = RREG32(mmVCE_STATUS);
-			if (status & 2)
-				break;
-			mdelay(10);
-		}
-		r = 0;
-		if (status & 2)
-			break;
-
-		DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
-		WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
-				~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-		mdelay(10);
-		WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-		mdelay(10);
-		r = -1;
-	}
+	r = vce_v2_0_firmware_loaded(adev);
 
 	/* clear BUSY flag */
 	WREG32_P(mmVCE_STATUS, 0, ~1);
@@ -338,31 +364,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated)
 
 static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated)
 {
-	u32 orig, tmp;
+	if (vce_v2_0_wait_for_idle(adev)) {
+		DRM_INFO("VCE is busy, Can't set clock gateing");
+		return;
+	}
 
-	tmp = RREG32(mmVCE_CLOCK_GATING_B);
-	tmp &= ~0x00060006;
+	WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100);
+
+	if (vce_v2_0_lmi_clean(adev)) {
+		DRM_INFO("LMI is busy, Can't set clock gateing");
+		return;
+	}
+
+	WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+	WREG32_P(mmVCE_SOFT_RESET,
+		 VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+		 ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+	WREG32(mmVCE_STATUS, 0);
+
+	if (gated)
+		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+	/* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */
 	if (gated) {
-		tmp |= 0xe10000;
+		/* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */
+		WREG32(mmVCE_CLOCK_GATING_B, 0xe90010);
 	} else {
-		tmp |= 0xe1;
-		tmp &= ~0xe10000;
+		/* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */
+		WREG32(mmVCE_CLOCK_GATING_B, 0x800f1);
 	}
-	WREG32(mmVCE_CLOCK_GATING_B, tmp);
 
-	orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
-	tmp &= ~0x1fe000;
-	tmp &= ~0xff000000;
-	if (tmp != orig)
-		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+	/* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/;
+	WREG32(mmVCE_UENC_CLOCK_GATING, 0x40);
 
-	orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
-	tmp &= ~0x3fc;
-	if (tmp != orig)
-		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+	/* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */
+	WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00);
 
-	if (gated)
-		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+	WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100);
+	if(!gated) {
+		WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+		mdelay(100);
+		WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+		vce_v2_0_firmware_loaded(adev);
+		WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK);
+	}
 }
 
 static void vce_v2_0_disable_cg(struct amdgpu_device *adev)
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1471524131-5357-1-git-send-email-Jim.Qu-5C7GfCeVMHo@public.gmane.org

#2 lore raw

Am 18.08.2016 um 14:42 schrieb jimqu:

Quoted 7 lines

phenomenon: software hang when device resume back, read UVD fence is 0xffffffff and read pcie pid is 0xffff.
the issue is caused by VCE reset when update cg setting. according to HW programming guide, adjust update VCE cg sequence.

Change-Id I18b12eea21c045908cdd23f93a0b196b87bfed6c

You should probably note in the commit log that this only applies to systems with VCE2, e.g. CIK.

Additional to that please make sure that it applies cleanly to the amd-staging-4.6 branch.

With that made sure the patch is Reviewed-by: Christian König <[email protected]>

Regards,
Christian.

Quoted 167 lines
---
  drivers/gpu/drm/amd/amdgpu/vce_v2_0.c | 123 +++++++++++++++++++++++-----------
  1 file changed, 84 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +77 -39
index 9e70df9..6ce7c07 100644
@@ -40,10 +40,12 @@
  #define VCE_V2_0_FW_SIZE	(256 * 1024)
  #define VCE_V2_0_STACK_SIZE	(64 * 1024)
  #define VCE_V2_0_DATA_SIZE	(23552 * AMDGPU_MAX_VCE_HANDLES)
+#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK	0x02
 
  static void vce_v2_0_mc_resume(struct amdgpu_device *adev);
  static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev);
  static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev);
+static int vce_v2_0_wait_for_idle(void *handle);
 
  /**
   * vce_v2_0_ring_get_rptr - get read pointer
@@ -96,6 +98,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
  		WREG32(mmVCE_RB_WPTR2, ring->wptr);
  }
 
