amd/amdgpu/amdgpu_gem.c

/*
 * Copyright 2008 Advanced Micro Devices, Inc.
 * Copyright 2008 Red Hat Inc.
 * Copyright 2009 Jerome Glisse.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 *
 * Authors: Dave Airlie
 *          Alex Deucher
 *          Jerome Glisse
 */
#include <linux/ktime.h>
#include <linux/module.h>
#include <linux/overflow.h>
#include <linux/pagemap.h>
#include <linux/pci.h>
#include <linux/dma-buf.h>
#include <linux/dma-fence-unwrap.h>
#include <linux/uaccess.h>

#include <drm/amdgpu_drm.h>
#include <drm/drm_drv.h>
#include <drm/drm_exec.h>
#include <drm/drm_gem_ttm_helper.h>
#include <drm/ttm/ttm_tt.h>
#include <drm/drm_syncobj.h>

#include "amdgpu.h"
#include "amdgpu_display.h"
#include "amdgpu_dma_buf.h"
#include "amdgpu_hmm.h"
#include "amdgpu_xgmi.h"
#include "amdgpu_vm.h"

static int
amdgpu_gem_add_input_fence(struct drm_file *filp,
			   uint64_t syncobj_handles_array,
			   uint32_t num_syncobj_handles)
{
	struct dma_fence *fence;
	uint32_t *syncobj_handles;
	int ret, i;

	if (!num_syncobj_handles)
		return 0;

	syncobj_handles = memdup_user(u64_to_user_ptr(syncobj_handles_array),
				      size_mul(sizeof(uint32_t), num_syncobj_handles));
	if (IS_ERR(syncobj_handles))
		return PTR_ERR(syncobj_handles);

	for (i = 0; i < num_syncobj_handles; i++) {

		if (!syncobj_handles[i]) {
			ret = -EINVAL;
			goto free_memdup;
		}

		ret = drm_syncobj_find_fence(filp, syncobj_handles[i], 0, 0, &fence);
		if (ret)
			goto free_memdup;

		dma_fence_wait(fence, false);

		/* TODO: optimize async handling */
		dma_fence_put(fence);
	}

free_memdup:
	kfree(syncobj_handles);
	return ret;
}

static int
amdgpu_gem_update_timeline_node(struct drm_file *filp,
				uint32_t syncobj_handle,
				uint64_t point,
				struct drm_syncobj **syncobj,
				struct dma_fence_chain **chain)
{
	if (!syncobj_handle)
		return 0;

	/* Find the sync object */
	*syncobj = drm_syncobj_find(filp, syncobj_handle);
	if (!*syncobj)
		return -ENOENT;

	if (!point)
		return 0;

	/* Allocate the chain node */
	*chain = dma_fence_chain_alloc();
	if (!*chain) {
		drm_syncobj_put(*syncobj);
		*syncobj = NULL;
		return -ENOMEM;
	}

	return 0;
}

static vm_fault_t amdgpu_gem_fault(struct vm_fault *vmf)
{
	struct ttm_buffer_object *bo = vmf->vma->vm_private_data;
	struct drm_device *ddev = bo->base.dev;
	vm_fault_t ret;
	int idx;

	ret = ttm_bo_vm_reserve(bo, vmf);
	if (ret)
		return ret;

	if (drm_dev_enter(ddev, &idx)) {
		ret = amdgpu_bo_fault_reserve_notify(bo);
		if (ret) {
			drm_dev_exit(idx);
			goto unlock;
		}

		ret = ttm_bo_vm_fault_reserved(vmf, vmf->vma->vm_page_prot,
					       TTM_BO_VM_NUM_PREFAULT);

		drm_dev_exit(idx);
	} else {
		ret = ttm_bo_vm_dummy_page(vmf, vmf->vma->vm_page_prot);
	}
	if (ret == VM_FAULT_RETRY && !(vmf->flags & FAULT_FLAG_RETRY_NOWAIT))
		return ret;

unlock:
	dma_resv_unlock(bo->base.resv);
	return ret;
}

static const struct vm_operations_struct amdgpu_gem_vm_ops = {
	.fault = amdgpu_gem_fault,
	.open = ttm_bo_vm_open,
	.close = ttm_bo_vm_close,
	.access = ttm_bo_vm_access
};

static void amdgpu_gem_object_free(struct drm_gem_object *gobj)
{
	struct amdgpu_bo *aobj = gem_to_amdgpu_bo(gobj);

	amdgpu_hmm_unregister(aobj);
	ttm_bo_fini(&aobj->tbo);
}

int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
			     int alignment, u32 initial_domain,
			     u64 flags, enum ttm_bo_type type,
			     struct dma_resv *resv,
			     struct drm_gem_object **obj, int8_t xcp_id_plus1)
{
	struct amdgpu_bo *bo;
	struct amdgpu_bo_user *ubo;
	struct amdgpu_bo_param bp;
	int r;

	memset(&bp, 0, sizeof(bp));
	*obj = NULL;
	flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;

	bp.size = size;
	bp.byte_align = alignment;
	bp.type = type;
	bp.resv = resv;
	bp.preferred_domain = initial_domain;
	bp.flags = flags;
	bp.domain = initial_domain;
	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
	bp.xcp_id_plus1 = xcp_id_plus1;

	r = amdgpu_bo_create_user(adev, &bp, &ubo);
	if (r)
		return r;

	bo = &ubo->bo;
	*obj = &bo->tbo.base;

	return 0;
}

void amdgpu_gem_force_release(struct amdgpu_device *adev)
{
	struct drm_device *ddev = adev_to_drm(adev);
	struct drm_file *file;

	mutex_lock(&ddev->filelist_mutex);

	list_for_each_entry(file, &ddev->filelist, lhead) {
		struct drm_gem_object *gobj;
		int handle;

