drm/amdgpu: nuke dynamic gfx scratch reg allocation

It's over a decade ago that this was actually used for more than ring and
IB tests. Just use the static register directly where needed and nuke the
now useless infrastructure.

Signed-off-by: Christian König <[e-mail]>
Acked-by: Lang Yu <[e-mail]>
Acked-by: Alex Deucher <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c

···
 }
 
 /**
- * amdgpu_gfx_scratch_get - Allocate a scratch register
- *
- * @adev: amdgpu_device pointer
- * @reg: scratch register mmio offset
- *
- * Allocate a CP scratch register for use by the driver (all asics).
- * Returns 0 on success or -EINVAL on failure.
- */
-int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg)
-{
-	int i;
-
-	i = ffs(adev->gfx.scratch.free_mask);
-	if (i != 0 && i <= adev->gfx.scratch.num_reg) {
-		i--;
-		adev->gfx.scratch.free_mask &= ~(1u << i);
-		*reg = adev->gfx.scratch.reg_base + i;
-		return 0;
-	}
-	return -EINVAL;
-}
-
-/**
- * amdgpu_gfx_scratch_free - Free a scratch register
- *
- * @adev: amdgpu_device pointer
- * @reg: scratch register mmio offset
- *
- * Free a CP scratch register allocated for use by the driver (all asics)
- */
-void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg)
-{
-	adev->gfx.scratch.free_mask |= 1u << (reg - adev->gfx.scratch.reg_base);
-}
-
-/**
  * amdgpu_gfx_parse_disable_cu - Parse the disable_cu module parameter
  *
  * @mask: array in which the per-shader array disable masks will be stored

drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h

···
 };
 
 /*
- * GPU scratch registers structures, functions & helpers
- */
-struct amdgpu_scratch {
-	unsigned		num_reg;
-	uint32_t                reg_base;
-	uint32_t		free_mask;
-};
-
-/*
  * GFX configurations
  */
 #define AMDGPU_GFX_MAX_SE 4
···
 	struct amdgpu_mec		mec;
 	struct amdgpu_kiq		kiq;
 	struct amdgpu_imu		imu;
-	struct amdgpu_scratch		scratch;
 	bool				rs64_enable; /* firmware format */
 	const struct firmware		*me_fw;	/* ME firmware */
 	uint32_t			me_fw_version;
···
 {
 	return (u32)((1ULL << bit_width) - 1);
 }
-
-int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg);
-void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg);
 
 void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se,
 				 unsigned max_sh);

drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c

···
 	gfx_v10_0_init_spm_golden_registers(adev);
 }
 
-static void gfx_v10_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 static void gfx_v10_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
 				       bool wc, uint32_t reg, uint32_t val)
 {
···
 static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r) {
-		DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
-		return r;
-	}
-
-	WREG32(scratch, 0xCAFEDEAD);
-
+	WREG32_SOC15(GC, 0, mmSCRATCH_REG0, 0xCAFEDEAD);
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r) {
 		DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
 			  ring->idx, r);
-		amdgpu_gfx_scratch_free(adev, scratch);
 		return r;
 	}
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0) -
+			  PACKET3_SET_UCONFIG_REG_START);
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
+		tmp = RREG32_SOC15(GC, 0, mmSCRATCH_REG0);
 		if (tmp == 0xDEADBEEF)
 			break;
 		if (amdgpu_emu_mode == 1)
···
 
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
-
-	amdgpu_gfx_scratch_free(adev, scratch);
 
 	return r;
 }
···
 		return r;
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
-
-	gfx_v10_0_scratch_init(adev);
 
 	r = gfx_v10_0_me_init(adev);
 	if (r)

drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c

···
 	gfx_v11_0_init_spm_golden_registers(adev);
 }
 
-static void gfx_v11_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 static void gfx_v11_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
 				       bool wc, uint32_t reg, uint32_t val)
 {
···
 static int gfx_v11_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
+	uint32_t scratch = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r) {
-		DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
-		return r;
-	}
-
 	WREG32(scratch, 0xCAFEDEAD);
-
 	r = amdgpu_ring_alloc(ring, 5);
 	if (r) {
 		DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
 			  ring->idx, r);
-		amdgpu_gfx_scratch_free(adev, scratch);
 		return r;
 	}
 
