[PATCH] drm/amdgpu: Properly allocate VM invalidate eng v2

[Christian König]

Am 14.11.2017 um 16:22 schrieb Oak Zeng:
> From: ozeng <oak.zeng at amd.com>
>
> v1: Properly allocate TLB invalidation engine to avoid conflict.
> v2: Added comments to codes
>
> Change-Id: I9a32cbcb133c07d7efad1a7bd07c6e2098c747e7
> Signed-off-by: Oak Zeng <Oak.Zeng at amd.com>

Reviewed-by: Christian König <christian.koenig at amd.com>

> ---
>   drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 15 ++++++++++++---
>   1 file changed, 12 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> index 16bba97..b067b46 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> @@ -392,7 +392,16 @@ static int gmc_v9_0_early_init(void *handle)
>   static int gmc_v9_0_late_init(void *handle)
>   {
>   	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
> -	unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 3, 3 };
> +	/*
> +	 * The latest engine allocation on gfx9 is:
> +	 * Engine 0, 1: idle
> +	 * Engine 2, 3: firmware
> +	 * Engine 4~13: amdgpu ring, subject to change when ring number changes
> +	 * Engine 14~15: idle
> +	 * Engine 16: kfd tlb invalidation
> +	 * Engine 17: Gart flushes
> +	 */
> +	unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 4, 4 };
>   	unsigned i;
>   
>   	for(i = 0; i < adev->num_rings; ++i) {
> @@ -405,9 +414,9 @@ static int gmc_v9_0_late_init(void *handle)
>   			 ring->funcs->vmhub);
>   	}
>   
> -	/* Engine 17 is used for GART flushes */
> +	/* Engine 16 is used for KFD and 17 for GART flushes */
>   	for(i = 0; i < AMDGPU_MAX_VMHUBS; ++i)
> -		BUG_ON(vm_inv_eng[i] > 17);
> +		BUG_ON(vm_inv_eng[i] > 16);
>   
>   	return amdgpu_irq_get(adev, &adev->mc.vm_fault, 0);
>   }