Merge tag 'amd-drm-fixes-6.17-2025-09-10' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.17-2025-09-10:

amdgpu:
- PSP 11.x fix
- DPCD quirk handing fix
- DCN 3.5 PG fix
- Audio suspend fix
- OEM i2c clean up fix
- Module unload memory leak fix
- DC delay fix
- ISP firmware fix
- VCN fixes

amdkfd:
- P2P topology fix
- APU mem limit calculation fix

Signed-off-by: Dave Airlie <[email protected]>

From: Alex Deucher <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Author: airlied

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

@@ -213,19 +213,35 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
 	if (kfd_mem_limit.system_mem_used + system_mem_needed >
-	    kfd_mem_limit.max_system_mem_limit)
+	    kfd_mem_limit.max_system_mem_limit) {
 		pr_debug("Set no_system_mem_limit=1 if using shared memory\n");
+		if (!no_system_mem_limit) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
 
-	if ((kfd_mem_limit.system_mem_used + system_mem_needed >
-	     kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) ||
-	    (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
-	     kfd_mem_limit.max_ttm_mem_limit) ||
-	    (adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] + vram_needed >
-	     vram_size - reserved_for_pt - reserved_for_ras - atomic64_read(&adev->vram_pin_size))) {
+	if (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
+		kfd_mem_limit.max_ttm_mem_limit) {
 		ret = -ENOMEM;
 		goto release;
 	}
 
+	/*if is_app_apu is false and apu_prefer_gtt is true, it is an APU with
+	 * carve out < gtt. In that case, VRAM allocation will go to gtt domain, skip
+	 * VRAM check since ttm_mem_limit check already cover this allocation
+	 */
+
+	if (adev && xcp_id >= 0 && (!adev->apu_prefer_gtt || adev->gmc.is_app_apu)) {
+		uint64_t vram_available =
+			vram_size - reserved_for_pt - reserved_for_ras -
+			atomic64_read(&adev->vram_pin_size);
+		if (adev->kfd.vram_used[xcp_id] + vram_needed > vram_available) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
+
 	/* Update memory accounting by decreasing available system
 	 * memory, TTM memory and GPU memory as computed above
 	 */
@@ -1626,11 +1642,15 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
 	uint64_t vram_available, system_mem_available, ttm_mem_available;
 
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
-	vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
-		- adev->kfd.vram_used_aligned[xcp_id]
-		- atomic64_read(&adev->vram_pin_size)
-		- reserved_for_pt
-		- reserved_for_ras;
+	if (adev->apu_prefer_gtt && !adev->gmc.is_app_apu)
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id];
+	else
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id]
+			- atomic64_read(&adev->vram_pin_size)
+			- reserved_for_pt
+			- reserved_for_ras;
 
 	if (adev->apu_prefer_gtt) {
 		system_mem_available = no_system_mem_limit ?

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

@@ -421,8 +421,6 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring)
 	dma_fence_put(ring->vmid_wait);
 	ring->vmid_wait = NULL;
 	ring->me = 0;
-
-	ring->adev->rings[ring->idx] = NULL;
 }
 
 /**

drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c

@@ -29,6 +29,8 @@
 #include "amdgpu.h"
 #include "isp_v4_1_1.h"
 
+MODULE_FIRMWARE("amdgpu/isp_4_1_1.bin");
+
 #define ISP_PERFORMANCE_STATE_LOW 0
 #define ISP_PERFORMANCE_STATE_HIGH 1
 

drivers/gpu/drm/amd/amdgpu/psp_v11_0.c

@@ -149,12 +149,12 @@ static int psp_v11_0_wait_for_bootloader(struct psp_context *psp)
 	int ret;
 	int retry_loop;
 
-	for (retry_loop = 0; retry_loop < 10; retry_loop++) {
+	for (retry_loop = 0; retry_loop < 20; retry_loop++) {
 		/* Wait for bootloader to signify that is
 		    ready having bit 31 of C2PMSG_35 set to 1 */
 		ret = psp_wait_for(
 			psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35),
-			0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE);
+			0x80000000, 0x8000FFFF, PSP_WAITREG_NOVERBOSE);
 
 		if (ret == 0)
 			return 0;
@@ -397,18 +397,6 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp)
 
 	msleep(500);
 
-	offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33);
-
-	ret = psp_wait_for(psp, offset, MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK,
-			   0);
-
-	if (ret) {
-		DRM_INFO("psp mode 1 reset failed!\n");
-		return -EINVAL;
-	}
-
-	DRM_INFO("psp mode1 reset succeed \n");
-
 	return 0;
 }
 
