On 010/22, Timur Kristóf wrote:
> After refactoring the DSC capability calculation, the
> get_min_slice_count_for_odm could crash on some GPUs due to a
> division by zero when max_total_throughput_mps was zero.
> As a result, DC was broken when connecting a GPU that doesn't
> support DSC to a monitor that supports DSC.
> Tested on Oland (DCE 6) and Fiji (DCE 10).
>
> This commit fixes it by returning zero instead.
>
> Fixes: 4909b8b3846c ("drm/amd/display: Refactor DSC cap calculations")
> Signed-off-by: Timur Kristóf <timur.kris...@gmail.com>
> ---
> drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c | 17 +++++++++--------
> 1 file changed, 9 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c
> b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c
> index a454d16e6586..4169ece9c535 100644
> --- a/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c
> +++ b/drivers/gpu/drm/amd/display/dc/dsc/dc_dsc.c
> @@ -642,14 +642,15 @@ static unsigned int get_min_slice_count_for_odm(
> unsigned int max_dispclk_khz;
>
> /* get max pixel rate and combine caps */
> - max_dispclk_khz = dsc_enc_caps->max_total_throughput_mps * 1000;
> - if (dsc && dsc->ctx->dc) {
> - if (dsc->ctx->dc->clk_mgr &&
> - dsc->ctx->dc->clk_mgr->funcs->get_max_clock_khz) {
> - /* dispclk is available */
> - max_dispclk_khz =
> dsc->ctx->dc->clk_mgr->funcs->get_max_clock_khz(dsc->ctx->dc->clk_mgr,
> CLK_TYPE_DISPCLK);
> - }
> - }
> + if (dsc && dsc->ctx->dc && dsc->ctx->dc->clk_mgr &&
> + dsc->ctx->dc->clk_mgr->funcs->get_max_clock_khz)
> + max_dispclk_khz =
> + dsc->ctx->dc->clk_mgr->funcs->get_max_clock_khz(
> + dsc->ctx->dc->clk_mgr, CLK_TYPE_DISPCLK);
> + else if (dsc_enc_caps->max_total_throughput_mps)
> + max_dispclk_khz = dsc_enc_caps->max_total_throughput_mps * 1000;
> + else
> + return 0;
>
> /* consider minimum odm slices required due to
> * 1) display pipe throughput (dispclk)
> --
> 2.50.1
>
This patch lgtm.
Reviewed-by: Rodrigo Siqueira <sique...@igalia.com>
I added other display folks to this patch.
I'm not sure if the original modification (4909b8b3846c) is already in
the stable kernel; if so, it could be a good idea to send this fix to
the stable kernel as well.
Thanks
--
Rodrigo Siqueira
