drm/radeon/kms: Add support for RLC init on SI

RLC handles the interrupt controller and other tasks
on the GPU.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

3 files changed, 162 insertions, 0 deletions

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index f1c2f58..8275b41 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -772,6 +772,18 @@ struct r600_blit {
 
 void r600_blit_suspend(struct radeon_device *rdev);
 
+/*
+ * SI RLC stuff
+ */
+struct si_rlc {
+	/* for power gating */
+	struct radeon_bo	*save_restore_obj;
+	uint64_t		save_restore_gpu_addr;
+	/* for clear state */
+	struct radeon_bo	*clear_state_obj;
+	uint64_t		clear_state_gpu_addr;
+};
+
 int radeon_ib_get(struct radeon_device *rdev, int ring,
 		  struct radeon_ib **ib, unsigned size);
 void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib **ib);
@@ -1532,6 +1544,7 @@ struct radeon_device {
 	struct r600_vram_scratch vram_scratch;
 	int msi_enabled; /* msi enabled */
 	struct r600_ih ih; /* r6/700 interrupt ring */
+	struct si_rlc rlc;
 	struct work_struct hotplug_work;
 	int num_crtc; /* number of crtcs */
 	struct mutex dc_hw_i2c_mutex; /* display controller hw i2c mutex */
diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c
index 408119a..4252cd0 100644
--- a/drivers/gpu/drm/radeon/si.c
+++ b/drivers/gpu/drm/radeon/si.c
@@ -2940,3 +2940,135 @@ void si_vm_tlb_flush(struct radeon_device *rdev, struct radeon_vm *vm)
 	WREG32(VM_INVALIDATE_REQUEST, 1 << vm->id);
 }
 
+/*
+ * RLC
+ */
+static void si_rlc_fini(struct radeon_device *rdev)
+{
+	int r;
+
+	/* save restore block */
+	if (rdev->rlc.save_restore_obj) {
+		r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(rdev->dev, "(%d) reserve RLC sr bo failed\n", r);
+		radeon_bo_unpin(rdev->rlc.save_restore_obj);
+		radeon_bo_unreserve(rdev->rlc.save_restore_obj);
+
+		radeon_bo_unref(&rdev->rlc.save_restore_obj);
+		rdev->rlc.save_restore_obj = NULL;
+	}
+
+	/* clear state block */
+	if (rdev->rlc.clear_state_obj) {
+		r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(rdev->dev, "(%d) reserve RLC c bo failed\n", r);
+		radeon_bo_unpin(rdev->rlc.clear_state_obj);
+		radeon_bo_unreserve(rdev->rlc.clear_state_obj);
+
+		radeon_bo_unref(&rdev->rlc.clear_state_obj);
+		rdev->rlc.clear_state_obj = NULL;
+	}
+}
+
+static int si_rlc_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* save restore block */
+	if (rdev->rlc.save_restore_obj == NULL) {
+		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true,
+				RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.save_restore_obj);
+		if (r) {
+			dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r);
+			return r;
+		}
+	}
+
+	r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false);
+	if (unlikely(r != 0)) {
+		si_rlc_fini(rdev);
+		return r;
+	}
+	r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM,
+			  &rdev->rlc.save_restore_gpu_addr);
+	if (r) {
+		radeon_bo_unreserve(rdev->rlc.save_restore_obj);
+		dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r);
+		si_rlc_fini(rdev);
+		return r;
+	}
+
+	/* clear state block */
+	if (rdev->rlc.clear_state_obj == NULL) {
+		r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true,
+				RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.clear_state_obj);
+		if (r) {
+			dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r);
+			si_rlc_fini(rdev);
+			return r;
+		}
+	}
+	r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false);
+	if (unlikely(r != 0)) {
+		si_rlc_fini(rdev);
+		return r;
+	}
+	r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM,
+			  &rdev->rlc.clear_state_gpu_addr);
+	if (r) {
+
+		radeon_bo_unreserve(rdev->rlc.clear_state_obj);
+		dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r);
+		si_rlc_fini(rdev);
+		return r;
+	}
+
+	return 0;
+}
+
+static void si_rlc_stop(struct radeon_device *rdev)
+{
+	WREG32(RLC_CNTL, 0);
+}
+
+static void si_rlc_start(struct radeon_device *rdev)
+{
+	WREG32(RLC_CNTL, RLC_ENABLE);
+}
+
+static int si_rlc_resume(struct radeon_device *rdev)
+{
+	u32 i;
+	const __be32 *fw_data;
+
+	if (!rdev->rlc_fw)
+		return -EINVAL;
+
+	si_rlc_stop(rdev);
+
+	WREG32(RLC_RL_BASE, 0);
+	WREG32(RLC_RL_SIZE, 0);
+	WREG32(RLC_LB_CNTL, 0);
+	WREG32(RLC_LB_CNTR_MAX, 0xffffffff);
+	WREG32(RLC_LB_CNTR_INIT, 0);
+
+	WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8);
+	WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8);
+
+	WREG32(RLC_MC_CNTL, 0);
+	WREG32(RLC_UCODE_CNTL, 0);
+
+	fw_data = (const __be32 *)rdev->rlc_fw->data;
+	for (i = 0; i < SI_RLC_UCODE_SIZE; i++) {
+		WREG32(RLC_UCODE_ADDR, i);
+		WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++));
+	}
+	WREG32(RLC_UCODE_ADDR, 0);
+
+	si_rlc_start(rdev);
+
+	return 0;
+}
+
diff --git a/drivers/gpu/drm/radeon/sid.h b/drivers/gpu/drm/radeon/sid.h
index 904c4fd..78b5649 100644
--- a/drivers/gpu/drm/radeon/sid.h
+++ b/drivers/gpu/drm/radeon/sid.h
@@ -513,6 +513,23 @@
 
 #define	CP_DEBUG					0xC1FC
 
+#define RLC_CNTL                                          0xC300
+#       define RLC_ENABLE                                 (1 << 0)
+#define RLC_RL_BASE                                       0xC304
+#define RLC_RL_SIZE                                       0xC308
+#define RLC_LB_CNTL                                       0xC30C
+#define RLC_SAVE_AND_RESTORE_BASE                         0xC310
+#define RLC_LB_CNTR_MAX                                   0xC314
+#define RLC_LB_CNTR_INIT                                  0xC318
+
+#define RLC_CLEAR_STATE_RESTORE_BASE                      0xC320
+
+#define RLC_UCODE_ADDR                                    0xC32C
+#define RLC_UCODE_DATA                                    0xC330
+
+#define RLC_MC_CNTL                                       0xC344
+#define RLC_UCODE_CNTL                                    0xC348
+
 #define VGT_EVENT_INITIATOR                             0x28a90
 #       define SAMPLE_STREAMOUTSTATS1                   (1 << 0)
 #       define SAMPLE_STREAMOUTSTATS2                   (2 << 0)