Merge tag 'amd-drm-fixes-6.10-2024-06-19' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.10-2024-06-19:

amdgpu:
- Fix display idle optimization race
- Fix GPUVM TLB flush locking scope
- IPS fix
- GFX 9.4.3 harvesting fix
- Runtime pm fix for shared buffers
- DCN 3.5.x fixes
- USB4 fix
- RISC-V clang fix
- Silence UBSAN warnings
- MES11 fix
- PSP 14.0.x fix

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240619223233.3116457-1-alexander.deucher@amd.com

Author: airlied

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

@@ -41,8 +41,6 @@
 #include <linux/dma-buf.h>
 #include <linux/dma-fence-array.h>
 #include <linux/pci-p2pdma.h>
-#include <linux/pm_runtime.h>
-#include "amdgpu_trace.h"
 
 /**
  * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation
@@ -58,42 +56,11 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
 	struct drm_gem_object *obj = dmabuf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	int r;
 
 	if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
 		attach->peer2peer = false;
 
-	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
-	trace_amdgpu_runpm_reference_dumps(1, __func__);
-	if (r < 0)
-		goto out;
-
 	return 0;
-
-out:
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
-	trace_amdgpu_runpm_reference_dumps(0, __func__);
-	return r;
-}
-
-/**
- * amdgpu_dma_buf_detach - &dma_buf_ops.detach implementation
- *
- * @dmabuf: DMA-buf where we remove the attachment from
- * @attach: the attachment to remove
- *
- * Called when an attachment is removed from the DMA-buf.
- */
-static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf,
-				  struct dma_buf_attachment *attach)
-{
-	struct drm_gem_object *obj = dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
-	trace_amdgpu_runpm_reference_dumps(0, __func__);
 }
 
 /**
@@ -267,7 +234,6 @@ static int amdgpu_dma_buf_begin_cpu_access(struct dma_buf *dma_buf,
 
 const struct dma_buf_ops amdgpu_dmabuf_ops = {
 	.attach = amdgpu_dma_buf_attach,
-	.detach = amdgpu_dma_buf_detach,
 	.pin = amdgpu_dma_buf_pin,
 	.unpin = amdgpu_dma_buf_unpin,
 	.map_dma_buf = amdgpu_dma_buf_map,

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

@@ -181,7 +181,6 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd
 	amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr,
 			       seq, flags | AMDGPU_FENCE_FLAG_INT);
 	pm_runtime_get_noresume(adev_to_drm(adev)->dev);
-	trace_amdgpu_runpm_reference_dumps(1, __func__);
 	ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask];
 	if (unlikely(rcu_dereference_protected(*ptr, 1))) {
 		struct dma_fence *old;
@@ -309,7 +308,6 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 		dma_fence_put(fence);
 		pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
-		trace_amdgpu_runpm_reference_dumps(0, __func__);
 	} while (last_seq != seq);
 
 	return true;

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

@@ -684,12 +684,17 @@ int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
 	struct amdgpu_ring *ring = &adev->gfx.kiq[inst].ring;
 	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
 	unsigned int ndw;
-	signed long r;
+	int r;
 	uint32_t seq;
 
-	if (!adev->gmc.flush_pasid_uses_kiq || !ring->sched.ready ||
-	    !down_read_trylock(&adev->reset_domain->sem)) {
+	/*
+	 * A GPU reset should flush all TLBs anyway, so no need to do
+	 * this while one is ongoing.
+	 */
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return 0;
 
+	if (!adev->gmc.flush_pasid_uses_kiq || !ring->sched.ready) {
 		if (adev->gmc.flush_tlb_needs_extra_type_2)
 			adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
 								 2, all_hub,
@@ -703,43 +708,40 @@ int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
 		adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
 							 flush_type, all_hub,
 							 inst);
-		return 0;
-	}
+		r = 0;
+	} else {
+		/* 2 dwords flush + 8 dwords fence */
+		ndw = kiq->pmf->invalidate_tlbs_size + 8;
 