		WARN_ONCE(1, "Still active user space clients!\n");
		spin_lock(&file->table_lock);
		idr_for_each_entry(&file->object_idr, gobj, handle) {
			WARN_ONCE(1, "And also active allocations!\n");
			drm_gem_object_put(gobj);
		}
		idr_destroy(&file->object_idr);
		spin_unlock(&file->table_lock);
	}

	mutex_unlock(&ddev->filelist_mutex);
}

/*
 * Call from drm_gem_handle_create which appear in both new and open ioctl
 * case.
 */
static int amdgpu_gem_object_open(struct drm_gem_object *obj,
				  struct drm_file *file_priv)
{
	struct amdgpu_bo *abo = gem_to_amdgpu_bo(obj);
	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
	struct amdgpu_vm *vm = &fpriv->vm;
	struct amdgpu_bo_va *bo_va;
	struct mm_struct *mm;
	struct drm_exec exec;
	int r;

	mm = amdgpu_ttm_tt_get_usermm(abo->tbo.ttm);
	if (mm && mm != current->mm)
		return -EPERM;

	if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID &&
	    !amdgpu_vm_is_bo_always_valid(vm, abo))
		return -EPERM;

	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
	drm_exec_until_all_locked(&exec) {
		r = drm_exec_prepare_obj(&exec, &abo->tbo.base, TTM_NUM_MOVE_FENCES + 1);
		drm_exec_retry_on_contention(&exec);
		if (unlikely(r))
			goto out_unlock;

		r = amdgpu_vm_lock_pd(vm, &exec, 0);
		drm_exec_retry_on_contention(&exec);
		if (unlikely(r))
			goto out_unlock;
	}

	amdgpu_vm_bo_update_shared(abo);
	bo_va = amdgpu_vm_bo_find(vm, abo);
	if (!bo_va) {
		bo_va = amdgpu_vm_bo_add(adev, vm, abo);
		r = amdgpu_evf_mgr_attach_fence(&fpriv->evf_mgr, abo);
		if (r)
			goto out_unlock;
	} else {
		++bo_va->ref_count;
	}

	drm_exec_fini(&exec);

	/* Validate and add eviction fence to DMABuf imports with dynamic
	 * attachment in compute VMs. Re-validation will be done by
	 * amdgpu_vm_validate. Fences are on the reservation shared with the
	 * export, which is currently required to be validated and fenced
	 * already by amdgpu_amdkfd_gpuvm_restore_process_bos.
	 *
	 * Nested locking below for the case that a GEM object is opened in
	 * kfd_mem_export_dmabuf. Since the lock below is only taken for imports,
	 * but not for export, this is a different lock class that cannot lead to
	 * circular lock dependencies.
	 */
	if (!vm->is_compute_context || !vm->process_info)
		return 0;
	if (!drm_gem_is_imported(obj) ||
	    !dma_buf_is_dynamic(obj->import_attach->dmabuf))
		return 0;
	mutex_lock_nested(&vm->process_info->lock, 1);
	if (!WARN_ON(!vm->process_info->eviction_fence)) {
		r = amdgpu_amdkfd_bo_validate_and_fence(abo, AMDGPU_GEM_DOMAIN_GTT,
							&vm->process_info->eviction_fence->base);
		if (r) {
			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);

			dev_warn(adev->dev, "validate_and_fence failed: %d\n", r);
			if (ti) {
				dev_warn(adev->dev, "pid %d\n", ti->task.pid);
				amdgpu_vm_put_task_info(ti);
			}
		}
	}
	mutex_unlock(&vm->process_info->lock);
	return r;

out_unlock:
	drm_exec_fini(&exec);
	return r;
}

static void amdgpu_gem_object_close(struct drm_gem_object *obj,
				    struct drm_file *file_priv)
{
	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
	struct amdgpu_vm *vm = &fpriv->vm;

	struct dma_fence *fence = NULL;
	struct amdgpu_bo_va *bo_va;
	struct drm_exec exec;
	long r;

	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
	drm_exec_until_all_locked(&exec) {
		r = drm_exec_prepare_obj(&exec, &bo->tbo.base, 1);
		drm_exec_retry_on_contention(&exec);
		if (unlikely(r))
			goto out_unlock;

		r = amdgpu_vm_lock_pd(vm, &exec, 0);
		drm_exec_retry_on_contention(&exec);
		if (unlikely(r))
			goto out_unlock;
	}

	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
		amdgpu_evf_mgr_detach_fence(&fpriv->evf_mgr, bo);

	bo_va = amdgpu_vm_bo_find(vm, bo);
	if (!bo_va || --bo_va->ref_count)
		goto out_unlock;

	amdgpu_vm_bo_del(adev, bo_va);
	amdgpu_vm_bo_update_shared(bo);
	if (!amdgpu_vm_ready(vm))
		goto out_unlock;

	r = amdgpu_vm_clear_freed(adev, vm, &fence);
	if (unlikely(r < 0) && !drm_dev_is_unplugged(adev_to_drm(adev)))
		dev_err(adev->dev, "failed to clear page "
			"tables on GEM object close (%ld)\n", r);
	if (r || !fence)
		goto out_unlock;

	amdgpu_bo_fence(bo, fence, true);
	dma_fence_put(fence);

out_unlock:
	if (r && !drm_dev_is_unplugged(adev_to_drm(adev)))
		dev_err(adev->dev, "leaking bo va (%ld)\n", r);
	drm_exec_fini(&exec);
}

static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
{
	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);