+static int vce_v2_0_lmi_clean(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			uint32_t status = RREG32(mmVCE_LMI_STATUS);
+
+			if (status & 0x337f)
+				return 0;
+			mdelay(10);
+		}
+	}
+
+	return -ETIMEDOUT;
+}
+
+static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	for (i = 0; i < 10; ++i) {
+		for (j = 0; j < 100; ++j) {
+			uint32_t status = RREG32(mmVCE_STATUS);
+
+			if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK)
+				return 0;
+			mdelay(10);
+		}
+
+		DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
+		WREG32_P(mmVCE_SOFT_RESET,
+			VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+		WREG32_P(mmVCE_SOFT_RESET, 0,
+			~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+	}
+
+	return -ETIMEDOUT;
+}
+
  /**
   * vce_v2_0_start - start VCE block
   *
@@ -106,7 +151,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
  static int vce_v2_0_start(struct amdgpu_device *adev)
  {
  	struct amdgpu_ring *ring;
-	int i, j, r;
+	int r;
 
  	vce_v2_0_mc_resume(adev);
 
@@ -137,26 +182,7 @@ static int vce_v2_0_start(struct amdgpu_device *adev)
 
  	WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
 
-	for (i = 0; i < 10; ++i) {
-		uint32_t status;
-		for (j = 0; j < 100; ++j) {
-			status = RREG32(mmVCE_STATUS);
-			if (status & 2)
-				break;
-			mdelay(10);
-		}
-		r = 0;
-		if (status & 2)
-			break;
-
-		DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
-		WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
-				~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-		mdelay(10);
-		WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-		mdelay(10);
-		r = -1;
-	}
+	r = vce_v2_0_firmware_loaded(adev);
 
  	/* clear BUSY flag */
  	WREG32_P(mmVCE_STATUS, 0, ~1);
@@ -338,31 +364,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated)
 
  static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated)
  {
-	u32 orig, tmp;
+	if (vce_v2_0_wait_for_idle(adev)) {
+		DRM_INFO("VCE is busy, Can't set clock gateing");
+		return;
+	}
 
-	tmp = RREG32(mmVCE_CLOCK_GATING_B);
-	tmp &= ~0x00060006;
+	WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100);
+
+	if (vce_v2_0_lmi_clean(adev)) {
+		DRM_INFO("LMI is busy, Can't set clock gateing");
+		return;
+	}
+
+	WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+	WREG32_P(mmVCE_SOFT_RESET,
+		 VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+		 ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+	WREG32(mmVCE_STATUS, 0);
+
+	if (gated)
+		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+	/* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */
  	if (gated) {
-		tmp |= 0xe10000;
+		/* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */
+		WREG32(mmVCE_CLOCK_GATING_B, 0xe90010);
  	} else {
-		tmp |= 0xe1;
-		tmp &= ~0xe10000;
+		/* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */
+		WREG32(mmVCE_CLOCK_GATING_B, 0x800f1);
  	}
-	WREG32(mmVCE_CLOCK_GATING_B, tmp);
 
-	orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
-	tmp &= ~0x1fe000;
-	tmp &= ~0xff000000;
-	if (tmp != orig)
-		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+	/* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/;
+	WREG32(mmVCE_UENC_CLOCK_GATING, 0x40);
 
-	orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
-	tmp &= ~0x3fc;
-	if (tmp != orig)
-		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+	/* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */
+	WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00);
 
-	if (gated)
-		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+	WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100);
+	if(!gated) {

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) · line 413

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +7 -0
+		WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+		mdelay(100);
+		WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+		vce_v2_0_firmware_loaded(adev);
+		WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK);
+	}
  }
 
  static void vce_v2_0_disable_cg(struct amdgpu_device *adev)

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 8b1a399d-0054-c356-8775-60cdd83d5004-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org

#4 lore raw

do we need to check those status(lmi status,firmware loaded, vce status and etc) when vce suspend?

it is weird that we didn't implement vce suspend function as hw spec suggested.

