On Wed, Apr 3, 2019 at 2:30 AM Pan, Xinhui <Xinhui.Pan@xxxxxxx> wrote:
>
> Unless IP has implemented its own ras, use ERROR_NONE as the default
> type.
>
> Signed-off-by: xinhui pan <xinhui.pan@xxxxxxx>
Reviewed-by: Alex Deucher <alexander.deucher@xxxxxxx>
In a future patch, I think we should probably just drop the default
enable and just explicitly enable each RAS feature. This catchall one
at the end is confusing and makes the logic hard to follow in the
common RAS code. For these other IPs that don't have explicit setup,
maybe just explicitly enable them in the soc IP modules (e.g.,
soc15.c).
Alex
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 24 +++++++++++++++---------
> 1 file changed, 15 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index fc4bf7237d4b..655d58b63405 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -685,11 +685,13 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev,
> struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
> int ras_block_count = AMDGPU_RAS_BLOCK_COUNT;
> int i;
> + const enum amdgpu_ras_error_type default_ras_type =
> + AMDGPU_RAS_ERROR__NONE;
>
> for (i = 0; i < ras_block_count; i++) {
> struct ras_common_if head = {
> .block = i,
> - .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
> + .type = default_ras_type,
> .sub_block_index = 0,
> };
> strcpy(head.name, ras_block_str(i));
> @@ -1495,9 +1497,6 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
>
> amdgpu_ras_mask &= AMDGPU_RAS_BLOCK_MASK;
>
> - if (con->flags & AMDGPU_RAS_FLAG_INIT_BY_VBIOS)
> - amdgpu_ras_enable_all_features(adev, 1);
> -
> if (amdgpu_ras_fs_init(adev))
> goto fs_out;
>
> @@ -1525,18 +1524,25 @@ void amdgpu_ras_post_init(struct amdgpu_device *adev)
> if (!con)
> return;
>
> - /* We enable ras on all hw_supported block, but as boot parameter might
> - * disable some of them and one or more IP has not implemented yet.
> - * So we disable them on behalf.
> - */
> if (con->flags & AMDGPU_RAS_FLAG_INIT_BY_VBIOS) {
> + /* Set up all other IPs which are not implemented. There is a
> + * tricky thing that IP's actual ras error type should be
> + * MULTI_UNCORRECTABLE, but as driver does not handle it, so
> + * ERROR_NONE make sense anyway.
> + */
> + amdgpu_ras_enable_all_features(adev, 1);
> +
> + /* We enable ras on all hw_supported block, but as boot
> + * parameter might disable some of them and one or more IP has
> + * not implemented yet. So we disable them on behalf.
> + */
> list_for_each_entry_safe(obj, tmp, &con->head, node) {
> if (!amdgpu_ras_is_supported(adev, obj->head.block)) {
> amdgpu_ras_feature_enable(adev, &obj->head, 0);
> /* there should be no any reference. */
> WARN_ON(alive_obj(obj));
> }
> - };
> + }
> }
> }
>
> --
> 2.17.1
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@xxxxxxxxxxxxxxxxxxxxx
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