-	/* 2 dwords flush + 8 dwords fence */
-	ndw = kiq->pmf->invalidate_tlbs_size + 8;
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			ndw += kiq->pmf->invalidate_tlbs_size;
 
-	if (adev->gmc.flush_tlb_needs_extra_type_2)
-		ndw += kiq->pmf->invalidate_tlbs_size;
+		if (adev->gmc.flush_tlb_needs_extra_type_0)
+			ndw += kiq->pmf->invalidate_tlbs_size;
 
-	if (adev->gmc.flush_tlb_needs_extra_type_0)
-		ndw += kiq->pmf->invalidate_tlbs_size;
+		spin_lock(&adev->gfx.kiq[inst].ring_lock);
+		amdgpu_ring_alloc(ring, ndw);
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 2, all_hub);
 
-	spin_lock(&adev->gfx.kiq[inst].ring_lock);
-	amdgpu_ring_alloc(ring, ndw);
-	if (adev->gmc.flush_tlb_needs_extra_type_2)
-		kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 2, all_hub);
+		if (flush_type == 2 && adev->gmc.flush_tlb_needs_extra_type_0)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 0, all_hub);
 
-	if (flush_type == 2 && adev->gmc.flush_tlb_needs_extra_type_0)
-		kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 0, all_hub);
+		kiq->pmf->kiq_invalidate_tlbs(ring, pasid, flush_type, all_hub);
+		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+		if (r) {
+			amdgpu_ring_undo(ring);
+			spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+			goto error_unlock_reset;
+		}
 
-	kiq->pmf->kiq_invalidate_tlbs(ring, pasid, flush_type, all_hub);
-	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-	if (r) {
-		amdgpu_ring_undo(ring);
+		amdgpu_ring_commit(ring);
 		spin_unlock(&adev->gfx.kiq[inst].ring_lock);
-		goto error_unlock_reset;
-	}
-
-	amdgpu_ring_commit(ring);
-	spin_unlock(&adev->gfx.kiq[inst].ring_lock);
-	r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
-	if (r < 1) {
-		dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-		r = -ETIME;
-		goto error_unlock_reset;
+		if (amdgpu_fence_wait_polling(ring, seq, usec_timeout) < 1) {
+			dev_err(adev->dev, "timeout waiting for kiq fence\n");
+			r = -ETIME;
+		}
 	}
-	r = 0;
 
 error_unlock_reset:
 	up_read(&adev->reset_domain->sem);

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

@@ -554,21 +554,6 @@ TRACE_EVENT(amdgpu_reset_reg_dumps,
 		      __entry->value)
 );
 
-TRACE_EVENT(amdgpu_runpm_reference_dumps,
-	    TP_PROTO(uint32_t index, const char *func),
-	    TP_ARGS(index, func),
-	    TP_STRUCT__entry(
-			     __field(uint32_t, index)
-			     __string(func, func)
-			     ),
-	    TP_fast_assign(
-			   __entry->index = index;
-			   __assign_str(func);
-			   ),
-	    TP_printk("amdgpu runpm reference dump 0x%x: 0x%s\n",
-		      __entry->index,
-		      __get_str(func))
-);
 #undef AMDGPU_JOB_GET_TIMELINE_NAME
 #endif
 

