On Tue, Mar 11, 2025 at 5:03 AM Kenneth Feng <kenneth.feng@xxxxxxx> wrote:
>
> This is a bug fix. The scenario is that the same client can add the certain
> workload type refcount multiple times. Then the same client can not remove
> this setting when it wants to get back to the default bootup workload.
This will break the ref count handling.
Alex
>
> Signed-off-by: Kenneth Feng <kenneth.feng@xxxxxxx>
> ---
> drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
> index 8cfb07549f54..592bf54015c6 100644
> --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
> +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c
> @@ -2273,7 +2273,8 @@ static int smu_bump_power_profile_mode(struct smu_context *smu,
> static void smu_power_profile_mode_get(struct smu_context *smu,
> enum PP_SMC_POWER_PROFILE profile_mode)
> {
> - smu->workload_refcount[profile_mode]++;
> + if (!smu->workload_refcount[profile_mode])
> + smu->workload_refcount[profile_mode]++;
> }
>
> static void smu_power_profile_mode_put(struct smu_context *smu,
> --
> 2.34.1
>
