1 files changed, 82 insertions, 49 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 86afd37b098d..4bfc0c8ab764 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -321,7 +321,7 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 	/* Return gpu_id as doorbell offset for mmap usage */
 	args->doorbell_offset = KFD_MMAP_TYPE_DOORBELL;
 	args->doorbell_offset |= KFD_MMAP_GPU_ID(args->gpu_id);
-	if (KFD_IS_SOC15(dev->device_info->asic_family))
+	if (KFD_IS_SOC15(dev))
 		/* On SOC15 ASICs, include the doorbell offset within the
 		 * process doorbell frame, which is 2 pages.
 		 */
@@ -405,7 +405,7 @@ static int kfd_ioctl_update_queue(struct file *filp, struct kfd_process *p,
 
 	mutex_lock(&p->mutex);
 
-	retval = pqm_update_queue(&p->pqm, args->queue_id, &properties);
+	retval = pqm_update_queue_properties(&p->pqm, args->queue_id, &properties);
 
 	mutex_unlock(&p->mutex);
 
@@ -418,7 +418,7 @@ static int kfd_ioctl_set_cu_mask(struct file *filp, struct kfd_process *p,
 	int retval;
 	const int max_num_cus = 1024;
 	struct kfd_ioctl_set_cu_mask_args *args = data;
-	struct queue_properties properties;
+	struct mqd_update_info minfo = {0};
 	uint32_t __user *cu_mask_ptr = (uint32_t __user *)args->cu_mask_ptr;
 	size_t cu_mask_size = sizeof(uint32_t) * (args->num_cu_mask / 32);
 
@@ -428,8 +428,8 @@ static int kfd_ioctl_set_cu_mask(struct file *filp, struct kfd_process *p,
 		return -EINVAL;
 	}
 
-	properties.cu_mask_count = args->num_cu_mask;
-	if (properties.cu_mask_count == 0) {
+	minfo.cu_mask.count = args->num_cu_mask;
+	if (minfo.cu_mask.count == 0) {
 		pr_debug("CU mask cannot be 0");
 		return -EINVAL;
 	}
@@ -438,32 +438,33 @@ static int kfd_ioctl_set_cu_mask(struct file *filp, struct kfd_process *p,
 	 * limit of max_num_cus bits.  We can then just drop any CU mask bits
 	 * past max_num_cus bits and just use the first max_num_cus bits.
 	 */
-	if (properties.cu_mask_count > max_num_cus) {
+	if (minfo.cu_mask.count > max_num_cus) {
 		pr_debug("CU mask cannot be greater than 1024 bits");
-		properties.cu_mask_count = max_num_cus;
+		minfo.cu_mask.count = max_num_cus;
 		cu_mask_size = sizeof(uint32_t) * (max_num_cus/32);
 	}
 
-	properties.cu_mask = kzalloc(cu_mask_size, GFP_KERNEL);
-	if (!properties.cu_mask)
+	minfo.cu_mask.ptr = kzalloc(cu_mask_size, GFP_KERNEL);
+	if (!minfo.cu_mask.ptr)
 		return -ENOMEM;
 
-	retval = copy_from_user(properties.cu_mask, cu_mask_ptr, cu_mask_size);
+	retval = copy_from_user(minfo.cu_mask.ptr, cu_mask_ptr, cu_mask_size);
 	if (retval) {
 		pr_debug("Could not copy CU mask from userspace");
-		kfree(properties.cu_mask);
-		return -EFAULT;
+		retval = -EFAULT;
+		goto out;
 	}
 
+	minfo.update_flag = UPDATE_FLAG_CU_MASK;
+
 	mutex_lock(&p->mutex);
 
-	retval = pqm_set_cu_mask(&p->pqm, args->queue_id, &properties);
+	retval = pqm_update_mqd(&p->pqm, args->queue_id, &minfo);
 
 	mutex_unlock(&p->mutex);
 
-	if (retval)
-		kfree(properties.cu_mask);
-
+out:
+	kfree(minfo.cu_mask.ptr);
 	return retval;
 }
 
