drm/amd/amdgpu: Add debugfs support for reading GPRs (v2)

Implemented for SGPRs for GFX v8 initially.

(v2) cleanup minor whitespace and remove sanity check and
     addressing is in dwords not bytes
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -842,6 +842,8 @@ struct amdgpu_gfx_funcs {
 	uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
 	void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance);
 	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields);
+	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t thread, uint32_t start, uint32_t size, uint32_t *dst);
+	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst);
 };
 
 struct amdgpu_gfx {

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

@@ -2985,6 +2985,66 @@ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf,
 	return result;
 }
 
+static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf,
+					size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = f->f_inode->i_private;
+	int r;
+	ssize_t result = 0;
+	uint32_t offset, se, sh, cu, wave, simd, thread, bank, *data;
+
+	if (size & 3 || *pos & 3)
+		return -EINVAL;
+
+	/* decode offset */
+	offset = (*pos & 0xFFF);       /* in dwords */
+	se = ((*pos >> 12) & 0xFF);
+	sh = ((*pos >> 20) & 0xFF);
+	cu = ((*pos >> 28) & 0xFF);
+	wave = ((*pos >> 36) & 0xFF);
+	simd = ((*pos >> 44) & 0xFF);
+	thread = ((*pos >> 52) & 0xFF);
+	bank = ((*pos >> 60) & 1);
+
+	data = kmalloc_array(1024, sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	/* switch to the specific se/sh/cu */
+	mutex_lock(&adev->grbm_idx_mutex);
+	amdgpu_gfx_select_se_sh(adev, se, sh, cu);
+
+	if (bank == 0) {
+		if (adev->gfx.funcs->read_wave_vgprs)
+			adev->gfx.funcs->read_wave_vgprs(adev, simd, wave, thread, offset, size>>2, data);
+	} else {
+		if (adev->gfx.funcs->read_wave_sgprs)
+			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset, size>>2, data);
+	}
+
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	while (size) {
+		uint32_t value;
+
+		value = data[offset++];
+		r = put_user(value, (uint32_t *)buf);
+		if (r) {
+			result = r;
+			goto err;
+		}
+
+		result += 4;
+		buf += 4;
+		size -= 4;
+	}
+
+err:
+	kfree(data);
+	return result;
+}
+
 static const struct file_operations amdgpu_debugfs_regs_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_debugfs_regs_read,
@@ -3027,6 +3087,11 @@ static const struct file_operations amdgpu_debugfs_wave_fops = {
 	.read = amdgpu_debugfs_wave_read,
 	.llseek = default_llseek
 };
+static const struct file_operations amdgpu_debugfs_gpr_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_gpr_read,
+	.llseek = default_llseek
+};
 
 static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_regs_fops,
@@ -3036,6 +3101,7 @@ static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_gca_config_fops,
 	&amdgpu_debugfs_sensors_fops,
 	&amdgpu_debugfs_wave_fops,
+	&amdgpu_debugfs_gpr_fops,
 };
 
 static const char *debugfs_regs_names[] = {
@@ -3046,6 +3112,7 @@ static const char *debugfs_regs_names[] = {
 	"amdgpu_gca_config",
 	"amdgpu_sensors",
 	"amdgpu_wave",
+	"amdgpu_gpr",
 };
 
 static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)

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

@@ -5184,6 +5184,21 @@ static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_
 	return RREG32(mmSQ_IND_DATA);
 }
 
+static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
+			   uint32_t wave, uint32_t thread,
+			   uint32_t regno, uint32_t num, uint32_t *out)
+{
+	WREG32(mmSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
+		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
+		(thread << SQ_IND_INDEX__THREAD_ID__SHIFT) |
+		(SQ_IND_INDEX__FORCE_READ_MASK) |
+		(SQ_IND_INDEX__AUTO_INCR_MASK));
+	while (num--)
+		*(out++) = RREG32(mmSQ_IND_DATA);
+}
+
 static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
@@ -5208,11 +5223,21 @@ static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_M0);
 }
 
+static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+				     uint32_t wave, uint32_t start,
+				     uint32_t size, uint32_t *dst)
+{
+	wave_read_regs(
+		adev, simd, wave, 0,
+		start + SQIND_WAVE_SGPRS_OFFSET, size, dst);
+}
+
 
 static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = {
 	.get_gpu_clock_counter = &gfx_v8_0_get_gpu_clock_counter,
 	.select_se_sh = &gfx_v8_0_select_se_sh,
 	.read_wave_data = &gfx_v8_0_read_wave_data,
+	.read_wave_sgprs = &gfx_v8_0_read_wave_sgprs,
 };
 
 static int gfx_v8_0_early_init(void *handle)