Best Regards

Rex

________________________________
From: amd-gfx <amd-gfx-bounces-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org> on behalf of Christian König <[email protected]>
Sent: Thursday, August 18, 2016 9:27:57 PM
To: Qu, Jim; [email protected]
Subject: Re: [PATCH] drm/amd/amdgpu: S3 resumed failed after 4-5 times loop

Am 18.08.2016 um 14:42 schrieb jimqu:

Quoted 7 lines

phenomenon: software hang when device resume back, read UVD fence is 0xffffffff and read pcie pid is 0xffff.
the issue is caused by VCE reset when update cg setting. according to HW programming guide, adjust update VCE cg sequence.

Change-Id I18b12eea21c045908cdd23f93a0b196b87bfed6c

You should probably note in the commit log that this only applies to systems with VCE2, e.g. CIK.

Additional to that please make sure that it applies cleanly to the amd-staging-4.6 branch.

With that made sure the patch is Reviewed-by: Christian König <[email protected]>

Regards,
Christian.

Quoted 167 lines
---
  drivers/gpu/drm/amd/amdgpu/vce_v2_0.c | 123 +++++++++++++++++++++++-----------
  1 file changed, 84 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +77 -39
index 9e70df9..6ce7c07 100644
@@ -40,10 +40,12 @@
  #define VCE_V2_0_FW_SIZE    (256 * 1024)
  #define VCE_V2_0_STACK_SIZE (64 * 1024)
  #define VCE_V2_0_DATA_SIZE  (23552 * AMDGPU_MAX_VCE_HANDLES)
+#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK        0x02
 
  static void vce_v2_0_mc_resume(struct amdgpu_device *adev);
  static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev);
  static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev);
+static int vce_v2_0_wait_for_idle(void *handle);
 
  /**
   * vce_v2_0_ring_get_rptr - get read pointer
@@ -96,6 +98,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
               WREG32(mmVCE_RB_WPTR2, ring->wptr);
  }
 
+static int vce_v2_0_lmi_clean(struct amdgpu_device *adev)
+{
+     int i, j;
+
+     for (i = 0; i < 10; ++i) {
+             for (j = 0; j < 100; ++j) {
+                     uint32_t status = RREG32(mmVCE_LMI_STATUS);
+
+                     if (status & 0x337f)
+                             return 0;
+                     mdelay(10);
+             }
+     }
+
+     return -ETIMEDOUT;
+}
+
+static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev)
+{
+     int i, j;
+
+     for (i = 0; i < 10; ++i) {
+             for (j = 0; j < 100; ++j) {
+                     uint32_t status = RREG32(mmVCE_STATUS);
+
+                     if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK)
+                             return 0;
+                     mdelay(10);
+             }
+
+             DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
+             WREG32_P(mmVCE_SOFT_RESET,
+                     VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+                     ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+             mdelay(10);
+             WREG32_P(mmVCE_SOFT_RESET, 0,
+                     ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+             mdelay(10);
+     }
+
+     return -ETIMEDOUT;
+}
+
  /**
   * vce_v2_0_start - start VCE block
   *
@@ -106,7 +151,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
  static int vce_v2_0_start(struct amdgpu_device *adev)
  {
       struct amdgpu_ring *ring;
-     int i, j, r;
+     int r;
 
       vce_v2_0_mc_resume(adev);
 
@@ -137,26 +182,7 @@ static int vce_v2_0_start(struct amdgpu_device *adev)
 
       WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
 
-     for (i = 0; i < 10; ++i) {
-             uint32_t status;
-             for (j = 0; j < 100; ++j) {
-                     status = RREG32(mmVCE_STATUS);
-                     if (status & 2)
-                             break;
-                     mdelay(10);
-             }
-             r = 0;
-             if (status & 2)
-                     break;
-
-             DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
-             WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
-                             ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-             mdelay(10);
-             WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-             mdelay(10);
-             r = -1;
-     }
+     r = vce_v2_0_firmware_loaded(adev);
 
       /* clear BUSY flag */
       WREG32_P(mmVCE_STATUS, 0, ~1);
@@ -338,31 +364,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated)
 
  static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated)
  {
-     u32 orig, tmp;
+     if (vce_v2_0_wait_for_idle(adev)) {
+             DRM_INFO("VCE is busy, Can't set clock gateing");
+             return;
+     }
 