@@ -579,7 +580,7 @@ static int kfd_ioctl_dbg_register(struct file *filep,
 	if (!dev)
 		return -EINVAL;
 
-	if (dev->device_info->asic_family == CHIP_CARRIZO) {
+	if (dev->adev->asic_type == CHIP_CARRIZO) {
 		pr_debug("kfd_ioctl_dbg_register not supported on CZ\n");
 		return -EINVAL;
 	}
@@ -630,7 +631,7 @@ static int kfd_ioctl_dbg_unregister(struct file *filep,
 	if (!dev || !dev->dbgmgr)
 		return -EINVAL;
 
-	if (dev->device_info->asic_family == CHIP_CARRIZO) {
+	if (dev->adev->asic_type == CHIP_CARRIZO) {
 		pr_debug("kfd_ioctl_dbg_unregister not supported on CZ\n");
 		return -EINVAL;
 	}
@@ -675,7 +676,7 @@ static int kfd_ioctl_dbg_address_watch(struct file *filep,
 	if (!dev)
 		return -EINVAL;
 
-	if (dev->device_info->asic_family == CHIP_CARRIZO) {
+	if (dev->adev->asic_type == CHIP_CARRIZO) {
 		pr_debug("kfd_ioctl_dbg_wave_control not supported on CZ\n");
 		return -EINVAL;
 	}
@@ -783,7 +784,7 @@ static int kfd_ioctl_dbg_wave_control(struct file *filep,
 	if (!dev)
 		return -EINVAL;
 
-	if (dev->device_info->asic_family == CHIP_CARRIZO) {
+	if (dev->adev->asic_type == CHIP_CARRIZO) {
 		pr_debug("kfd_ioctl_dbg_wave_control not supported on CZ\n");
 		return -EINVAL;
 	}
@@ -850,7 +851,7 @@ static int kfd_ioctl_get_clock_counters(struct file *filep,
 	dev = kfd_device_by_id(args->gpu_id);
 	if (dev)
 		/* Reading GPU clock counter from KGD */
-		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->kgd);
+		args->gpu_clock_counter = amdgpu_amdkfd_get_gpu_clock_counter(dev->adev);
 	else
 		/* Node without GPU resource */
 		args->gpu_clock_counter = 0;
@@ -1011,11 +1012,6 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
 		void *mem, *kern_addr;
 		uint64_t size;
 
-		if (p->signal_page) {
-			pr_err("Event page is already set\n");
-			return -EINVAL;
-		}
-
 		kfd = kfd_device_by_id(GET_GPU_ID(args->event_page_offset));
 		if (!kfd) {
 			pr_err("Getting device by id failed in %s\n", __func__);
@@ -1023,6 +1019,13 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
 		}
 
 		mutex_lock(&p->mutex);
+
+		if (p->signal_page) {
+			pr_err("Event page is already set\n");
+			err = -EINVAL;
+			goto out_unlock;
+		}
+
 		pdd = kfd_bind_process_to_device(kfd, p);
 		if (IS_ERR(pdd)) {
 			err = PTR_ERR(pdd);
@@ -1037,20 +1040,24 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
 			err = -EINVAL;
 			goto out_unlock;
 		}
-		mutex_unlock(&p->mutex);
 
-		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd,
+		err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->adev,
 						mem, &kern_addr, &size);
 		if (err) {
 			pr_err("Failed to map event page to kernel\n");
-			return err;
+			goto out_unlock;
 		}
 
 		err = kfd_event_page_set(p, kern_addr, size);
 		if (err) {
 			pr_err("Failed to set event page\n");
-			return err;
+			amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kfd->adev, mem);
+			goto out_unlock;
 		}
+
+		p->signal_handle = args->event_page_offset;
+
+		mutex_unlock(&p->mutex);
 	}
 
 	err = kfd_event_create(filp, p, args->event_type,
@@ -1130,7 +1137,7 @@ static int kfd_ioctl_set_scratch_backing_va(struct file *filep,
 	if (dev->dqm->sched_policy == KFD_SCHED_POLICY_NO_HWS &&
 	    pdd->qpd.vmid != 0 && dev->kfd2kgd->set_scratch_backing_va)
 		dev->kfd2kgd->set_scratch_backing_va(
-			dev->kgd, args->va_addr, pdd->qpd.vmid);
+			dev->adev, args->va_addr, pdd->qpd.vmid);
 
 	return 0;
 