	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
		return -EPERM;
	if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)
		return -EPERM;

	/* Workaround for Thunk bug creating PROT_NONE,MAP_PRIVATE mappings
	 * for debugger access to invisible VRAM. Should have used MAP_SHARED
	 * instead. Clearing VM_MAYWRITE prevents the mapping from ever
	 * becoming writable and makes is_cow_mapping(vm_flags) false.
	 */
	if (is_cow_mapping(vma->vm_flags) &&
	    !(vma->vm_flags & VM_ACCESS_FLAGS))
		vm_flags_clear(vma, VM_MAYWRITE);

	return drm_gem_ttm_mmap(obj, vma);
}

const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
	.free = amdgpu_gem_object_free,
	.open = amdgpu_gem_object_open,
	.close = amdgpu_gem_object_close,
	.export = amdgpu_gem_prime_export,
	.vmap = drm_gem_ttm_vmap,
	.vunmap = drm_gem_ttm_vunmap,
	.mmap = amdgpu_gem_object_mmap,
	.vm_ops = &amdgpu_gem_vm_ops,
};

/*
 * GEM ioctls.
 */
int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
			    struct drm_file *filp)
{
	struct amdgpu_device *adev = drm_to_adev(dev);
	struct amdgpu_fpriv *fpriv = filp->driver_priv;
	struct amdgpu_vm *vm = &fpriv->vm;
	union drm_amdgpu_gem_create *args = data;
	uint64_t flags = args->in.domain_flags;
	uint64_t size = args->in.bo_size;
	struct dma_resv *resv = NULL;
	struct drm_gem_object *gobj;
	uint32_t handle, initial_domain;
	int r;

	/* reject invalid gem flags */
	if (flags & ~AMDGPU_GEM_CREATE_SETTABLE_MASK)
		return -EINVAL;

	/* reject invalid gem domains */
	if (args->in.domains & ~AMDGPU_GEM_DOMAIN_MASK)
		return -EINVAL;

	if (!amdgpu_is_tmz(adev) && (flags & AMDGPU_GEM_CREATE_ENCRYPTED)) {
		DRM_NOTE_ONCE("Cannot allocate secure buffer since TMZ is disabled\n");
		return -EINVAL;
	}

	/* always clear VRAM */
	flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;

	/* create a gem object to contain this object in */
	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
		if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
			/* if gds bo is created from user space, it must be
			 * passed to bo list
			 */
			DRM_ERROR("GDS bo cannot be per-vm-bo\n");
			return -EINVAL;
		}
		flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS;
	}

	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
		r = amdgpu_bo_reserve(vm->root.bo, false);
		if (r)
			return r;

		resv = vm->root.bo->tbo.base.resv;
	}

	initial_domain = (u32)(0xffffffff & args->in.domains);
retry:
	r = amdgpu_gem_object_create(adev, size, args->in.alignment,
				     initial_domain,
				     flags, ttm_bo_type_device, resv, &gobj, fpriv->xcp_id + 1);
	if (r && r != -ERESTARTSYS) {
		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) {
			flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
			goto retry;
		}

		if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) {
			initial_domain |= AMDGPU_GEM_DOMAIN_GTT;
			goto retry;
		}
		DRM_DEBUG("Failed to allocate GEM object (%llu, %d, %llu, %d)\n",
				size, initial_domain, args->in.alignment, r);
	}

	if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) {
		if (!r) {
			struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);

			abo->parent = amdgpu_bo_ref(vm->root.bo);
		}
		amdgpu_bo_unreserve(vm->root.bo);
	}
	if (r)
		return r;

	r = drm_gem_handle_create(filp, gobj, &handle);
	/* drop reference from allocate - handle holds it now */
	drm_gem_object_put(gobj);
	if (r)
		return r;

	memset(args, 0, sizeof(*args));
	args->out.handle = handle;
	return 0;
}

int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
			     struct drm_file *filp)
{
	struct ttm_operation_ctx ctx = { true, false };
	struct amdgpu_device *adev = drm_to_adev(dev);
	struct drm_amdgpu_gem_userptr *args = data;
	struct amdgpu_fpriv *fpriv = filp->driver_priv;
	struct drm_gem_object *gobj;
	struct amdgpu_hmm_range *range;
	struct amdgpu_bo *bo;
	uint32_t handle;
	int r;

	args->addr = untagged_addr(args->addr);

	if (offset_in_page(args->addr | args->size))
		return -EINVAL;

	if (!access_ok((void __user *)(uintptr_t)args->addr, args->size))
		return -EFAULT;

	/* reject unknown flag values */
	if (args->flags & ~(AMDGPU_GEM_USERPTR_READONLY |
	    AMDGPU_GEM_USERPTR_ANONONLY | AMDGPU_GEM_USERPTR_VALIDATE |
	    AMDGPU_GEM_USERPTR_REGISTER))
		return -EINVAL;

	if (!(args->flags & AMDGPU_GEM_USERPTR_READONLY) &&
	     !(args->flags & AMDGPU_GEM_USERPTR_REGISTER)) {