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c

@@ -4195,9 +4195,10 @@ static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev, int xcc_i
 static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
 				 struct amdgpu_cu_info *cu_info)
 {
-	int i, j, k, counter, xcc_id, active_cu_number = 0;
-	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
+	int i, j, k, prev_counter, counter, xcc_id, active_cu_number = 0;
+	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0, tmp;
 	unsigned disable_masks[4 * 4];
+	bool is_symmetric_cus;
 
 	if (!adev || !cu_info)
 		return -EINVAL;
@@ -4215,6 +4216,7 @@ static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (xcc_id = 0; xcc_id < NUM_XCC(adev->gfx.xcc_mask); xcc_id++) {
+		is_symmetric_cus = true;
 		for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 			for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 				mask = 1;
@@ -4242,6 +4244,15 @@ static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
 					ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
 				cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 			}
+			if (i && is_symmetric_cus && prev_counter != counter)
+				is_symmetric_cus = false;
+			prev_counter = counter;
+		}
+		if (is_symmetric_cus) {
+			tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_RELAUNCH_DISABLE, 1);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_DISPATCH_DISABLE, 1);
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG, tmp);
 		}
 		gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
 					    xcc_id);

drivers/gpu/drm/amd/amdgpu/mes_v11_0.c

@@ -154,18 +154,18 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 						    void *pkt, int size,
 						    int api_status_off)
 {
-	int ndw = size / 4;
-	signed long r;
-	union MESAPI__MISC *x_pkt = pkt;
-	struct MES_API_STATUS *api_status;
+	union MESAPI__QUERY_MES_STATUS mes_status_pkt;
+	signed long timeout = 3000000; /* 3000 ms */
 	struct amdgpu_device *adev = mes->adev;
 	struct amdgpu_ring *ring = &mes->ring;
-	unsigned long flags;
-	signed long timeout = 3000000; /* 3000 ms */
+	struct MES_API_STATUS *api_status;
+	union MESAPI__MISC *x_pkt = pkt;
 	const char *op_str, *misc_op_str;
-	u32 fence_offset;
-	u64 fence_gpu_addr;
-	u64 *fence_ptr;
+	unsigned long flags;
+	u64 status_gpu_addr;
+	u32 status_offset;
+	u64 *status_ptr;
+	signed long r;
 	int ret;
 
 	if (x_pkt->header.opcode >= MES_SCH_API_MAX)
@@ -177,44 +177,55 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 		/* Worst case in sriov where all other 15 VF timeout, each VF needs about 600ms */
 		timeout = 15 * 600 * 1000;
 	}
-	BUG_ON(size % 4 != 0);
 
-	ret = amdgpu_device_wb_get(adev, &fence_offset);
+	ret = amdgpu_device_wb_get(adev, &status_offset);
 	if (ret)
 		return ret;
-	fence_gpu_addr =
-		adev->wb.gpu_addr + (fence_offset * 4);
-	fence_ptr = (u64 *)&adev->wb.wb[fence_offset];
-	*fence_ptr = 0;
+
+	status_gpu_addr = adev->wb.gpu_addr + (status_offset * 4);
+	status_ptr = (u64 *)&adev->wb.wb[status_offset];
+	*status_ptr = 0;
 
 	spin_lock_irqsave(&mes->ring_lock, flags);
-	if (amdgpu_ring_alloc(ring, ndw)) {
-		spin_unlock_irqrestore(&mes->ring_lock, flags);
-		amdgpu_device_wb_free(adev, fence_offset);
-		return -ENOMEM;
-	}
+	r = amdgpu_ring_alloc(ring, (size + sizeof(mes_status_pkt)) / 4);
+	if (r)
+		goto error_unlock_free;
 
 	api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off);
-	api_status->api_completion_fence_addr = fence_gpu_addr;
+	api_status->api_completion_fence_addr = status_gpu_addr;
 	api_status->api_completion_fence_value = 1;
 