@@ -1151,7 +1158,7 @@ static int kfd_ioctl_get_tile_config(struct file *filep,
 	if (!dev)
 		return -EINVAL;
 
-	amdgpu_amdkfd_get_tile_config(dev->kgd, &config);
+	amdgpu_amdkfd_get_tile_config(dev->adev, &config);
 
 	args->gb_addr_config = config.gb_addr_config;
 	args->num_banks = config.num_banks;
@@ -1237,7 +1244,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev)
 	if (dev->use_iommu_v2)
 		return false;
 
-	amdgpu_amdkfd_get_local_mem_info(dev->kgd, &mem_info);
+	amdgpu_amdkfd_get_local_mem_info(dev->adev, &mem_info);
 	if (mem_info.local_mem_size_private == 0 &&
 			mem_info.local_mem_size_public > 0)
 		return true;
@@ -1259,6 +1266,23 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	if (args->size == 0)
 		return -EINVAL;
 
+#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
+	/* Flush pending deferred work to avoid racing with deferred actions
+	 * from previous memory map changes (e.g. munmap).
+	 */
+	svm_range_list_lock_and_flush_work(&p->svms, current->mm);
+	mutex_lock(&p->svms.lock);
+	mmap_write_unlock(current->mm);
+	if (interval_tree_iter_first(&p->svms.objects,
+				     args->va_addr >> PAGE_SHIFT,
+				     (args->va_addr + args->size - 1) >> PAGE_SHIFT)) {
+		pr_err("Address: 0x%llx already allocated by SVM\n",
+			args->va_addr);
+		mutex_unlock(&p->svms.lock);
+		return -EADDRINUSE;
+	}
+	mutex_unlock(&p->svms.lock);
+#endif
 	dev = kfd_device_by_id(args->gpu_id);
 	if (!dev)
 		return -EINVAL;
@@ -1289,7 +1313,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 			err = -EINVAL;
 			goto err_unlock;
 		}
-		offset = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
+		offset = dev->adev->rmmio_remap.bus_addr;
 		if (!offset) {
 			err = -ENOMEM;
 			goto err_unlock;
@@ -1297,7 +1321,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	}
 
 	err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
-		dev->kgd, args->va_addr, args->size,
+		dev->adev, args->va_addr, args->size,
 		pdd->drm_priv, (struct kgd_mem **) &mem, &offset,
 		flags);
 
@@ -1329,7 +1353,7 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
 	return 0;
 
 err_free:
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem,
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem,
 					       pdd->drm_priv, NULL);
 err_unlock:
 	mutex_unlock(&p->mutex);
@@ -1351,6 +1375,15 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep,
 		return -EINVAL;
 
 	mutex_lock(&p->mutex);
+	/*
+	 * Safeguard to prevent user space from freeing signal BO.
+	 * It will be freed at process termination.
+	 */
+	if (p->signal_handle && (p->signal_handle == args->handle)) {
+		pr_err("Free signal BO is not allowed\n");
+		ret = -EPERM;
+		goto err_unlock;
+	}
 
 	pdd = kfd_get_process_device_data(dev, p);
 	if (!pdd) {
@@ -1366,7 +1399,7 @@ static int kfd_ioctl_free_memory_of_gpu(struct file *filep,
 		goto err_unlock;
 	}
 
-	ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd,
+	ret = amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev,
 				(struct kgd_mem *)mem, pdd->drm_priv, &size);
 
