Re: [PATCH] drm/amdkfd: fix svm_migrate_fini warning

[Felix Kuehling <felix.kuehling@amd.com>]

Am 2021-09-20 um 5:55 p.m. schrieb Philip Yang:
> Don't use devm_request_free_mem_region to alloc VRAM region for pgmap,
devm_... refers to a device manager that automatically releases
device-specific resources when a driver disconnects from a device. So
maybe that just means, our devm_memunmap_pages and
devm_release_mem_region calls in svm_migrate_fini are redundant, and the
best solution is to remove those calls.

See Documentation/driver-api/driver-model/devres.rst

Regards,
  Felix


> because driver__detach releases all device resource region, then calls
> amdgpu_device_fini_sw, which calls devm_memunmap_pages generating below
> warning trace:
>
> WARNING: CPU: 1 PID: 3646 at drivers/base/devres.c:795
> devm_release_action+0x51/0x60
> Call Trace:
>     ? memunmap_pages+0x360/0x360
>     svm_migrate_fini+0x2d/0x60 [amdgpu]
>     kgd2kfd_device_exit+0x23/0xa0 [amdgpu]
>     amdgpu_amdkfd_device_fini_sw+0x1d/0x30 [amdgpu]
>     amdgpu_device_fini_sw+0x45/0x290 [amdgpu]
>     amdgpu_driver_release_kms+0x12/0x30 [amdgpu]
>     drm_dev_release+0x20/0x40 [drm]
>     release_nodes+0x196/0x1e0
>     device_release_driver_internal+0x104/0x1d0
>     driver_detach+0x47/0x90
>     bus_remove_driver+0x7a/0xd0
>     pci_unregister_driver+0x3d/0x90
>     amdgpu_exit+0x11/0x20 [amdgpu]
> Trying to free nonexistent resource <000007fc00000000-000007fdffffffff>
>
> Signed-off-by: Philip Yang <Philip.Yang@amd.com>
> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> index 21f745e0b86c..aa96767920a9 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> @@ -880,7 +880,7 @@ int svm_migrate_init(struct amdgpu_device *adev)
>  	 * should remove reserved size
>  	 */
>  	size = ALIGN(adev->gmc.real_vram_size, 2ULL << 20);
> -	res = devm_request_free_mem_region(adev->dev, &iomem_resource, size);
> +	res = request_free_mem_region(&iomem_resource, size, "amdgpu_vram");
>  	if (IS_ERR(res))
>  		return -ENOMEM;
>  
> @@ -891,14 +891,13 @@ int svm_migrate_init(struct amdgpu_device *adev)
>  	pgmap->ops = &svm_migrate_pgmap_ops;
>  	pgmap->owner = SVM_ADEV_PGMAP_OWNER(adev);
>  	pgmap->flags = MIGRATE_VMA_SELECT_DEVICE_PRIVATE;
> -	r = devm_memremap_pages(adev->dev, pgmap);
> +	r = memremap_pages(pgmap, dev_to_node(adev->dev));
>  	if (IS_ERR(r)) {
>  		pr_err("failed to register HMM device memory\n");
>  
>  		/* Disable SVM support capability */
>  		pgmap->type = 0;
> -		devm_release_mem_region(adev->dev, res->start,
> -					res->end - res->start + 1);
> +		release_mem_region(res->start, res->end - res->start + 1);
>  		return PTR_ERR(r);
>  	}
>  
> @@ -919,7 +918,7 @@ void svm_migrate_fini(struct amdgpu_device *adev)
>  	if (!KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev))
>  		return;
>  
> -	devm_memunmap_pages(adev->dev, pgmap);
> -	devm_release_mem_region(adev->dev, pgmap->range.start,
> -				pgmap->range.end - pgmap->range.start + 1);
> +	memunmap_pages(pgmap);
> +	release_mem_region(pgmap->range.start,
> +			   pgmap->range.end - pgmap->range.start + 1);
>  }