-	amdgpu_ring_write_multiple(ring, pkt, ndw);
+	amdgpu_ring_write_multiple(ring, pkt, size / 4);
+
+	memset(&mes_status_pkt, 0, sizeof(mes_status_pkt));
+	mes_status_pkt.header.type = MES_API_TYPE_SCHEDULER;
+	mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS;
+	mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS;
+	mes_status_pkt.api_status.api_completion_fence_addr =
+		ring->fence_drv.gpu_addr;
+	mes_status_pkt.api_status.api_completion_fence_value =
+		++ring->fence_drv.sync_seq;
+
+	amdgpu_ring_write_multiple(ring, &mes_status_pkt,
+				   sizeof(mes_status_pkt) / 4);
+
 	amdgpu_ring_commit(ring);
 	spin_unlock_irqrestore(&mes->ring_lock, flags);
 
 	op_str = mes_v11_0_get_op_string(x_pkt);
 	misc_op_str = mes_v11_0_get_misc_op_string(x_pkt);
 
 	if (misc_op_str)
-		dev_dbg(adev->dev, "MES msg=%s (%s) was emitted\n", op_str, misc_op_str);
+		dev_dbg(adev->dev, "MES msg=%s (%s) was emitted\n", op_str,
+			misc_op_str);
 	else if (op_str)
 		dev_dbg(adev->dev, "MES msg=%s was emitted\n", op_str);
 	else
-		dev_dbg(adev->dev, "MES msg=%d was emitted\n", x_pkt->header.opcode);
+		dev_dbg(adev->dev, "MES msg=%d was emitted\n",
+			x_pkt->header.opcode);
 
-	r = amdgpu_mes_fence_wait_polling(fence_ptr, (u64)1, timeout);
-	amdgpu_device_wb_free(adev, fence_offset);
-	if (r < 1) {
+	r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, timeout);
+	if (r < 1 || !*status_ptr) {
 
 		if (misc_op_str)
 			dev_err(adev->dev, "MES failed to respond to msg=%s (%s)\n",
@@ -229,10 +240,19 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 		while (halt_if_hws_hang)
 			schedule();
 
-		return -ETIMEDOUT;
+		r = -ETIMEDOUT;
+		goto error_wb_free;
 	}
 
+	amdgpu_device_wb_free(adev, status_offset);
 	return 0;
+
+error_unlock_free:
+	spin_unlock_irqrestore(&mes->ring_lock, flags);
+
+error_wb_free:
+	amdgpu_device_wb_free(adev, status_offset);
+	return r;
 }
 
 static int convert_to_mes_queue_type(int queue_type)

drivers/gpu/drm/amd/amdgpu/psp_v14_0.c

@@ -32,7 +32,9 @@
 #include "mp/mp_14_0_2_sh_mask.h"
 
 MODULE_FIRMWARE("amdgpu/psp_14_0_2_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_2_ta.bin");
 MODULE_FIRMWARE("amdgpu/psp_14_0_3_sos.bin");
+MODULE_FIRMWARE("amdgpu/psp_14_0_3_ta.bin");
 
 /* For large FW files the time to complete can be very long */
 #define USBC_PD_POLLING_LIMIT_S 240
@@ -64,6 +66,9 @@ static int psp_v14_0_init_microcode(struct psp_context *psp)
 	case IP_VERSION(14, 0, 2):
 	case IP_VERSION(14, 0, 3):
 		err = psp_init_sos_microcode(psp, ucode_prefix);
+		if (err)
+			return err;
+		err = psp_init_ta_microcode(psp, ucode_prefix);
 		if (err)
 			return err;
 		break;

drivers/gpu/drm/amd/display/Kconfig

@@ -8,7 +8,7 @@ config DRM_AMD_DC
 	depends on BROKEN || !CC_IS_CLANG || ARM64 || RISCV || SPARC64 || X86_64
 	select SND_HDA_COMPONENT if SND_HDA_CORE
 	# !CC_IS_CLANG: https://github.com/ClangBuiltLinux/linux/issues/1752
-	select DRM_AMD_DC_FP if ARCH_HAS_KERNEL_FPU_SUPPORT && (!ARM64 || !CC_IS_CLANG)
+	select DRM_AMD_DC_FP if ARCH_HAS_KERNEL_FPU_SUPPORT && !(CC_IS_CLANG && (ARM64 || RISCV))
 	help
 	  Choose this option if you want to use the new display engine
 	  support for AMDGPU. This adds required support for Vega and