@@ -665,7 +653,8 @@ static const struct psp_funcs psp_v11_0_funcs = {
 	.ring_get_wptr = psp_v11_0_ring_get_wptr,
 	.ring_set_wptr = psp_v11_0_ring_set_wptr,
 	.load_usbc_pd_fw = psp_v11_0_load_usbc_pd_fw,
-	.read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw
+	.read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw,
+	.wait_for_bootloader = psp_v11_0_wait_for_bootloader
 };
 
 void psp_v11_0_set_psp_funcs(struct psp_context *psp)

drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c

@@ -1888,15 +1888,19 @@ static int vcn_v3_0_limit_sched(struct amdgpu_cs_parser *p,
 				struct amdgpu_job *job)
 {
 	struct drm_gpu_scheduler **scheds;
-
-	/* The create msg must be in the first IB submitted */
-	if (atomic_read(&job->base.entity->fence_seq))
-		return -EINVAL;
+	struct dma_fence *fence;
 
 	/* if VCN0 is harvested, we can't support AV1 */
 	if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0)
 		return -EINVAL;
 
+	/* wait for all jobs to finish before switching to instance 0 */
+	fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull);
+	if (fence) {
+		dma_fence_wait(fence, false);
+		dma_fence_put(fence);
+	}
+
 	scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_DEC]
 		[AMDGPU_RING_PRIO_DEFAULT].sched;
 	drm_sched_entity_modify_sched(job->base.entity, scheds, 1);

drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c

@@ -1808,15 +1808,19 @@ static int vcn_v4_0_limit_sched(struct amdgpu_cs_parser *p,
 				struct amdgpu_job *job)
 {
 	struct drm_gpu_scheduler **scheds;
-
-	/* The create msg must be in the first IB submitted */
-	if (atomic_read(&job->base.entity->fence_seq))
-		return -EINVAL;
+	struct dma_fence *fence;
 
 	/* if VCN0 is harvested, we can't support AV1 */
 	if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0)
 		return -EINVAL;
 
+	/* wait for all jobs to finish before switching to instance 0 */
+	fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull);
+	if (fence) {
+		dma_fence_wait(fence, false);
+		dma_fence_put(fence);
+	}
+
 	scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_ENC]
 		[AMDGPU_RING_PRIO_0].sched;
 	drm_sched_entity_modify_sched(job->base.entity, scheds, 1);
@@ -1907,22 +1911,16 @@ static int vcn_v4_0_dec_msg(struct amdgpu_cs_parser *p, struct amdgpu_job *job,
 
 #define RADEON_VCN_ENGINE_TYPE_ENCODE			(0x00000002)
 #define RADEON_VCN_ENGINE_TYPE_DECODE			(0x00000003)
-
 #define RADEON_VCN_ENGINE_INFO				(0x30000001)
-#define RADEON_VCN_ENGINE_INFO_MAX_OFFSET		16
-
 #define RENCODE_ENCODE_STANDARD_AV1			2
 #define RENCODE_IB_PARAM_SESSION_INIT			0x00000003
-#define RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET	64
 
-/* return the offset in ib if id is found, -1 otherwise
- * to speed up the searching we only search upto max_offset
- */
-static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int max_offset)
+/* return the offset in ib if id is found, -1 otherwise */
+static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int start)
 {
 	int i;
 
-	for (i = 0; i < ib->length_dw && i < max_offset && ib->ptr[i] >= 8; i += ib->ptr[i]/4) {
+	for (i = start; i < ib->length_dw && ib->ptr[i] >= 8; i += ib->ptr[i] / 4) {
 		if (ib->ptr[i + 1] == id)
 			return i;
 	}
@@ -1937,33 +1935,29 @@ static int vcn_v4_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
 	struct amdgpu_vcn_decode_buffer *decode_buffer;
 	uint64_t addr;
 	uint32_t val;
-	int idx;
+	int idx = 0, sidx;
 
 	/* The first instance can decode anything */
 	if (!ring->me)
 		return 0;
 
