[PATCH] drm/amdgpu: Add guest driver CUID support

[PATCH] drm/amdgpu: Add guest driver CUID support

[chong li]

1. Add guest driver CUID support
2. Sync the feature flag between guest driver and host driver.

Signed-off-by: chong li <chongli2@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h         |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 55 +++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c    |  8 ++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h    |  4 ++
 drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h | 12 ++++-
 5 files changed, 76 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 39894e38fee4..27522f7b12da 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1112,6 +1112,7 @@ struct amdgpu_device {
 	long				psp_timeout;
 
 	uint64_t			unique_id;
+	uint8_t				unitid;
 	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
 
 	/* enable runtime pm on the device */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 979032ecaf79..366e37db69d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -2114,6 +2114,45 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
 DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
 			amdgpu_debugfs_sclk_set, "%llu\n");
 
+
+static ssize_t amdgpu_debugfs_unique_id_read(struct file *f, char __user *buf,
+					 size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	char tmp[34];
+	int len;
+
+	len = scnprintf(tmp, sizeof(tmp), "0x%016llx\n",
+			adev->unique_id);
+
+	return simple_read_from_buffer(buf, size, pos, tmp, len);
+}
+
+static ssize_t amdgpu_debugfs_unitid_read(struct file *f, char __user *buf,
+					 size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	char tmp[34];
+	int len;
+
+	len = scnprintf(tmp, sizeof(tmp), "0x%02x\n",
+			adev->unitid);
+
+	return simple_read_from_buffer(buf, size, pos, tmp, len);
+}
+
+static const struct file_operations amdgpu_debugfs_unique_id_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_unique_id_read,
+	.llseek = default_llseek,
+};
+
+static const struct file_operations amdgpu_debugfs_unitid_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_unitid_read,
+	.llseek = default_llseek,
+};
+
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
 	struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
@@ -2126,6 +2165,22 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 	debugfs_create_x32("amdgpu_smu_debug", 0600, root,
 			   &adev->pm.smu_debug_mask);
 
+	ent = debugfs_create_file("unique_id", 0444, root, adev,
+			    &amdgpu_debugfs_unique_id_fops);
+	if (IS_ERR(ent)) {
+		drm_err(adev_to_drm(adev),
+			"unable to create unique_id debugfs file\n");
+		return PTR_ERR(ent);
+	}
+
+	ent = debugfs_create_file("unitid", 0444, root, adev,
+			    &amdgpu_debugfs_unitid_fops);
+	if (IS_ERR(ent)) {
+		drm_err(adev_to_drm(adev),
+			"unable to create unitid debugfs file\n");
+		return PTR_ERR(ent);
+	}
+
 	ent = debugfs_create_file("amdgpu_preempt_ib", 0600, root, adev,
 				  &fops_ib_preempt);
 	if (IS_ERR(ent)) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 6974b1c5b56c..45e89e104b5a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -534,8 +534,12 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 		if ((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
 			adev->virt.is_mm_bw_enabled = true;
 
-		adev->unique_id =
-			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+		adev->unique_id = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+
+		adev->unitid = 0;
+		if (amdgpu_sriov_is_unitid_support(adev))
+			adev->unitid = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->unitid;
+
 		adev->virt.ras_en_caps.all = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_en_caps.all;
 		adev->virt.ras_telemetry_en_caps.all =
 			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_telemetry_en_caps.all;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
index 9da0c6e9b869..5f889382fe4a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
@@ -163,6 +163,8 @@ enum AMDGIM_FEATURE_FLAG {
 	AMDGIM_FEATURE_RAS_CPER = (1 << 11),
 	AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK = (1 << 12),
 	AMDGIM_FEATURE_XGMI_CONNECTED_TO_CPU = (1 << 13),
+	AMDGIM_FEATURE_PTL_SUPPORT = (1 << 14),
+	AMDGIM_FEATURE_UNITID = (1 << 15),
 };
 
 enum AMDGIM_REG_ACCESS_FLAG {
@@ -441,6 +443,8 @@ static inline bool is_virtual_machine(void)
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_VCN_RB_DECOUPLE)
 #define amdgpu_sriov_is_mes_info_enable(adev) \
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_MES_INFO_ENABLE)
+#define amdgpu_sriov_is_unitid_support(adev) \
+	((adev)->virt.gim_feature & AMDGIM_FEATURE_UNITID)
 
 #define amdgpu_virt_xgmi_migrate_enabled(adev) \
 	((adev)->virt.is_xgmi_node_migrate_enabled && (adev)->gmc.xgmi.node_segment_size != 0)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
index 847cfd1fd004..7d79fb785e12 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
@@ -162,7 +162,9 @@ union amd_sriov_msg_feature_flags {
 		uint32_t ras_cper		: 1;
 		uint32_t xgmi_ta_ext_peer_link	: 1;
 		uint32_t xgmi_connected_to_cpu  : 1;
-		uint32_t reserved		: 18;
+		uint32_t ptl_support		: 1;
+		uint32_t unitid_support		: 1;
+		uint32_t reserved		: 16;
 	} flags;
 	uint32_t all;
 };
@@ -256,7 +258,7 @@ struct amd_sriov_msg_pf2vf_info_header {
 	uint32_t reserved[2];
 };
 