		/* if we want to write to it we must install a MMU notifier */
		return -EACCES;
	}

	/* create a gem object to contain this object in */
	r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU,
				     0, ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
	if (r)
		return r;

	bo = gem_to_amdgpu_bo(gobj);
	bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT;
	bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT;
	bo->parent = amdgpu_bo_ref(fpriv->vm.root.bo);
	r = amdgpu_ttm_tt_set_userptr(&bo->tbo, args->addr, args->flags);
	if (r)
		goto release_object;

	r = amdgpu_hmm_register(bo, args->addr);
	if (r)
		goto release_object;

	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
		range = amdgpu_hmm_range_alloc(NULL);
		if (unlikely(!range))
			return -ENOMEM;
		r = amdgpu_ttm_tt_get_user_pages(bo, range);
		if (r) {
			amdgpu_hmm_range_free(range);
			goto release_object;
		}
		r = amdgpu_bo_reserve(bo, true);
		if (r)
			goto user_pages_done;

		amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);

		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
		amdgpu_bo_unreserve(bo);
		if (r)
			goto user_pages_done;
	}

	r = drm_gem_handle_create(filp, gobj, &handle);
	if (r)
		goto user_pages_done;

	args->handle = handle;

user_pages_done:
	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE)
		amdgpu_hmm_range_free(range);
release_object:
	drm_gem_object_put(gobj);

	return r;
}

int amdgpu_mode_dumb_mmap(struct drm_file *filp,
			  struct drm_device *dev,
			  uint32_t handle, uint64_t *offset_p)
{
	struct drm_gem_object *gobj;
	struct amdgpu_bo *robj;

	gobj = drm_gem_object_lookup(filp, handle);
	if (!gobj)
		return -ENOENT;

	robj = gem_to_amdgpu_bo(gobj);
	if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm) ||
	    (robj->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
		drm_gem_object_put(gobj);
		return -EPERM;
	}
	*offset_p = amdgpu_bo_mmap_offset(robj);
	drm_gem_object_put(gobj);
	return 0;
}

int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
			  struct drm_file *filp)
{
	union drm_amdgpu_gem_mmap *args = data;
	uint32_t handle = args->in.handle;

	memset(args, 0, sizeof(*args));
	return amdgpu_mode_dumb_mmap(filp, dev, handle, &args->out.addr_ptr);
}

/**
 * amdgpu_gem_timeout - calculate jiffies timeout from absolute value
 *
 * @timeout_ns: timeout in ns
 *
 * Calculate the timeout in jiffies from an absolute timeout in ns.
 */
unsigned long amdgpu_gem_timeout(uint64_t timeout_ns)
{
	unsigned long timeout_jiffies;
	ktime_t timeout;

	/* clamp timeout if it's to large */
	if (((int64_t)timeout_ns) < 0)
		return MAX_SCHEDULE_TIMEOUT;

	timeout = ktime_sub(ns_to_ktime(timeout_ns), ktime_get());
	if (ktime_to_ns(timeout) < 0)
		return 0;

	timeout_jiffies = nsecs_to_jiffies(ktime_to_ns(timeout));
	/*  clamp timeout to avoid unsigned-> signed overflow */
	if (timeout_jiffies > MAX_SCHEDULE_TIMEOUT)
		return MAX_SCHEDULE_TIMEOUT - 1;

	return timeout_jiffies;
}

int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
			      struct drm_file *filp)
{
	union drm_amdgpu_gem_wait_idle *args = data;
	struct drm_gem_object *gobj;
	struct amdgpu_bo *robj;
	uint32_t handle = args->in.handle;
	unsigned long timeout = amdgpu_gem_timeout(args->in.timeout);
	int r = 0;
	long ret;

	gobj = drm_gem_object_lookup(filp, handle);
	if (!gobj)
		return -ENOENT;

	robj = gem_to_amdgpu_bo(gobj);
	ret = dma_resv_wait_timeout(robj->tbo.base.resv, DMA_RESV_USAGE_READ,
				    true, timeout);

	/* ret == 0 means not signaled,
	 * ret > 0 means signaled
	 * ret < 0 means interrupted before timeout
	 */
	if (ret >= 0) {
		memset(args, 0, sizeof(*args));
		args->out.status = (ret == 0);
	} else
		r = ret;

	drm_gem_object_put(gobj);
	return r;
}

int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
				struct drm_file *filp)
{
	struct drm_amdgpu_gem_metadata *args = data;
	struct drm_gem_object *gobj;
	struct amdgpu_bo *robj;
	int r = -1;

	DRM_DEBUG("%d\n", args->handle);
	gobj = drm_gem_object_lookup(filp, args->handle);
	if (gobj == NULL)
		return -ENOENT;
	robj = gem_to_amdgpu_bo(gobj);

	r = amdgpu_bo_reserve(robj, false);
	if (unlikely(r != 0))
		goto out;