-     tmp = RREG32(mmVCE_CLOCK_GATING_B);
-     tmp &= ~0x00060006;
+     WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100);
+
+     if (vce_v2_0_lmi_clean(adev)) {
+             DRM_INFO("LMI is busy, Can't set clock gateing");
+             return;
+     }
+
+     WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+     WREG32_P(mmVCE_SOFT_RESET,
+              VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+              ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+     WREG32(mmVCE_STATUS, 0);
+
+     if (gated)
+             WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+     /* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */
       if (gated) {
-             tmp |= 0xe10000;
+             /* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */
+             WREG32(mmVCE_CLOCK_GATING_B, 0xe90010);
       } else {
-             tmp |= 0xe1;
-             tmp &= ~0xe10000;
+             /* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */
+             WREG32(mmVCE_CLOCK_GATING_B, 0x800f1);
       }
-     WREG32(mmVCE_CLOCK_GATING_B, tmp);
 
-     orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
-     tmp &= ~0x1fe000;
-     tmp &= ~0xff000000;
-     if (tmp != orig)
-             WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+     /* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/;
+     WREG32(mmVCE_UENC_CLOCK_GATING, 0x40);
 
-     orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
-     tmp &= ~0x3fc;
-     if (tmp != orig)
-             WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+     /* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */
+     WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00);
 
-     if (gated)
-             WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+     WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100);
+     if(!gated) {

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) · line 413

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +7 -0
+             WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+             mdelay(100);
+             WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+             vce_v2_0_firmware_loaded(adev);
+             WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK);
+     }
  }
 
  static void vce_v2_0_disable_cg(struct amdgpu_device *adev)

_______________________________________________
amd-gfx mailing list [email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): CY4PR12MB1687869929F2F96EF9541E0BFB150-rpdhrqHFk06Y0SjTqZDccQdYzm3356FpvxpqHgZTriW3zl9H0oFU5g@public.gmane.org

author
#6 lore raw

if it is required, we can continue to optimize driver in the future.

Thanks
JimQu

________________________________________
发件人: Zhu, Rex
发送时间: 2016年8月18日 23:24:19
收件人: Christian König; Qu, Jim; [email protected] 主题: Re: [PATCH] drm/amd/amdgpu: S3 resumed failed after 4-5 times loop

do we need to check those status(lmi status,firmware loaded, vce status and etc) when vce suspend?

it is weird that we didn't implement vce suspend function as hw spec suggested.

Best Regards

Rex

________________________________
From: amd-gfx <[email protected]> on behalf of Christian König <[email protected]>
Sent: Thursday, August 18, 2016 9:27:57 PM
To: Qu, Jim; [email protected]
Subject: Re: [PATCH] drm/amd/amdgpu: S3 resumed failed after 4-5 times loop

Am 18.08.2016 um 14:42 schrieb jimqu:

Quoted 7 lines

phenomenon: software hang when device resume back, read UVD fence is 0xffffffff and read pcie pid is 0xffff.
the issue is caused by VCE reset when update cg setting. according to HW programming guide, adjust update VCE cg sequence.

Change-Id I18b12eea21c045908cdd23f93a0b196b87bfed6c

You should probably note in the commit log that this only applies to systems with VCE2, e.g. CIK.

Additional to that please make sure that it applies cleanly to the amd-staging-4.6 branch.

With that made sure the patch is Reviewed-by: Christian König <[email protected]>

Regards,
Christian.

Quoted 167 lines
---
  drivers/gpu/drm/amd/amdgpu/vce_v2_0.c | 123 +++++++++++++++++++++++-----------
  1 file changed, 84 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +77 -39
index 9e70df9..6ce7c07 100644
@@ -40,10 +40,12 @@
  #define VCE_V2_0_FW_SIZE    (256 * 1024)
  #define VCE_V2_0_STACK_SIZE (64 * 1024)
  #define VCE_V2_0_DATA_SIZE  (23552 * AMDGPU_MAX_VCE_HANDLES)
+#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK        0x02
 
  static void vce_v2_0_mc_resume(struct amdgpu_device *adev);
  static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev);
  static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev);
+static int vce_v2_0_wait_for_idle(void *handle);
 