···
 		gfx_v11_0_ring_emit_wreg(ring, scratch, 0xDEADBEEF);
 	} else {
 		amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-		amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+		amdgpu_ring_write(ring, scratch -
+				  PACKET3_SET_UCONFIG_REG_START);
 		amdgpu_ring_write(ring, 0xDEADBEEF);
 	}
 	amdgpu_ring_commit(ring);
···
 
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
-
-	amdgpu_gfx_scratch_free(adev, scratch);
-
 	return r;
 }
 
···
 		return r;
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
-
-	gfx_v11_0_scratch_init(adev);
 
 	if (adev->gfx.imu.funcs) {
 		if (adev->gfx.imu.funcs->init_microcode) {

drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c

···
 	udelay(50);
 }
 
-
-static void gfx_v6_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = mmSCRATCH_REG0;
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
-	if (r)
-		return r;
-
-	WREG32(scratch, 0xCAFEDEAD);
+	WREG32(mmSCRATCH_REG0, 0xCAFEDEAD);
 
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
-		goto error_free_scratch;
+		return r;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
-	amdgpu_ring_write(ring, (scratch - PACKET3_SET_CONFIG_REG_START));
+	amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_CONFIG_REG_START);
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
+		tmp = RREG32(mmSCRATCH_REG0);
 		if (tmp == 0xDEADBEEF)
 			break;
 		udelay(1);
···
 
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
-
-error_free_scratch:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 {
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_ib ib;
 	struct dma_fence *f = NULL;
-	uint32_t scratch;
+	struct amdgpu_ib ib;
 	uint32_t tmp = 0;
 	long r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	WREG32(mmSCRATCH_REG0, 0xCAFEDEAD);
+	memset(&ib, 0, sizeof(ib));
+	r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		return r;
 
-	WREG32(scratch, 0xCAFEDEAD);
-	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 256,
-					AMDGPU_IB_POOL_DIRECT, &ib);
-	if (r)
-		goto err1;
-
 	ib.ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1);
-	ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_START));
+	ib.ptr[1] = mmSCRATCH_REG0 - PACKET3_SET_CONFIG_REG_START;
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
 
 	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
-		goto err2;
+		goto error;
 
 	r = dma_fence_wait_timeout(f, false, timeout);
 	if (r == 0) {
 		r = -ETIMEDOUT;
-		goto err2;
+		goto error;
 	} else if (r < 0) {
-		goto err2;
+		goto error;
 	}
-	tmp = RREG32(scratch);
+	tmp = RREG32(mmSCRATCH_REG0);
 	if (tmp == 0xDEADBEEF)
 		r = 0;
 	else
 		r = -EINVAL;
 
-err2:
+error:
 	amdgpu_ib_free(adev, &ib, NULL);
 	dma_fence_put(f);
-err1:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq);
 	if (r)
 		return r;
-
-	gfx_v6_0_scratch_init(adev);
 
 	r = gfx_v6_0_init_microcode(adev);
 	if (r) {

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

···
 	udelay(50);
 }
 
-/*
- * GPU scratch registers helpers function.
- */
-/**
- * gfx_v7_0_scratch_init - setup driver info for CP scratch regs
- *
- * @adev: amdgpu_device pointer
- *
- * Set up the number and offset of the CP scratch registers.
- * NOTE: use of CP scratch registers is a legacy interface and
- * is not used by default on newer asics (r6xx+).  On newer asics,
- * memory buffers are used for fences rather than scratch regs.
- */
-static void gfx_v7_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = mmSCRATCH_REG0;
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 /**
  * gfx_v7_0_ring_test_ring - basic gfx ring test
  *
···
 static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	WREG32(mmSCRATCH_REG0, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
 