	/* Reject MMIO_REMAP BOs at IOCTL level: metadata/tiling does not apply. */
	if (robj->tbo.resource &&
	    robj->tbo.resource->mem_type == AMDGPU_PL_MMIO_REMAP) {
		DRM_WARN("metadata ioctl on MMIO_REMAP BO (handle %d)\n",
			 args->handle);
		r = -EINVAL;
		goto unreserve;
	}

	if (args->op == AMDGPU_GEM_METADATA_OP_GET_METADATA) {
		amdgpu_bo_get_tiling_flags(robj, &args->data.tiling_info);
		r = amdgpu_bo_get_metadata(robj, args->data.data,
					   sizeof(args->data.data),
					   &args->data.data_size_bytes,
					   &args->data.flags);
	} else if (args->op == AMDGPU_GEM_METADATA_OP_SET_METADATA) {
		if (args->data.data_size_bytes > sizeof(args->data.data)) {
			r = -EINVAL;
			goto unreserve;
		}
		r = amdgpu_bo_set_tiling_flags(robj, args->data.tiling_info);
		if (!r)
			r = amdgpu_bo_set_metadata(robj, args->data.data,
						   args->data.data_size_bytes,
						   args->data.flags);
	}

unreserve:
	amdgpu_bo_unreserve(robj);
out:
	drm_gem_object_put(gobj);
	return r;
}

/**
 * amdgpu_gem_va_update_vm -update the bo_va in its VM
 *
 * @adev: amdgpu_device pointer
 * @vm: vm to update
 * @bo_va: bo_va to update
 * @operation: map, unmap or clear
 *
 * Update the bo_va directly after setting its address. Errors are not
 * vital here, so they are not reported back to userspace.
 *
 * Returns resulting fence if freed BO(s) got cleared from the PT.
 * otherwise stub fence in case of error.
 */
static struct dma_fence *
amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
			struct amdgpu_vm *vm,
			struct amdgpu_bo_va *bo_va,
			uint32_t operation)
{
	struct dma_fence *fence;
	int r = 0;

	/* If the VM is not ready return only a stub. */
	if (!amdgpu_vm_ready(vm))
		return dma_fence_get_stub();


	/*
	 * First clean up any freed mappings in the VM.
	 *
	 * amdgpu_vm_clear_freed() may replace @fence with a new fence if it
	 * schedules GPU work. If nothing needs clearing, @fence can remain as
	 * the original vm->last_update.
	 */
	r = amdgpu_vm_clear_freed(adev, vm, &vm->last_update);
	if (r)
		goto error;

	/* For MAP/REPLACE we also need to update the BO mappings. */
	if (operation == AMDGPU_VA_OP_MAP ||
	    operation == AMDGPU_VA_OP_REPLACE) {
		r = amdgpu_vm_bo_update(adev, bo_va, false);
		if (r)
			goto error;
	}

	/* Always update PDEs after we touched the mappings. */
	r = amdgpu_vm_update_pdes(adev, vm, false);
	if (r)
		goto error;

	if ((operation == AMDGPU_VA_OP_MAP ||
	     operation == AMDGPU_VA_OP_REPLACE) &&
	    !amdgpu_vm_is_bo_always_valid(vm, bo_va->base.bo)) {

		/*
		 * For MAP/REPLACE of non per-VM BOs we need to sync to both the
		 * bo_va->last_pt_update and vm->last_update or otherwise we
		 * potentially miss the PDE updates.
		 */
		fence = dma_fence_unwrap_merge(vm->last_update,
					       bo_va->last_pt_update);
		if (!fence) {
			/* As fallback in OOM situations */
			dma_fence_wait(vm->last_update, false);
			dma_fence_wait(bo_va->last_pt_update, false);
			fence = dma_fence_get_stub();
		}
	} else {
		fence = dma_fence_get(vm->last_update);
	}

	return fence;

error:
	if (r && r != -ERESTARTSYS)
		DRM_ERROR("Couldn't update BO_VA (%d)\n", r);

	return dma_fence_get(vm->last_update);
}

int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
			  struct drm_file *filp)
{
	const uint32_t valid_flags = AMDGPU_VM_DELAY_UPDATE |
		AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_PAGE_WRITEABLE |
		AMDGPU_VM_PAGE_EXECUTABLE | AMDGPU_VM_MTYPE_MASK |
		AMDGPU_VM_PAGE_NOALLOC;
	const uint32_t prt_flags = AMDGPU_VM_DELAY_UPDATE |
		AMDGPU_VM_PAGE_PRT;

	struct drm_amdgpu_gem_va *args = data;
	struct drm_gem_object *gobj;
	struct amdgpu_device *adev = drm_to_adev(dev);
	struct amdgpu_fpriv *fpriv = filp->driver_priv;
	struct amdgpu_bo *abo;
	struct amdgpu_bo_va *bo_va;
	struct drm_syncobj *timeline_syncobj = NULL;
	struct dma_fence_chain *timeline_chain = NULL;
	struct drm_exec exec;
	uint64_t vm_size, tmp;
	int r = 0;