  /**
   * vce_v2_0_ring_get_rptr - get read pointer
@@ -96,6 +98,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
               WREG32(mmVCE_RB_WPTR2, ring->wptr);
  }
 
+static int vce_v2_0_lmi_clean(struct amdgpu_device *adev)
+{
+     int i, j;
+
+     for (i = 0; i < 10; ++i) {
+             for (j = 0; j < 100; ++j) {
+                     uint32_t status = RREG32(mmVCE_LMI_STATUS);
+
+                     if (status & 0x337f)
+                             return 0;
+                     mdelay(10);
+             }
+     }
+
+     return -ETIMEDOUT;
+}
+
+static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev)
+{
+     int i, j;
+
+     for (i = 0; i < 10; ++i) {
+             for (j = 0; j < 100; ++j) {
+                     uint32_t status = RREG32(mmVCE_STATUS);
+
+                     if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK)
+                             return 0;
+                     mdelay(10);
+             }
+
+             DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
+             WREG32_P(mmVCE_SOFT_RESET,
+                     VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+                     ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+             mdelay(10);
+             WREG32_P(mmVCE_SOFT_RESET, 0,
+                     ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+             mdelay(10);
+     }
+
+     return -ETIMEDOUT;
+}
+
  /**
   * vce_v2_0_start - start VCE block
   *
@@ -106,7 +151,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
  static int vce_v2_0_start(struct amdgpu_device *adev)
  {
       struct amdgpu_ring *ring;
-     int i, j, r;
+     int r;
 
       vce_v2_0_mc_resume(adev);
 
@@ -137,26 +182,7 @@ static int vce_v2_0_start(struct amdgpu_device *adev)
 
       WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
 
-     for (i = 0; i < 10; ++i) {
-             uint32_t status;
-             for (j = 0; j < 100; ++j) {
-                     status = RREG32(mmVCE_STATUS);
-                     if (status & 2)
-                             break;
-                     mdelay(10);
-             }
-             r = 0;
-             if (status & 2)
-                     break;
-
-             DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
-             WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
-                             ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-             mdelay(10);
-             WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
-             mdelay(10);
-             r = -1;
-     }
+     r = vce_v2_0_firmware_loaded(adev);
 
       /* clear BUSY flag */
       WREG32_P(mmVCE_STATUS, 0, ~1);
@@ -338,31 +364,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated)
 
  static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated)
  {
-     u32 orig, tmp;
+     if (vce_v2_0_wait_for_idle(adev)) {
+             DRM_INFO("VCE is busy, Can't set clock gateing");
+             return;
+     }
 
-     tmp = RREG32(mmVCE_CLOCK_GATING_B);
-     tmp &= ~0x00060006;
+     WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100);
+
+     if (vce_v2_0_lmi_clean(adev)) {
+             DRM_INFO("LMI is busy, Can't set clock gateing");
+             return;
+     }
+
+     WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+     WREG32_P(mmVCE_SOFT_RESET,
+              VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+              ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+     WREG32(mmVCE_STATUS, 0);
+
+     if (gated)
+             WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+     /* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */
       if (gated) {
-             tmp |= 0xe10000;
+             /* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */
+             WREG32(mmVCE_CLOCK_GATING_B, 0xe90010);
       } else {
-             tmp |= 0xe1;
-             tmp &= ~0xe10000;
+             /* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */
+             WREG32(mmVCE_CLOCK_GATING_B, 0x800f1);
       }
-     WREG32(mmVCE_CLOCK_GATING_B, tmp);
 
-     orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
-     tmp &= ~0x1fe000;
-     tmp &= ~0xff000000;
-     if (tmp != orig)
-             WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+     /* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/;
+     WREG32(mmVCE_UENC_CLOCK_GATING, 0x40);
 
-     orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
-     tmp &= ~0x3fc;
-     if (tmp != orig)
-             WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+     /* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */
+     WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00);
 
-     if (gated)
-             WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+     WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100);
+     if(!gated) {

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) · line 413

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +7 -0
+             WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+             mdelay(100);
+             WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+             vce_v2_0_firmware_loaded(adev);
+             WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK);
+     }
  }
 
  static void vce_v2_0_disable_cg(struct amdgpu_device *adev)

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx