[PATCH] drm/amdgpu: Use SKU instead of DID for FRU check

[Felix Kuehling]

Am 2020-09-29 um 7:31 a.m. schrieb Kent Russell:
> The VG20 DIDs 66a0, 66a1 and 66a4 are used for various SKUs that may or may
> not have the FRU EEPROM on it. Parse the VBIOS to check for server SKU
> variants (D131 or D134) until a more general solution can be determined.
>
> Signed-off-by: Kent Russell <kent.russell at amd.com>
> ---
>  .../gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c    | 37 +++++++++++++------
>  1 file changed, 26 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
> index e811fecc540f..82207b758800 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
> @@ -34,18 +34,33 @@
>  
>  static bool is_fru_eeprom_supported(struct amdgpu_device *adev)
>  {
> -	/* TODO: Gaming SKUs don't have the FRU EEPROM.
> -	 * Use this hack to address hangs on modprobe on gaming SKUs
> -	 * until a proper solution can be implemented by only supporting
> -	 * the explicit chip IDs for VG20 Server cards
> -	 *
> -	 * TODO: Add list of supported Arcturus DIDs once confirmed
> +	/* TODO: See if we can figure this out dynamically instead of
> +	 * having to parse VBIOS versions.
>  	 */
> -	if ((adev->asic_type == CHIP_VEGA20 && adev->pdev->device == 0x66a0) ||
> -	    (adev->asic_type == CHIP_VEGA20 && adev->pdev->device == 0x66a1) ||
> -	    (adev->asic_type == CHIP_VEGA20 && adev->pdev->device == 0x66a4))
> -		return true;
> -	return false;
> +	struct atom_context *atom_ctx = adev->mode_info.atom_context;
> +	char model[3];
> +	int modelnum;
> +
> +	/* VBIOS is of the format ###-DXXXXYY-XX. We only want the first

I'm not sure I understand the format here. There are four X, but you
only want the first three of them?

What are the two X in the end? are they related to the first four X?


> +	 * 3 digits after the D, and if we convert it to a hex integer,
> +	 * we can use switch for ease/speed/readability and potential
> +	 * expandability if required
> +	 */
> +	strncpy(model, atom_ctx->vbios_version + 5, 3);

This produces a string in _model_ that's not 0-terminated.


> +	modelnum = strtoul(model, NULL, 16);

Using strtoul on a non-terminated string is not reliable.

Regards,
  Felix


> +	switch (adev->asic_type) {
> +	case CHIP_VEGA20:
> +		switch (modelnum) {
> +		/* These are the server VG20 SKUs */
> +		case 0x161:
> +		case 0x163:
> +			return true;
> +		default:
> +			return false;
> +		}
> +	default:
> +		return false;
> +	}
>  }
>  
>  static int amdgpu_fru_read_eeprom(struct amdgpu_device *adev, uint32_t addrptr,