Skip to contentSkip to conversation

drm/amdgpu: Add user queue instance count in HW IP info

PATCH Superseded Jesse Zhang 2 participants 2 messages Jun 25, 2025

Subsystems (3)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: Add user queue instance count in HW IP info

This change exposes the number of available user queue instances for each hardware IP type (GFX, COMPUTE, SDMA) through the drm_amdgpu_info_hw_ip interface.

Key changes:
1. Added userq_num_instance field to drm_amdgpu_info_hw_ip structure
2. Implemented counting of available HQD slots using:
- mes.gfx_hqd_mask for GFX queues
- mes.compute_hqd_mask for COMPUTE queues
- mes.sdma_hqd_mask for SDMA queues
3. Only counts available instances when user queues are enabled (!disable_uq)

v2: using the adev->mes.gfx_hqd_mask[]/compute_hqd_mask[]/sdma_hqd_mask[] masks
to determine the number of queue slots available for each engine type (Alex)

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 17 +++++++++++++++++
 include/uapi/drm/amdgpu_drm.h           |  2 ++
 2 files changed, 19 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +17 -0
index 195ed81d39ff..d682a92f086b 100644
@@ -399,6 +399,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	uint32_t ib_size_alignment = 0;
 	enum amd_ip_block_type type;
 	unsigned int num_rings = 0;
+	uint32_t num_instance = 0;
 	unsigned int i, j;
 
 	if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
@@ -411,6 +412,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->gfx.gfx_ring[i].sched.ready &&
 			    !adev->gfx.gfx_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq)
+			for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
+				num_instance += hweight32(adev->mes.gfx_hqd_mask[i]);
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
@@ -420,6 +426,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->gfx.compute_ring[i].sched.ready &&
 			    !adev->gfx.compute_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->sdma.disable_uq)
+			for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++)
+				num_instance += hweight32(adev->mes.compute_hqd_mask[i]);
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
@@ -429,6 +440,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->sdma.instance[i].ring.sched.ready &&
 			    !adev->sdma.instance[i].ring.no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq)
+			for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++)
+				num_instance += hweight32(adev->mes.sdma_hqd_mask[i]);
+
 		ib_start_alignment = 256;
 		ib_size_alignment = 4;
 		break;
@@ -570,6 +586,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	}
 	result->capabilities_flags = 0;
 	result->available_rings = (1 << num_rings) - 1;
+	result->userq_num_instance = num_instance;
 	result->ib_start_alignment = ib_start_alignment;
 	result->ib_size_alignment = ib_size_alignment;
 	return 0;
include/uapi/drm/amdgpu_drm.h +2 -0
index 45c4fa13499c..8082270a7ceb 100644
@@ -1493,6 +1493,8 @@ struct drm_amdgpu_info_hw_ip {
 	__u32  available_rings;
 	/** version info: bits 23:16 major, 15:8 minor, 7:0 revision */
 	__u32  ip_discovery_version;
+	/* Userq available instance */
+	__u32  userq_num_instance;
 };
 
 /* GFX metadata BO sizes and alignment info (in bytes) */
Signature · 1 line
2.34.1
#1 lore raw
Quoted 85 lines · Jesse Zhang

This change exposes the number of available user queue instances for each hardware IP type (GFX, COMPUTE, SDMA) through the drm_amdgpu_info_hw_ip interface.

Key changes:
1. Added userq_num_instance field to drm_amdgpu_info_hw_ip structure
2. Implemented counting of available HQD slots using:
- mes.gfx_hqd_mask for GFX queues
- mes.compute_hqd_mask for COMPUTE queues
- mes.sdma_hqd_mask for SDMA queues
3. Only counts available instances when user queues are enabled (!disable_uq)

v2: using the adev->mes.gfx_hqd_mask[]/compute_hqd_mask[]/sdma_hqd_mask[] masks
to determine the number of queue slots available for each engine type (Alex)

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 17 +++++++++++++++++
 include/uapi/drm/amdgpu_drm.h           |  2 ++
 2 files changed, 19 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +17 -0
index 195ed81d39ff..d682a92f086b 100644
@@ -399,6 +399,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
        uint32_t ib_size_alignment = 0;
        enum amd_ip_block_type type;
        unsigned int num_rings = 0;
+       uint32_t num_instance = 0;
        unsigned int i, j;
 
        if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
@@ -411,6 +412,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
                        if (adev->gfx.gfx_ring[i].sched.ready &&
                            !adev->gfx.gfx_ring[i].no_user_submission)
                                ++num_rings;
+
+               if (!adev->gfx.disable_uq)
+                       for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
+                               num_instance += hweight32(adev->mes.gfx_hqd_mask[i]);
+
                ib_start_alignment = 32;
                ib_size_alignment = 32;
                break;
@@ -420,6 +426,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
                        if (adev->gfx.compute_ring[i].sched.ready &&
                            !adev->gfx.compute_ring[i].no_user_submission)
                                ++num_rings;
+
+               if (!adev->sdma.disable_uq)
+                       for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++)
+                               num_instance += hweight32(adev->mes.compute_hqd_mask[i]);
+
                ib_start_alignment = 32;
                ib_size_alignment = 32;
                break;
@@ -429,6 +440,11 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
                        if (adev->sdma.instance[i].ring.sched.ready &&
                            !adev->sdma.instance[i].ring.no_user_submission)
                                ++num_rings;
+
+               if (!adev->gfx.disable_uq)
+                       for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++)
+                               num_instance += hweight32(adev->mes.sdma_hqd_mask[i]);
+
                ib_start_alignment = 256;
                ib_size_alignment = 4;
                break;
@@ -570,6 +586,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
        }
        result->capabilities_flags = 0;
        result->available_rings = (1 << num_rings) - 1;
+       result->userq_num_instance = num_instance;
        result->ib_start_alignment = ib_start_alignment;
        result->ib_size_alignment = ib_size_alignment;
        return 0;
include/uapi/drm/amdgpu_drm.h +0 -0
index 45c4fa13499c..8082270a7ceb 100644

include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_info_hw_ip { · line 1497

Quoted 5 lines
include/uapi/drm/amdgpu_drm.h +2 -0
@@ -1493,6 +1493,8 @@ struct drm_amdgpu_info_hw_ip {
        __u32  available_rings;
        /** version info: bits 23:16 major, 15:8 minor, 7:0 revision */
        __u32  ip_discovery_version;
+       /* Userq available instance */
+       __u32  userq_num_instance;

I would call this something like userq_num_slots or userq_num_hqds. Other than that, it looks good to me.

Alex

Quoted 6 lines
include/uapi/drm/amdgpu_drm.h +0 -1
 };
 
 /* GFX metadata BO sizes and alignment info (in bytes) */
--

2.34.1