-	WREG32(scratch, 0xCAFEDEAD);
-	r = amdgpu_ring_alloc(ring, 3);
-	if (r)
-		goto error_free_scratch;
-
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START);
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
+		tmp = RREG32(mmSCRATCH_REG0);
 		if (tmp == 0xDEADBEEF)
 			break;
 		udelay(1);
 	}
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
-
-error_free_scratch:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib ib;
 	struct dma_fence *f = NULL;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	long r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	WREG32(mmSCRATCH_REG0, 0xCAFEDEAD);
+	memset(&ib, 0, sizeof(ib));
+	r = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		return r;
 
-	WREG32(scratch, 0xCAFEDEAD);
-	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 256,
-					AMDGPU_IB_POOL_DIRECT, &ib);
-	if (r)
-		goto err1;
-
 	ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1);
-	ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START));
+	ib.ptr[1] = mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START;
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
 
 	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
 	if (r)
-		goto err2;
+		goto error;
 
 	r = dma_fence_wait_timeout(f, false, timeout);
 	if (r == 0) {
 		r = -ETIMEDOUT;
-		goto err2;
+		goto error;
 	} else if (r < 0) {
-		goto err2;
+		goto error;
 	}
-	tmp = RREG32(scratch);
+	tmp = RREG32(mmSCRATCH_REG0);
 	if (tmp == 0xDEADBEEF)
 		r = 0;
 	else
 		r = -EINVAL;
 
-err2:
+error:
 	amdgpu_ib_free(adev, &ib, NULL);
 	dma_fence_put(f);
-err1:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 			      &adev->gfx.priv_inst_irq);
 	if (r)
 		return r;
-
-	gfx_v7_0_scratch_init(adev);
 
 	r = gfx_v7_0_init_microcode(adev);
 	if (r) {

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

···
 	}
 }
 
-static void gfx_v8_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = mmSCRATCH_REG0;
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	WREG32(mmSCRATCH_REG0, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
 
-	WREG32(scratch, 0xCAFEDEAD);
-	r = amdgpu_ring_alloc(ring, 3);
-	if (r)
-		goto error_free_scratch;
-
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(ring, mmSCRATCH_REG0 - PACKET3_SET_UCONFIG_REG_START);
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
+		tmp = RREG32(mmSCRATCH_REG0);
 		if (tmp == 0xDEADBEEF)
 			break;
 		udelay(1);
···
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
 
-error_free_scratch:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 	INIT_WORK(&adev->gfx.sq_work.work, gfx_v8_0_sq_irq_work_func);
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
-
-	gfx_v8_0_scratch_init(adev);
 
 	r = gfx_v8_0_init_microcode(adev);
 	if (r) {

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

···
 						(const u32)ARRAY_SIZE(golden_settings_gc_9_x_common));
 }
 
-static void gfx_v9_0_scratch_init(struct amdgpu_device *adev)
-{
-	adev->gfx.scratch.num_reg = 8;
-	adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
-	adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1;
-}
-
 static void gfx_v9_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
 				       bool wc, uint32_t reg, uint32_t val)
 {
···
 static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t scratch;
 	uint32_t tmp = 0;
 	unsigned i;
 	int r;
 
-	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	WREG32_SOC15(GC, 0, mmSCRATCH_REG0, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
 
-	WREG32(scratch, 0xCAFEDEAD);
-	r = amdgpu_ring_alloc(ring, 3);
-	if (r)
-		goto error_free_scratch;
-
 	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
-	amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0) -
+			  PACKET3_SET_UCONFIG_REG_START);
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(scratch);
+		tmp = RREG32_SOC15(GC, 0, mmSCRATCH_REG0);
 		if (tmp == 0xDEADBEEF)
 			break;
 		udelay(1);
···
 
 	if (i >= adev->usec_timeout)
 		r = -ETIMEDOUT;
-
-error_free_scratch:
-	amdgpu_gfx_scratch_free(adev, scratch);
 	return r;
 }
 
···
 		return r;
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
-
-	gfx_v9_0_scratch_init(adev);
 
 	r = gfx_v9_0_init_microcode(adev);
 	if (r) {