	/* Validate virtual address range against reserved regions. */
	if (args->va_address < AMDGPU_VA_RESERVED_BOTTOM) {
		dev_dbg(dev->dev,
			"va_address 0x%llx is in reserved area 0x%llx\n",
			args->va_address, AMDGPU_VA_RESERVED_BOTTOM);
		return -EINVAL;
	}

	if (args->va_address >= AMDGPU_GMC_HOLE_START &&
	    args->va_address < AMDGPU_GMC_HOLE_END) {
		dev_dbg(dev->dev,
			"va_address 0x%llx is in VA hole 0x%llx-0x%llx\n",
			args->va_address, AMDGPU_GMC_HOLE_START,
			AMDGPU_GMC_HOLE_END);
		return -EINVAL;
	}

	args->va_address &= AMDGPU_GMC_HOLE_MASK;

	vm_size = adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE;
	vm_size -= AMDGPU_VA_RESERVED_TOP;
	if (check_add_overflow(args->va_address, args->map_size, &tmp) || tmp > vm_size) {
		dev_dbg(dev->dev,
			"va_address 0x%llx is in top reserved area 0x%llx\n",
			args->va_address + args->map_size, vm_size);
		return -EINVAL;
	}

	if ((args->flags & ~valid_flags) && (args->flags & ~prt_flags)) {
		dev_dbg(dev->dev, "invalid flags combination 0x%08X\n",
			args->flags);
		return -EINVAL;
	}

	/* Validate operation type. */
	switch (args->operation) {
	case AMDGPU_VA_OP_MAP:
	case AMDGPU_VA_OP_UNMAP:
	case AMDGPU_VA_OP_CLEAR:
	case AMDGPU_VA_OP_REPLACE:
		break;
	default:
		dev_dbg(dev->dev, "unsupported operation %d\n",
			args->operation);
		return -EINVAL;
	}

	if (args->flags & AMDGPU_VM_DELAY_UPDATE &&
	    args->vm_timeline_syncobj_out)
		return -EINVAL;

	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
		gobj = drm_gem_object_lookup(filp, args->handle);
		if (gobj == NULL)
			return -ENOENT;
		abo = gem_to_amdgpu_bo(gobj);
	} else {
		gobj = NULL;
		abo = NULL;
	}

	/* Add input syncobj fences (if any) for synchronization. */
	r = amdgpu_gem_add_input_fence(filp,
				       args->input_fence_syncobj_handles,
				       args->num_syncobj_handles);
	if (r)
		goto error_put_gobj;

	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
		      DRM_EXEC_IGNORE_DUPLICATES, 0);
	drm_exec_until_all_locked(&exec) {
		if (gobj) {
			r = drm_exec_lock_obj(&exec, gobj);
			drm_exec_retry_on_contention(&exec);
			if (unlikely(r))
				goto error;
		}

		r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2);
		drm_exec_retry_on_contention(&exec);
		if (unlikely(r))
			goto error;
	}

	/* Resolve the BO-VA mapping for this VM/BO combination. */
	if (abo) {
		bo_va = amdgpu_vm_bo_find(&fpriv->vm, abo);
		if (!bo_va) {
			r = -ENOENT;
			goto error;
		}
	} else if (args->operation != AMDGPU_VA_OP_CLEAR) {
		bo_va = fpriv->prt_va;
	} else {
		bo_va = NULL;
	}

	/*
	 * Prepare the timeline syncobj node if the user requested a VM
	 * timeline update. This only allocates/looks up the syncobj and
	 * chain node; the actual fence is attached later.
	 */
	r = amdgpu_gem_update_timeline_node(filp,
					    args->vm_timeline_syncobj_out,
					    args->vm_timeline_point,
					    &timeline_syncobj,
					    &timeline_chain);
	if (r)
		goto error;

	switch (args->operation) {
	case AMDGPU_VA_OP_MAP:
		r = amdgpu_vm_bo_map(adev, bo_va, args->va_address,
				     args->offset_in_bo, args->map_size,
				     args->flags);
		break;
	case AMDGPU_VA_OP_UNMAP:
		r = amdgpu_vm_bo_unmap(adev, bo_va, args->va_address);
		break;

	case AMDGPU_VA_OP_CLEAR:
		r = amdgpu_vm_bo_clear_mappings(adev, &fpriv->vm,
						args->va_address,
						args->map_size);
		break;
	case AMDGPU_VA_OP_REPLACE:
		r = amdgpu_vm_bo_replace_map(adev, bo_va, args->va_address,
					     args->offset_in_bo, args->map_size,
					     args->flags);
		break;
	default:
		break;
	}

	/*
	 * Once the VA operation is done, update the VM and obtain the fence
	 * that represents the last relevant update for this mapping. This
	 * fence can then be exported to the user-visible VM timeline.
	 */
	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) &&
	    (!adev->debug_vm || timeline_syncobj)) {
		struct dma_fence *fence;

		fence = amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va,
						args->operation);
		if (timeline_syncobj) {
			if (!args->vm_timeline_point) {
				/* Replace the existing fence when no point is given. */
				drm_syncobj_replace_fence(timeline_syncobj,
							  fence);
			} else {
				/* Attach the last-update fence at a specific point. */
				drm_syncobj_add_point(timeline_syncobj,
						      timeline_chain,
						      fence,
						      args->vm_timeline_point);
				timeline_chain = NULL;
			}
		}
		dma_fence_put(fence);

	}

error:
	dma_fence_chain_free(timeline_chain);
	if (timeline_syncobj)
		drm_syncobj_put(timeline_syncobj);
	drm_exec_fini(&exec);
error_put_gobj:
	drm_gem_object_put(gobj);
	return r;
}

int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
			struct drm_file *filp)
{
	struct drm_amdgpu_gem_op *args = data;
	struct drm_gem_object *gobj;
	struct amdgpu_vm_bo_base *base;
	struct amdgpu_bo *robj;
	struct drm_exec exec;
	struct amdgpu_fpriv *fpriv = filp->driver_priv;
	int r;

	if (args->padding)
		return -EINVAL;

	gobj = drm_gem_object_lookup(filp, args->handle);
	if (!gobj)
		return -ENOENT;

	robj = gem_to_amdgpu_bo(gobj);