 	/* If freeing the buffer failed, leave the handle in place for
@@ -1451,7 +1484,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 			goto get_mem_obj_from_handle_failed;
 		}
 		err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
-			peer->kgd, (struct kgd_mem *)mem,
+			peer->adev, (struct kgd_mem *)mem,
 			peer_pdd->drm_priv, &table_freed);
 		if (err) {
 			pr_err("Failed to map to gpu %d/%d\n",
@@ -1463,7 +1496,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 
 	mutex_unlock(&p->mutex);
 
-	err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd, (struct kgd_mem *) mem, true);
+	err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev, (struct kgd_mem *) mem, true);
 	if (err) {
 		pr_debug("Sync memory failed, wait interrupted by user signal\n");
 		goto sync_memory_failed;
@@ -1560,7 +1593,7 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 			goto get_mem_obj_from_handle_failed;
 		}
 		err = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
-			peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv);
+			peer->adev, (struct kgd_mem *)mem, peer_pdd->drm_priv);
 		if (err) {
 			pr_err("Failed to unmap from gpu %d/%d\n",
 			       i, args->n_devices);
@@ -1570,8 +1603,8 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 	}
 	mutex_unlock(&p->mutex);
 
-	if (dev->device_info->asic_family == CHIP_ALDEBARAN) {
-		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd,
+	if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2)) {
+		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->adev,
 				(struct kgd_mem *) mem, true);
 		if (err) {
 			pr_debug("Sync memory failed, wait interrupted by user signal\n");
@@ -1647,7 +1680,7 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 {
 	struct kfd_ioctl_get_dmabuf_info_args *args = data;
 	struct kfd_dev *dev = NULL;
-	struct kgd_dev *dma_buf_kgd;
+	struct amdgpu_device *dmabuf_adev;
 	void *metadata_buffer = NULL;
 	uint32_t flags;
 	unsigned int i;
@@ -1667,15 +1700,15 @@ static int kfd_ioctl_get_dmabuf_info(struct file *filep,
 	}
 
 	/* Get dmabuf info from KGD */
-	r = amdgpu_amdkfd_get_dmabuf_info(dev->kgd, args->dmabuf_fd,
-					  &dma_buf_kgd, &args->size,
+	r = amdgpu_amdkfd_get_dmabuf_info(dev->adev, args->dmabuf_fd,
+					  &dmabuf_adev, &args->size,
 					  metadata_buffer, args->metadata_size,
 					  &args->metadata_size, &flags);
 	if (r)
 		goto exit;
 
 	/* Reverse-lookup gpu_id from kgd pointer */
-	dev = kfd_device_by_kgd(dma_buf_kgd);
+	dev = kfd_device_by_adev(dmabuf_adev);
 	if (!dev) {
 		r = -EINVAL;
 		goto exit;
@@ -1725,7 +1758,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 		goto err_unlock;
 	}
 
-	r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->kgd, dmabuf,
+	r = amdgpu_amdkfd_gpuvm_import_dmabuf(dev->adev, dmabuf,
 					      args->va_addr, pdd->drm_priv,
 					      (struct kgd_mem **)&mem, &size,
 					      NULL);
@@ -1746,7 +1779,7 @@ static int kfd_ioctl_import_dmabuf(struct file *filep,
 	return 0;
 
 err_free:
-	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, (struct kgd_mem *)mem,
+	amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, (struct kgd_mem *)mem,
 					       pdd->drm_priv, NULL);
 err_unlock:
 	mutex_unlock(&p->mutex);
@@ -2033,7 +2066,7 @@ static int kfd_mmio_mmap(struct kfd_dev *dev, struct kfd_process *process,
 	if (vma->vm_end - vma->vm_start != PAGE_SIZE)
 		return -EINVAL;
 
-	address = amdgpu_amdkfd_get_mmio_remap_phys_addr(dev->kgd);
+	address = dev->adev->rmmio_remap.bus_addr;
 
 	vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND | VM_NORESERVE |
 				VM_DONTDUMP | VM_PFNMAP;