-#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (55)
+#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (59)
 struct amd_sriov_msg_pf2vf_info {
 	/* header contains size and version */
 	struct amd_sriov_msg_pf2vf_info_header header;
@@ -314,6 +316,12 @@ struct amd_sriov_msg_pf2vf_info {
 	uint32_t more_bp;	//Reserved for future use.
 	union amd_sriov_ras_caps ras_en_caps;
 	union amd_sriov_ras_caps ras_telemetry_en_caps;
+	uint8_t unitid;
+	uint8_t padding[3];  //use the 3 bytes to align
+	/* PTL status response for guest */
+	uint32_t ptl_enabled;        // PTL enable status: 0=disabled, 1=enabled
+	uint32_t ptl_pref_format1;   // Current preferred format 1
+	uint32_t ptl_pref_format2;   // Current preferred format 2
 
 	/* reserved */
 	uint32_t reserved[256 - AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE];
-- 
2.48.1

[PATCH] drm/amdgpu: Add guest driver CUID support

[chong li]

1. Add guest driver CUID support
2. Do not expose vf index(variable "fcn_idx") to customers,
   replace the fcn_idx with pad.
   Only expose the unitid to customers.

Signed-off-by: chong li <chongli2@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h         |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 55 +++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c    |  8 ++-
 drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h |  2 +-
 4 files changed, 63 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 80b18bbd7f3a..98549a148695 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1074,6 +1074,7 @@ struct amdgpu_device {
 	long				psp_timeout;
 
 	uint64_t			unique_id;
+	uint8_t				unitid;
 	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
 
 	/* enable runtime pm on the device */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index 979032ecaf79..366e37db69d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -2114,6 +2114,45 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
 DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
 			amdgpu_debugfs_sclk_set, "%llu\n");
 
+
+static ssize_t amdgpu_debugfs_unique_id_read(struct file *f, char __user *buf,
+					 size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	char tmp[34];
+	int len;
+
+	len = scnprintf(tmp, sizeof(tmp), "0x%016llx\n",
+			adev->unique_id);
+
+	return simple_read_from_buffer(buf, size, pos, tmp, len);
+}
+
+static ssize_t amdgpu_debugfs_unitid_read(struct file *f, char __user *buf,
+					 size_t size, loff_t *pos)
+{
+	struct amdgpu_device *adev = file_inode(f)->i_private;
+	char tmp[34];
+	int len;
+
+	len = scnprintf(tmp, sizeof(tmp), "0x%02x\n",
+			adev->unitid);
+
+	return simple_read_from_buffer(buf, size, pos, tmp, len);
+}
+
+static const struct file_operations amdgpu_debugfs_unique_id_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_unique_id_read,
+	.llseek = default_llseek,
+};
+
+static const struct file_operations amdgpu_debugfs_unitid_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_unitid_read,
+	.llseek = default_llseek,
+};
+
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
 	struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
@@ -2126,6 +2165,22 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 	debugfs_create_x32("amdgpu_smu_debug", 0600, root,
 			   &adev->pm.smu_debug_mask);
 
+	ent = debugfs_create_file("unique_id", 0444, root, adev,
+			    &amdgpu_debugfs_unique_id_fops);
+	if (IS_ERR(ent)) {
+		drm_err(adev_to_drm(adev),
+			"unable to create unique_id debugfs file\n");
+		return PTR_ERR(ent);
+	}
+
+	ent = debugfs_create_file("unitid", 0444, root, adev,
+			    &amdgpu_debugfs_unitid_fops);
+	if (IS_ERR(ent)) {
+		drm_err(adev_to_drm(adev),
+			"unable to create unitid debugfs file\n");
+		return PTR_ERR(ent);
+	}
+
 	ent = debugfs_create_file("amdgpu_preempt_ib", 0600, root, adev,
 				  &fops_ib_preempt);
 	if (IS_ERR(ent)) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 6974b1c5b56c..45e89e104b5a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -534,8 +534,12 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 		if ((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
 			adev->virt.is_mm_bw_enabled = true;
 
-		adev->unique_id =
-			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+		adev->unique_id = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+
+		adev->unitid = 0;
+		if (amdgpu_sriov_is_unitid_support(adev))
+			adev->unitid = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->unitid;
+
 		adev->virt.ras_en_caps.all = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_en_caps.all;
 		adev->virt.ras_telemetry_en_caps.all =
 			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_telemetry_en_caps.all;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
index 9dcf0b07d513..d80f01c0e754 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
@@ -295,7 +295,7 @@ struct amd_sriov_msg_pf2vf_info {
 	uint32_t vf2pf_update_interval_ms;
 	/* identification in ROCm SMI */
 	uint64_t uuid;
-	uint32_t fcn_idx;
+	uint32_t pad;
 	/* flags to indicate which register access method VF should use */
 	union amd_sriov_reg_access_flags reg_access_flags;
 	/* MM BW management */
-- 
2.48.1

