[1/3] drm/amd/amdgpu: Add debugfs support for reading GPRs

Submitted by Tom St Denis on Dec. 5, 2016, 6:36 p.m.

Details

Message ID 20161205183652.29325-2-tom.stdenis@amd.com
State New
Headers show
Series "Series without cover letter" ( rev: 1 ) in AMD X.Org drivers

Not browsing as part of any series.

Commit Message

Tom St Denis Dec. 5, 2016, 6:36 p.m.
Implemented for SGPRs for GFX v8 initially.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  2 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 72 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c      | 25 +++++++++++
 3 files changed, 99 insertions(+)

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 4f64bb16a8d3..2834451eef8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -844,6 +844,8 @@  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 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 25ad736b5ca5..7f9cb1c85d75 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3055,6 +3055,71 @@  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);
+	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);
+
+	/* sanity check offset/size */
+	if (offset + size > (adev->gfx.config.max_gprs << 2))
+		return -EINVAL;
+
+	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>>2, size>>2, data);
+	} else {
+		if (adev->gfx.funcs->read_wave_sgprs)
+			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset>>2, 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 >> 2];
+		r = put_user(value, (uint32_t *)buf);
+		if (r) {
+			result = r;
+			goto err;
+		}
+
+		result += 4;
+		buf += 4;
+		offset += 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,
@@ -3097,6 +3162,11 @@  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,
@@ -3106,6 +3176,7 @@  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[] = {
@@ -3116,6 +3187,7 @@  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)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 9b8d3fe67adb..180309f5784c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/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)

Comments

Hi Tom,

On 12/06/2016 05:36 AM, Tom St Denis wrote:
> Implemented for SGPRs for GFX v8 initially.
> 
> Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu.h        |  2 +
>  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 72 ++++++++++++++++++++++++++++++
>  drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c      | 25 +++++++++++
>  3 files changed, 99 insertions(+)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index 4f64bb16a8d3..2834451eef8a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -844,6 +844,8 @@ 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 {
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index 25ad736b5ca5..7f9cb1c85d75 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -3055,6 +3055,71 @@ 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;

Could f potentially be null here?

> +	int r;
> +	ssize_t result=0;

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);
> +	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);
> +
> +	/* sanity check offset/size */
> +	if (offset + size > (adev->gfx.config.max_gprs << 2))
> +		return -EINVAL;
> +
> +	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>>2, size>>2, data);
> +	} else {
> +		if (adev->gfx.funcs->read_wave_sgprs)
> +			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset>>2, 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 >> 2];
> +		r = put_user(value, (uint32_t *)buf);
> +		if (r) {
> +			result = r;
> +			goto err;
> +		}
> +
> +		result += 4;
> +		buf += 4;
> +		offset += 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,
> @@ -3097,6 +3162,11 @@ 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,
> @@ -3106,6 +3176,7 @@ 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[] = {
> @@ -3116,6 +3187,7 @@ 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)
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> index 9b8d3fe67adb..180309f5784c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
> +++ b/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,

Just fyi, these & are actually redundant when it comes to func ptr's.

>  };
>  
>  static int gfx_v8_0_early_init(void *handle)
> 

Kind Regards,
Edward.
Hi Edward,


I don't think 'f' could be NULL there nor could the private data be NULL because it's initialized to point to the amdgpu_device when the entry is added to the vfs.


I can sort out the whitespace though.


Tom