drm/amdgpu: fix non-x86 GPU VCRAT parsing

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

@@ -88,6 +88,10 @@ static int aldebaran_mode2_suspend_ip(struct amdgpu_device *adev)
 	uint32_t ip_block;
 	int r, i;
 
+	/* Skip suspend of SDMA IP versions >= 4.4.2. They are multi-aid */
+	if (adev->aid_mask)
+		ip_block_mask &= ~BIT(AMD_IP_BLOCK_TYPE_SDMA);
+
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
 

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

@@ -1364,6 +1364,10 @@ static int unmap_bo_from_gpuvm(struct kgd_mem *mem,
 
 	(void)amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
 
+	/* VM entity stopped if process killed, don't clear freed pt bo */
+	if (!amdgpu_vm_ready(vm))
+		return 0;
+
 	(void)amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
 
 	(void)amdgpu_sync_fence(sync, bo_va->last_pt_update, GFP_KERNEL);

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

@@ -181,19 +181,22 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev)
 	u8 frev, crev;
 	int usage_bytes = 0;
 
-	if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
-		if (frev == 2 && crev == 1) {
-			fw_usage_v2_1 =
-				(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_1(adev,
-					fw_usage_v2_1,
-					&usage_bytes);
-		} else if (frev >= 2 && crev >= 2) {
-			fw_usage_v2_2 =
-				(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_2(adev,
-					fw_usage_v2_2,
-					&usage_bytes);
+	/* Skip atomfirmware allocation for SRIOV VFs when dynamic crit regn is enabled */
+	if (!(amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled)) {
+		if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
+			if (frev == 2 && crev == 1) {
+				fw_usage_v2_1 =
+					(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_1(adev,
+						fw_usage_v2_1,
+						&usage_bytes);
+			} else if (frev >= 2 && crev >= 2) {
+				fw_usage_v2_2 =
+					(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_2(adev,
+						fw_usage_v2_2,
+						&usage_bytes);
+			}
 		}
 	}
 

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

@@ -96,11 +96,12 @@ void amdgpu_bios_release(struct amdgpu_device *adev)
  * part of the system bios.  On boot, the system bios puts a
  * copy of the igp rom at the start of vram if a discrete card is
  * present.
- * For SR-IOV, the vbios image is also put in VRAM in the VF.
+ * For SR-IOV, if dynamic critical region is not enabled,
+ * the vbios image is also put at the start of VRAM in the VF.
  */
 static bool amdgpu_read_bios_from_vram(struct amdgpu_device *adev)
 {
-	uint8_t __iomem *bios;
+	uint8_t __iomem *bios = NULL;
 	resource_size_t vram_base;
 	resource_size_t size = 256 * 1024; /* ??? */
 
@@ -114,18 +115,33 @@ static bool amdgpu_read_bios_from_vram(struct amdgpu_device *adev)
 
 	adev->bios = NULL;
 	vram_base = pci_resource_start(adev->pdev, 0);
-	bios = ioremap_wc(vram_base, size);
-	if (!bios)
-		return false;
 
 	adev->bios = kmalloc(size, GFP_KERNEL);
-	if (!adev->bios) {
-		iounmap(bios);
+	if (!adev->bios)
 		return false;
+
+	/* For SRIOV with dynamic critical region is enabled,
+	 * the vbios image is put at a dynamic offset of VRAM in the VF.
+	 * If dynamic critical region is disabled, follow the existing logic as on baremetal.
+	 */
+	if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+		if (amdgpu_virt_get_dynamic_data_info(adev,
+				AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID, adev->bios, (uint64_t *)&size)) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+	} else {
+		bios = ioremap_wc(vram_base, size);
+		if (!bios) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+
+		memcpy_fromio(adev->bios, bios, size);
+		iounmap(bios);
 	}
+
 	adev->bios_size = size;
-	memcpy_fromio(adev->bios, bios, size);
-	iounmap(bios);
 
 	if (!check_atom_bios(adev, size)) {
 		amdgpu_bios_release(adev);

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

@@ -2779,6 +2779,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 		r = amdgpu_virt_request_full_gpu(adev, true);
 		if (r)
 			return r;
+
+		r = amdgpu_virt_init_critical_region(adev);
+		if (r)
+			return r;
 	}
 
 	switch (adev->asic_type) {
@@ -7329,10 +7333,17 @@ void amdgpu_device_flush_hdp(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.connected_to_cpu)
 		return;
 
-	if (ring && ring->funcs->emit_hdp_flush)
+	if (ring && ring->funcs->emit_hdp_flush) {
 		amdgpu_ring_emit_hdp_flush(ring);
-	else
-		amdgpu_asic_flush_hdp(adev, ring);
+		return;
+	}
+
+	if (!ring && amdgpu_sriov_runtime(adev)) {
+		if (!amdgpu_kiq_hdp_flush(adev))
+			return;
+	}
+
+	amdgpu_asic_flush_hdp(adev, ring);
 }
 
 void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev,

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

@@ -299,9 +299,25 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
 		vram_size <<= 20;
 
 	if (sz_valid) {
-		uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
-		amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
-					  adev->mman.discovery_tmr_size, false);
+		if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+			/* For SRIOV VFs with dynamic critical region enabled,
+			 * we will get the IPD binary via below call.
+			 * If dynamic critical is disabled, fall through to normal seq.
+			 */
+			if (amdgpu_virt_get_dynamic_data_info(adev,
+						AMD_SRIOV_MSG_IPD_TABLE_ID, binary,
+						(uint64_t *)&adev->mman.discovery_tmr_size)) {
+				dev_err(adev->dev,
+						"failed to read discovery info from dynamic critical region.");
+				ret = -EINVAL;
+				goto exit;
+			}
+		} else {
+			uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
+
+			amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
+					adev->mman.discovery_tmr_size, false);
+		}
 	} else {
 		ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary);
 	}
@@ -310,7 +326,7 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
 		dev_err(adev->dev,
 			"failed to read discovery info from memory, vram size read: %llx",
 			vram_size);
-
+exit:
 	return ret;
 }
 

