Re: [PATCH] drm/amdgpu: Delay unmapping MMIO VRAM to amdgpu_ttm_fini() in GPU initialization failure

[Andrey Grodzovsky <andrey.grodzovsky@xxxxxxx>]

On 2022-01-03 9:30 p.m., Leslie Shi wrote:
> If the driver loads failed during hw_init(), delay unmapping MMIO VRAM to amdgpu_ttm_fini().
> Its prevents accessing invalid memory address in vcn_v3_0_sw_fini().
>
> Signed-off-by: Leslie Shi <Yuliang.Shi@xxxxxxx>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 16 +++++++++-------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  4 ++++
>   2 files changed, 13 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index ce93a304292c..d6006de57af5 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -3830,7 +3830,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>   	return r;
>   }
>   
> -static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
> +static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev, bool unmap_mmio_vram)
>   {
>   	/* Clear all CPU mappings pointing to this device */
>   	unmap_mapping_range(adev->ddev.anon_inode->i_mapping, 0, 0, 1);
> @@ -3840,9 +3840,12 @@ static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
>   
>   	iounmap(adev->rmmio);
>   	adev->rmmio = NULL;
> -	if (adev->mman.aper_base_kaddr)
> -		iounmap(adev->mman.aper_base_kaddr);
> -	adev->mman.aper_base_kaddr = NULL;


Why only VRAM ? Why not register BAR above ? In general I don't see why 
not just follow
what i suggested here https://www.spinics.net/lists/amd-gfx/msg72217.html

Andrey


> +
> +	if (unmap_mmio_vram) {
> +		if (adev->mman.aper_base_kaddr)
> +			iounmap(adev->mman.aper_base_kaddr);
> +		adev->mman.aper_base_kaddr = NULL;
> +	}
>   
>   	/* Memory manager related */
>   	if (!adev->gmc.xgmi.connected_to_cpu) {
> @@ -3905,8 +3908,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
>   
>   	amdgpu_gart_dummy_page_fini(adev);
>   
> -	if (drm_dev_is_unplugged(adev_to_drm(adev)))
> -		amdgpu_device_unmap_mmio(adev);
> +	amdgpu_device_unmap_mmio(adev, drm_dev_is_unplugged(adev_to_drm(adev)));
>   
>   }
>   
> @@ -5727,7 +5729,7 @@ void amdgpu_device_halt(struct amdgpu_device *adev)
>   
>   	adev->no_hw_access = true;
>   
> -	amdgpu_device_unmap_mmio(adev);
> +	amdgpu_device_unmap_mmio(adev, true);
>   
>   	pci_disable_device(pdev);
>   	pci_wait_for_pending_transaction(pdev);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 367abed1d6e6..67cd12caf019 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -1815,6 +1815,10 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
>   				      NULL, NULL);
>   	amdgpu_ttm_fw_reserve_vram_fini(adev);
>   
> +        if (adev->mman.aper_base_kaddr)
> +                iounmap(adev->mman.aper_base_kaddr);
> +        adev->mman.aper_base_kaddr = NULL;
> +
>   	amdgpu_vram_mgr_fini(adev);
>   	amdgpu_gtt_mgr_fini(adev);
>   	amdgpu_preempt_mgr_fini(adev);