	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
			  DRM_EXEC_IGNORE_DUPLICATES, 0);
	drm_exec_until_all_locked(&exec) {
		r = drm_exec_lock_obj(&exec, gobj);
		drm_exec_retry_on_contention(&exec);
		if (r)
			goto out_exec;

		if (args->op == AMDGPU_GEM_OP_GET_MAPPING_INFO) {
			r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 0);
			drm_exec_retry_on_contention(&exec);
			if (r)
				goto out_exec;
		}
	}

	switch (args->op) {
	case AMDGPU_GEM_OP_GET_GEM_CREATE_INFO: {
		struct drm_amdgpu_gem_create_in info;
		void __user *out = u64_to_user_ptr(args->value);

		info.bo_size = robj->tbo.base.size;
		info.alignment = robj->tbo.page_alignment << PAGE_SHIFT;
		info.domains = robj->preferred_domains;
		info.domain_flags = robj->flags;
		drm_exec_fini(&exec);
		if (copy_to_user(out, &info, sizeof(info)))
			r = -EFAULT;
		break;
	}
	case AMDGPU_GEM_OP_SET_PLACEMENT:
		if (drm_gem_is_imported(&robj->tbo.base) &&
		    args->value & AMDGPU_GEM_DOMAIN_VRAM) {
			r = -EINVAL;
			goto out_exec;
		}
		if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm)) {
			r = -EPERM;
			goto out_exec;
		}
		for (base = robj->vm_bo; base; base = base->next)
			if (amdgpu_xgmi_same_hive(amdgpu_ttm_adev(robj->tbo.bdev),
				amdgpu_ttm_adev(base->vm->root.bo->tbo.bdev))) {
				r = -EINVAL;
				goto out_exec;
			}


		robj->preferred_domains = args->value & (AMDGPU_GEM_DOMAIN_VRAM |
							AMDGPU_GEM_DOMAIN_GTT |
							AMDGPU_GEM_DOMAIN_CPU);
		robj->allowed_domains = robj->preferred_domains;
		if (robj->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)
			robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;

		if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
			amdgpu_vm_bo_invalidate(robj, true);
		drm_exec_fini(&exec);
		break;
	case AMDGPU_GEM_OP_GET_MAPPING_INFO: {
		struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(&fpriv->vm, robj);
		struct drm_amdgpu_gem_vm_entry *vm_entries;
		struct amdgpu_bo_va_mapping *mapping;
		int num_mappings = 0;
		/*
		 * num_entries is set as an input to the size of the user-allocated array of
		 * drm_amdgpu_gem_vm_entry stored at args->value.
		 * num_entries is sent back as output as the number of mappings the bo has.
		 * If that number is larger than the size of the array, the ioctl must
		 * be retried.
		 */
		if (!bo_va) {
			r = -ENOENT;
			goto out_exec;
		}

		if (args->num_entries > INT_MAX / sizeof(*vm_entries)) {
			r = -EINVAL;
			goto out_exec;
		}

		vm_entries = kvcalloc(args->num_entries, sizeof(*vm_entries), GFP_KERNEL);
		if (!vm_entries) {
			r = -ENOMEM;
			goto out_exec;
		}

		amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) {
			if (num_mappings < args->num_entries) {
				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
				vm_entries[num_mappings].offset = mapping->offset;
				vm_entries[num_mappings].flags = mapping->flags;
			}
			num_mappings += 1;
		}

		amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) {
			if (num_mappings < args->num_entries) {
				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
				vm_entries[num_mappings].offset = mapping->offset;
				vm_entries[num_mappings].flags = mapping->flags;
			}
			num_mappings += 1;
		}

		drm_exec_fini(&exec);

		if (num_mappings > 0 && num_mappings <= args->num_entries)
			if (copy_to_user(u64_to_user_ptr(args->value), vm_entries, num_mappings * sizeof(*vm_entries)))
				r = -EFAULT;

		args->num_entries = num_mappings;

		kvfree(vm_entries);
		break;
	}
	default:
		drm_exec_fini(&exec);
		r = -EINVAL;
	}

	drm_gem_object_put(gobj);
	return r;
out_exec:
	drm_exec_fini(&exec);
	drm_gem_object_put(gobj);
	return r;
}

/**
 * amdgpu_gem_list_handles_ioctl - get information about a process' buffer objects
 *
 * @dev: drm device pointer
 * @data: drm_amdgpu_gem_list_handles
 * @filp: drm file pointer
 *
 * num_entries is set as an input to the size of the entries array.
 * num_entries is sent back as output as the number of bos in the process.
 * If that number is larger than the size of the array, the ioctl must
 * be retried.
 *
 * Returns:
 * 0 for success, -errno for errors.
 */
int amdgpu_gem_list_handles_ioctl(struct drm_device *dev, void *data,
				  struct drm_file *filp)
{
	struct drm_amdgpu_gem_list_handles *args = data;
	struct drm_amdgpu_gem_list_handles_entry *bo_entries;
	struct drm_gem_object *gobj;
	int id, ret = 0;
	int bo_index = 0;
	int num_bos = 0;

	spin_lock(&filp->table_lock);
	idr_for_each_entry(&filp->object_idr, gobj, id)
		num_bos += 1;
	spin_unlock(&filp->table_lock);

