[PATCH v2 2/3] drm/amd: Poll for GFX core to be off

[Mario Limonciello]

If GFXOFF was flushed during suspend entry it may take some time
for GFX core to be powered down.  Ensure that it's powered off
before continuing any operations that may try to utilize related
IP. This avoids hangs from stopping RLC as well as problems with
fence interrupts timing out during s2idle entry and exit.

Cc: stable at vger.kernel.org # 6.1
Suggested-by: Tim Huang <tim.huang at amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello at amd.com>
---
v1->v2:
 * Only poll in the s0ix case not all GFXOFF cases
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 24 ++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c     | 18 ++++++++++++++++
 drivers/gpu/drm/amd/include/amd_shared.h   |  1 +
 3 files changed, 43 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 059139f1f973..59d5fc65276c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3063,6 +3063,26 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
 		adev->gfx.gfx_off_state = true;
 }
 
+static int amdgpu_device_ensure_gfx_off(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	if (!adev->in_s0ix)
+		return 0;
+
+	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
+		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_GFX)
+			continue;
+		if (!adev->ip_blocks[i].version->funcs->wait_for_off)
+			continue;
+		r = adev->ip_blocks[i].version->funcs->wait_for_off((void *)adev);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_device_ip_suspend_phase1 - run suspend for hardware IPs (phase 1)
  *
@@ -4318,6 +4338,10 @@ int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
 	cancel_delayed_work_sync(&adev->delayed_init_work);
 	flush_delayed_work(&adev->gfx.gfx_off_delay_work);
 
+	r = amdgpu_device_ensure_gfx_off(adev);
+	if (r)
+		return r;
+
 	amdgpu_ras_suspend(adev);
 
 	amdgpu_device_ip_suspend_phase1(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 4b7224de879e..dcbdb2641086 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -4434,6 +4434,23 @@ static int gfx_v11_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
+
+static int gfx_v11_0_wait_for_off(void *handle)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	u32 tmp;
+	int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regGFX_IMU_MSG_FLAGS);
+		if (!(tmp & 0x06))
+			return 0;
+		udelay(1);
+	}
+	dev_dbg(adev->dev, "GFX IMU is %x\n", tmp);
+	return -ETIMEDOUT;
+}
+
 static int gfx_v11_0_soft_reset(void *handle)
 {
 	u32 grbm_soft_reset = 0;
@@ -6109,6 +6126,7 @@ static const struct amd_ip_funcs gfx_v11_0_ip_funcs = {
 	.resume = gfx_v11_0_resume,
 	.is_idle = gfx_v11_0_is_idle,
 	.wait_for_idle = gfx_v11_0_wait_for_idle,
+	.wait_for_off = gfx_v11_0_wait_for_off,
 	.soft_reset = gfx_v11_0_soft_reset,
 	.check_soft_reset = gfx_v11_0_check_soft_reset,
 	.post_soft_reset = gfx_v11_0_post_soft_reset,
diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h
index f175e65b853a..ce2e2b6fd6ff 100644
--- a/drivers/gpu/drm/amd/include/amd_shared.h
+++ b/drivers/gpu/drm/amd/include/amd_shared.h
@@ -298,6 +298,7 @@ struct amd_ip_funcs {
 	int (*resume)(void *handle);
 	bool (*is_idle)(void *handle);
 	int (*wait_for_idle)(void *handle);
+	int (*wait_for_off)(void *handle);
 	bool (*check_soft_reset)(void *handle);
 	int (*pre_soft_reset)(void *handle);
 	int (*soft_reset)(void *handle);
-- 
2.34.1