-	/* RADEON_VCN_ENGINE_INFO is at the top of ib block */
-	idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO,
-			RADEON_VCN_ENGINE_INFO_MAX_OFFSET);
-	if (idx < 0) /* engine info is missing */
-		return 0;
-
-	val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */
-	if (val == RADEON_VCN_ENGINE_TYPE_DECODE) {
-		decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6];
-
-		if (!(decode_buffer->valid_buf_flag  & 0x1))
-			return 0;
-
-		addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 |
-			decode_buffer->msg_buffer_address_lo;
-		return vcn_v4_0_dec_msg(p, job, addr);
-	} else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) {
-		idx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT,
-			RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET);
-		if (idx >= 0 && ib->ptr[idx + 2] == RENCODE_ENCODE_STANDARD_AV1)
-			return vcn_v4_0_limit_sched(p, job);
+	while ((idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO, idx)) >= 0) {
+		val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */
+		if (val == RADEON_VCN_ENGINE_TYPE_DECODE) {
+			decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6];
+
+			if (!(decode_buffer->valid_buf_flag & 0x1))
+				return 0;
+
+			addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 |
+				decode_buffer->msg_buffer_address_lo;
+			return vcn_v4_0_dec_msg(p, job, addr);
+		} else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) {
+			sidx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT, idx);
+			if (sidx >= 0 && ib->ptr[sidx + 2] == RENCODE_ENCODE_STANDARD_AV1)
+				return vcn_v4_0_limit_sched(p, job);
+		}
+		idx += ib->ptr[idx] / 4;
 	}
 	return 0;
 }

drivers/gpu/drm/amd/amdkfd/kfd_topology.c

@@ -1587,7 +1587,8 @@ static int kfd_dev_create_p2p_links(void)
 			break;
 		if (!dev->gpu || !dev->gpu->adev ||
 		    (dev->gpu->kfd->hive_id &&
-		     dev->gpu->kfd->hive_id == new_dev->gpu->kfd->hive_id))
+		     dev->gpu->kfd->hive_id == new_dev->gpu->kfd->hive_id &&
+		     amdgpu_xgmi_get_is_sharing_enabled(dev->gpu->adev, new_dev->gpu->adev)))
 			goto next;
 
 		/* check if node(s) is/are peer accessible in one direction or bi-direction */

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

@@ -2913,6 +2913,17 @@ static int dm_oem_i2c_hw_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+static void dm_oem_i2c_hw_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_display_manager *dm = &adev->dm;
+
+	if (dm->oem_i2c) {
+		i2c_del_adapter(&dm->oem_i2c->base);
+		kfree(dm->oem_i2c);
+		dm->oem_i2c = NULL;
+	}
+}
+
 /**
  * dm_hw_init() - Initialize DC device
  * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
@@ -2963,7 +2974,7 @@ static int dm_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->dm.oem_i2c);
+	dm_oem_i2c_hw_fini(adev);
 
 	amdgpu_dm_hpd_fini(adev);
 
@@ -3127,25 +3138,6 @@ static void dm_destroy_cached_state(struct amdgpu_device *adev)
 	dm->cached_state = NULL;
 }
 
-static void dm_complete(struct amdgpu_ip_block *ip_block)
-{
-	struct amdgpu_device *adev = ip_block->adev;
-
-	dm_destroy_cached_state(adev);
-}
-
-static int dm_prepare_suspend(struct amdgpu_ip_block *ip_block)
-{
-	struct amdgpu_device *adev = ip_block->adev;
-
-	if (amdgpu_in_reset(adev))
-		return 0;
-
-	WARN_ON(adev->dm.cached_state);
-
-	return dm_cache_state(adev);
-}
-
 static int dm_suspend(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_device *adev = ip_block->adev;
@@ -3571,10 +3563,8 @@ static const struct amd_ip_funcs amdgpu_dm_funcs = {
 	.early_fini = amdgpu_dm_early_fini,
 	.hw_init = dm_hw_init,
 	.hw_fini = dm_hw_fini,
-	.prepare_suspend = dm_prepare_suspend,
 	.suspend = dm_suspend,
 	.resume = dm_resume,
-	.complete = dm_complete,
 	.is_idle = dm_is_idle,
 	.wait_for_idle = dm_wait_for_idle,
 	.check_soft_reset = dm_check_soft_reset,

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c

@@ -809,6 +809,7 @@ void amdgpu_dm_initialize_dp_connector(struct amdgpu_display_manager *dm,
 	drm_dp_aux_init(&aconnector->dm_dp_aux.aux);
 	drm_dp_cec_register_connector(&aconnector->dm_dp_aux.aux,
 				      &aconnector->base);
+	drm_dp_dpcd_set_probe(&aconnector->dm_dp_aux.aux, false);
 
 	if (aconnector->base.connector_type == DRM_MODE_CONNECTOR_eDP)
 		return;

drivers/gpu/drm/amd/display/dc/dc.h

@@ -1145,6 +1145,7 @@ struct dc_debug_options {
 	bool enable_hblank_borrow;
 	bool force_subvp_df_throttle;
 	uint32_t acpi_transition_bitmasks[MAX_PIPES];
+	bool enable_pg_cntl_debug_logs;
 };