	if (args->num_entries < num_bos) {
		args->num_entries = num_bos;
		return 0;
	}

	if (num_bos == 0) {
		args->num_entries = 0;
		return 0;
	}

	bo_entries = kvzalloc_objs(*bo_entries, num_bos);
	if (!bo_entries)
		return -ENOMEM;

	spin_lock(&filp->table_lock);
	idr_for_each_entry(&filp->object_idr, gobj, id) {
		struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
		struct drm_amdgpu_gem_list_handles_entry *bo_entry;

		if (bo_index >= num_bos) {
			ret = -EAGAIN;
			break;
		}

		bo_entry = &bo_entries[bo_index];

		bo_entry->size = amdgpu_bo_size(bo);
		bo_entry->alloc_flags = bo->flags & AMDGPU_GEM_CREATE_SETTABLE_MASK;
		bo_entry->preferred_domains = bo->preferred_domains;
		bo_entry->gem_handle = id;
		bo_entry->alignment = bo->tbo.page_alignment;

		if (bo->tbo.base.import_attach)
			bo_entry->flags |= AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;

		bo_index += 1;
	}
	spin_unlock(&filp->table_lock);

	args->num_entries = bo_index;

	if (!ret)
		if (copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)))
			ret = -EFAULT;

	kvfree(bo_entries);

	return ret;
}

static unsigned int amdgpu_gem_align_pitch(struct amdgpu_device *adev,
					   unsigned int width,
					   unsigned int cpp,
					   bool tiled)
{
	unsigned int aligned = width;
	unsigned int pitch_mask = 0;
	unsigned int pitch;

	switch (cpp) {
	case 1:
		pitch_mask = 255;
		break;
	case 2:
		pitch_mask = 127;
		break;
	case 3:
	case 4:
		pitch_mask = 63;
		break;
	}

	if (check_add_overflow(aligned, pitch_mask, &aligned))
		return 0;
	aligned &= ~pitch_mask;
	if (check_mul_overflow(aligned, cpp, &pitch))
		return 0;
	return pitch;
}

int amdgpu_mode_dumb_create(struct drm_file *file_priv,
			    struct drm_device *dev,
			    struct drm_mode_create_dumb *args)
{
	struct amdgpu_device *adev = drm_to_adev(dev);
	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
	struct drm_gem_object *gobj;
	uint32_t handle;
	u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
		    AMDGPU_GEM_CREATE_CPU_GTT_USWC |
		    AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
	u32 domain;
	int r;

	/*
	 * The buffer returned from this function should be cleared, but
	 * it can only be done if the ring is enabled or we'll fail to
	 * create the buffer.
	 */
	if (adev->mman.buffer_funcs_enabled)
		flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;

	args->pitch = amdgpu_gem_align_pitch(adev, args->width,
					     DIV_ROUND_UP(args->bpp, 8), 0);
	if (!args->pitch)
		return -EINVAL;
	args->size = (u64)args->pitch * args->height;
	args->size = ALIGN(args->size, PAGE_SIZE);
	if (!args->size)
		return -EINVAL;
	domain = amdgpu_bo_get_preferred_domain(adev,
				amdgpu_display_supported_domains(adev, flags));
	r = amdgpu_gem_object_create(adev, args->size, 0, domain, flags,
				     ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
	if (r)
		return -ENOMEM;

	r = drm_gem_handle_create(file_priv, gobj, &handle);
	/* drop reference from allocate - handle holds it now */
	drm_gem_object_put(gobj);
	if (r)
		return r;

	args->handle = handle;
	return 0;
}

#if defined(CONFIG_DEBUG_FS)
static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
{
	struct amdgpu_device *adev = m->private;
	struct drm_device *dev = adev_to_drm(adev);
	struct drm_file *file;
	int r;

	r = mutex_lock_interruptible(&dev->filelist_mutex);
	if (r)
		return r;

	list_for_each_entry(file, &dev->filelist, lhead) {
		struct task_struct *task;
		struct drm_gem_object *gobj;
		struct pid *pid;
		int id;

		/*
		 * Although we have a valid reference on file->pid, that does
		 * not guarantee that the task_struct who called get_pid() is
		 * still alive (e.g. get_pid(current) => fork() => exit()).
		 * Therefore, we need to protect this ->comm access using RCU.
		 */
		rcu_read_lock();
		pid = rcu_dereference(file->pid);
		task = pid_task(pid, PIDTYPE_TGID);
		seq_printf(m, "pid %8d command %s:\n", pid_nr(pid),
			   task ? task->comm : "<unknown>");
		rcu_read_unlock();

		spin_lock(&file->table_lock);
		idr_for_each_entry(&file->object_idr, gobj, id) {
			struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);

			amdgpu_bo_print_info(id, bo, m);
		}
		spin_unlock(&file->table_lock);
	}

	mutex_unlock(&dev->filelist_mutex);
	return 0;
}

DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_gem_info);

#endif

void amdgpu_debugfs_gem_init(struct amdgpu_device *adev)
{
#if defined(CONFIG_DEBUG_FS)
	struct drm_minor *minor = adev_to_drm(adev)->primary;
	struct dentry *root = minor->debugfs_root;

	debugfs_create_file("amdgpu_gem_info", 0444, root, adev,
			    &amdgpu_debugfs_gem_info_fops);
#endif
}