| Message ID | 20250222033221.554976-1-dhananjay.ugwekar@amd.com |
|---|---|
| State | New |
| Headers | show |
| Series | cpufreq/amd-pstate: Fix the clamping of perf values | expand |
On 2/21/2025 21:32, Dhananjay Ugwekar wrote: > The clamping in freq_to_perf() is broken right now, as we first typecast > (read wraparound) the overflowing value into a u8 and then clamp it down. > So, use a u32 to store the >255 value in certain edge cases and then clamp > it down into a u8. > > Also, use a "explicit typecast + clamp" instead of just a "clamp_t" as the > latter typecasts first and then clamps between the limits, which defeats > our purpose. > > Fixes: 305621eb6a8b ("cpufreq/amd-pstate: Modularize perf<->freq conversion") > Signed-off-by: Dhananjay Ugwekar <dhananjay.ugwekar@amd.com> Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> I'll rebase my series on this for v5. > --- > drivers/cpufreq/amd-pstate.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c > index 9ab95ec1f828..4705a644db6d 100644 > --- a/drivers/cpufreq/amd-pstate.c > +++ b/drivers/cpufreq/amd-pstate.c > @@ -144,10 +144,10 @@ static struct quirk_entry quirk_amd_7k62 = { > > static inline u8 freq_to_perf(struct amd_cpudata *cpudata, unsigned int freq_val) > { > - u8 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * cpudata->nominal_perf, > + u32 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * cpudata->nominal_perf, > cpudata->nominal_freq); > > - return clamp_t(u8, perf_val, cpudata->lowest_perf, cpudata->highest_perf); > + return (u8)clamp(perf_val, cpudata->lowest_perf, cpudata->highest_perf); > } > > static inline u32 perf_to_freq(struct amd_cpudata *cpudata, u8 perf_val)
diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c index 9ab95ec1f828..4705a644db6d 100644 --- a/drivers/cpufreq/amd-pstate.c +++ b/drivers/cpufreq/amd-pstate.c @@ -144,10 +144,10 @@ static struct quirk_entry quirk_amd_7k62 = { static inline u8 freq_to_perf(struct amd_cpudata *cpudata, unsigned int freq_val) { - u8 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * cpudata->nominal_perf, + u32 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * cpudata->nominal_perf, cpudata->nominal_freq); - return clamp_t(u8, perf_val, cpudata->lowest_perf, cpudata->highest_perf); + return (u8)clamp(perf_val, cpudata->lowest_perf, cpudata->highest_perf); } static inline u32 perf_to_freq(struct amd_cpudata *cpudata, u8 perf_val)
The clamping in freq_to_perf() is broken right now, as we first typecast (read wraparound) the overflowing value into a u8 and then clamp it down. So, use a u32 to store the >255 value in certain edge cases and then clamp it down into a u8. Also, use a "explicit typecast + clamp" instead of just a "clamp_t" as the latter typecasts first and then clamps between the limits, which defeats our purpose. Fixes: 305621eb6a8b ("cpufreq/amd-pstate: Modularize perf<->freq conversion") Signed-off-by: Dhananjay Ugwekar <dhananjay.ugwekar@amd.com> --- drivers/cpufreq/amd-pstate.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)