Re: [PATCH] drm/amdgpu: Add guest driver CUID support

[Christian König]

On 5/8/26 09:53, chong li wrote:
> 1. Add guest driver CUID support
> 2. Do not expose vf index(variable "fcn_idx") to customers,
>    replace the fcn_idx with pad.
>    Only expose the unitid to customers.
> 
> Signed-off-by: chong li <chongli2@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu.h         |  1 +
>  drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 55 +++++++++++++++++++++
>  drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c    |  8 ++-
>  drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h |  2 +-
>  4 files changed, 63 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index 80b18bbd7f3a..98549a148695 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -1074,6 +1074,7 @@ struct amdgpu_device {
>  	long				psp_timeout;
>  
>  	uint64_t			unique_id;
> +	uint8_t				unitid;
>  	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
>  
>  	/* enable runtime pm on the device */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> index 979032ecaf79..366e37db69d9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
> @@ -2114,6 +2114,45 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
>  DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
>  			amdgpu_debugfs_sclk_set, "%llu\n");
>  

> +
> +static ssize_t amdgpu_debugfs_unique_id_read(struct file *f, char __user *buf,
> +					 size_t size, loff_t *pos)
> +{
> +	struct amdgpu_device *adev = file_inode(f)->i_private;
> +	char tmp[34];
> +	int len;
> +
> +	len = scnprintf(tmp, sizeof(tmp), "0x%016llx\n",
> +			adev->unique_id);
> +
> +	return simple_read_from_buffer(buf, size, pos, tmp, len);
> +}
> +
> +static ssize_t amdgpu_debugfs_unitid_read(struct file *f, char __user *buf,
> +					 size_t size, loff_t *pos)
> +{
> +	struct amdgpu_device *adev = file_inode(f)->i_private;
> +	char tmp[34];
> +	int len;
> +
> +	len = scnprintf(tmp, sizeof(tmp), "0x%02x\n",
> +			adev->unitid);
> +
> +	return simple_read_from_buffer(buf, size, pos, tmp, len);
> +}
> +
> +static const struct file_operations amdgpu_debugfs_unique_id_fops = {
> +	.owner = THIS_MODULE,
> +	.read = amdgpu_debugfs_unique_id_read,
> +	.llseek = default_llseek,
> +};
> +
> +static const struct file_operations amdgpu_debugfs_unitid_fops = {
> +	.owner = THIS_MODULE,
> +	.read = amdgpu_debugfs_unitid_read,
> +	.llseek = default_llseek,
> +};
> +

As far as I can see that is overkill, please use debugfs_create_x64() and debugfs_create_x8() instead.


>  int amdgpu_debugfs_init(struct amdgpu_device *adev)
>  {
>  	struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
> @@ -2126,6 +2165,22 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
>  	debugfs_create_x32("amdgpu_smu_debug", 0600, root,
>  			   &adev->pm.smu_debug_mask);
>  
> +	ent = debugfs_create_file("unique_id", 0444, root, adev,
> +			    &amdgpu_debugfs_unique_id_fops);

> +	if (IS_ERR(ent)) {
> +		drm_err(adev_to_drm(adev),
> +			"unable to create unique_id debugfs file\n");
> +		return PTR_ERR(ent);
> +	}

Debugfs should never have error handling. Please drop that.

> +
> +	ent = debugfs_create_file("unitid", 0444, root, adev,
> +			    &amdgpu_debugfs_unitid_fops);
> +	if (IS_ERR(ent)) {
> +		drm_err(adev_to_drm(adev),
> +			"unable to create unitid debugfs file\n");
> +		return PTR_ERR(ent);
> +	}
> +
>  	ent = debugfs_create_file("amdgpu_preempt_ib", 0600, root, adev,
>  				  &fops_ib_preempt);
>  	if (IS_ERR(ent)) {
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> index 6974b1c5b56c..45e89e104b5a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
> @@ -534,8 +534,12 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
>  		if ((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
>  			adev->virt.is_mm_bw_enabled = true;
>  
> -		adev->unique_id =
> -			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
> +		adev->unique_id = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;

Please make sure that this passes checkpatch.pl. Of hand it looks like the line is to long.

Regards,
Christian.

> +
> +		adev->unitid = 0;
> +		if (amdgpu_sriov_is_unitid_support(adev))
> +			adev->unitid = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->unitid;
> +
>  		adev->virt.ras_en_caps.all = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_en_caps.all;
>  		adev->virt.ras_telemetry_en_caps.all =
>  			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_telemetry_en_caps.all;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
> index 9dcf0b07d513..d80f01c0e754 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
> @@ -295,7 +295,7 @@ struct amd_sriov_msg_pf2vf_info {
>  	uint32_t vf2pf_update_interval_ms;
>  	/* identification in ROCm SMI */
>  	uint64_t uuid;
> -	uint32_t fcn_idx;
> +	uint32_t pad;
>  	/* flags to indicate which register access method VF should use */
>  	union amd_sriov_reg_access_flags reg_access_flags;
>  	/* MM BW management */