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

@@ -260,11 +260,24 @@ static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 
 #ifdef HAVE_STRUCT_DMA_BUF_ATTACH_OPS_ALLOW_PEER2PEER
+	/*
+	 * Disable peer-to-peer access for DCC-enabled VRAM surfaces on GFX12+.
+	 * Such buffers cannot be safely accessed over P2P due to device-local
+	 * compression metadata. Fallback to system-memory path instead.
+	 * Device supports GFX12 (GC 12.x or newer)
+	 * BO was created with the AMDGPU_GEM_CREATE_GFX12_DCC flag
+	 *
+	 */
+	if ((adev->ip_versions[GC_HWIP][0] >= IP_VERSION(12, 0, 0)) &&
+		bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC) {
+		attach->peer2peer = false;
+		goto update_vm;
+	}
 	if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) &&
 	    pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
 		attach->peer2peer = false;
 #endif
-
+update_vm:
 	amdgpu_vm_bo_update_shared(bo);
 
 	return 0;

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

@@ -33,6 +33,7 @@
 #include "amdgpu_reset.h"
 #include "amdgpu_xcp.h"
 #include "amdgpu_xgmi.h"
+#include "amdgpu_mes.h"
 #include "nvd.h"
 
 /* delay 0.1 second to enable gfx off feature */
@@ -1194,6 +1195,75 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint3
 	dev_err(adev->dev, "failed to write reg:%x\n", reg);
 }
 
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev)
+{
+	signed long r, cnt = 0;
+	unsigned long flags;
+	uint32_t seq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *ring = &kiq->ring;
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (adev->enable_mes_kiq && adev->mes.ring[0].sched.ready)
+		return amdgpu_mes_hdp_flush(adev);
+
+	if (!ring->funcs->emit_hdp_flush) {
+		return -EOPNOTSUPP;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
+	amdgpu_ring_emit_hdp_flush(ring);
+	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+	if (r)
+		goto failed_undo;
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+	/* don't wait anymore for gpu reset case because this way may
+	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
+	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
+	 * never return if we keep waiting in virt_kiq_rreg, which cause
+	 * gpu_recover() hang there.
+	 *
+	 * also don't wait anymore for IRQ context
+	 * */
+	if (r < 1 && (amdgpu_in_reset(adev) || in_interrupt()))
+		goto failed_kiq_hdp_flush;
+
+	might_sleep();
+	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_hdp_flush;
+
+		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+	}
+
+	if (cnt > MAX_KIQ_REG_TRY) {
+		dev_err(adev->dev, "failed to flush HDP via KIQ timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+
+failed_undo:
+	amdgpu_ring_undo(ring);
+failed_unlock:
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+failed_kiq_hdp_flush:
+	dev_err(adev->dev, "failed to flush HDP via KIQ\n");
+	return r < 0 ? r : -EIO;
+}
+
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev)
 {
 	if (amdgpu_num_kcq == -1) {
@@ -2484,3 +2554,4 @@ void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev)
 			    &amdgpu_debugfs_compute_sched_mask_fops);
 #endif
 }
+

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

@@ -626,6 +626,7 @@ int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
 				  struct amdgpu_iv_entry *entry);
 uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id);
 void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id);
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev);
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev);
 void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev, uint32_t ucode_id);
 

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

@@ -370,7 +370,7 @@ static int amdgpu_debugfs_jpeg_sched_mask_set(void *data, u64 val)
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
 		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
 			ring = &adev->jpeg.inst[i].ring_dec[j];
-			if (val & (BIT_ULL(1) << ((i * adev->jpeg.num_jpeg_rings) + j)))
+			if (val & (BIT_ULL((i * adev->jpeg.num_jpeg_rings) + j)))
 				ring->sched.ready = true;
 			else
 				ring->sched.ready = false;

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

@@ -523,6 +523,18 @@ int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 	return r;
 }
 
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev)
+{
+	uint32_t hdp_flush_req_offset, hdp_flush_done_offset, ref_and_mask;
+
+	hdp_flush_req_offset = adev->nbio.funcs->get_hdp_flush_req_offset(adev);
+	hdp_flush_done_offset = adev->nbio.funcs->get_hdp_flush_done_offset(adev);
+	ref_and_mask = adev->nbio.hdp_flush_reg->ref_and_mask_cp0;
+
+	return amdgpu_mes_reg_write_reg_wait(adev, hdp_flush_req_offset, hdp_flush_done_offset,
+					     ref_and_mask, ref_and_mask);
+}
+
 int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
 				uint64_t process_context_addr,
 				uint32_t spi_gdbg_per_vmid_cntl,

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

@@ -426,6 +426,7 @@ int amdgpu_mes_wreg(struct amdgpu_device *adev,
 int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 				  uint32_t reg0, uint32_t reg1,
 				  uint32_t ref, uint32_t mask);
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev);
 int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
 				uint64_t process_context_addr,
 				uint32_t spi_gdbg_per_vmid_cntl,