Re: [PATCH v4 0/2] cpufreq: spacemit: Add cpufreq support for K1 SoC

On 6/26/2026 4:10 PM, Shuwei Wu wrote:
> This series enables CPU DVFS for the SpacemiT K1 SoC using the generic
> cpufreq-dt driver.
> 
> K1 has two CPU clock clusters. The two clusters have separate CPU clocks,
> so they are represented as two cpufreq policies: policy0 for CPUs 0-3 and
> policy4 for CPUs 4-7.
> 
> The CPU voltage rail is shared between the clusters. To model this with two
> policies, the OPP entries describe voltage ranges instead of a single fixed
> voltage, so the shared regulator can keep the rail within a range acceptable
> for the active OPP constraints.
> 
> Tested on Banana Pi BPI-F3:
> 
> ~ # cat /sys/devices/system/cpu/online
> 0-7
> 
> ~ # ls /sys/devices/system/cpu/cpufreq/
> policy0  policy4
> 
> ~ # cat /sys/devices/system/cpu/cpufreq/policy0/scaling_driver
> cpufreq-dt
> ~ # cat /sys/devices/system/cpu/cpufreq/policy0/affected_cpus
> 0 1 2 3
> 
> ~ # cat /sys/devices/system/cpu/cpufreq/policy4/scaling_driver
> cpufreq-dt
> ~ # cat /sys/devices/system/cpu/cpufreq/policy4/affected_cpus
> 4 5 6 7
> 
> Both policies expose the same OPP frequencies:
> 
> ~ # cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
> 614400 819000 1000000 1228800 1600000
> 
> ~ # cat /sys/devices/system/cpu/cpufreq/policy4/scaling_available_frequencies
> 614400 819000 1000000 1228800 1600000
> 
> For each policy, scaling_setspeed was set to each supported OPP and the
> workload was pinned to one CPU covered by that policy with taskset.
> CPU0 was used for policy0, and CPU4 was used for policy4. The clock rates below
> are from /sys/kernel/debug/clk/clk_summary.
> 
> policy0 / CPU0:
> ----------------------------------------------------------
> Frequency    |  cpu_c0_core_clk  |  Real (s)  |  User (s)
> (kHz)        |  (Hz)             |            |
> -------------+-------------------+------------+-----------
> 1,600,000    |  1,600,000,000    |    1.81    |    1.80
> 1,228,800    |  1,228,800,000    |    2.37    |    2.37
> 1,000,000    |  1,000,000,000    |    2.89    |    2.89
>    819,000    |    819,200,000    |    3.56    |    3.55
>    614,400    |    614,400,000    |    4.71    |    4.71
> ----------------------------------------------------------
> 
> policy4 / CPU4:
> ----------------------------------------------------------
> Frequency    |  cpu_c1_core_clk  |  Real (s)  |  User (s)
> (kHz)        |  (Hz)             |            |
> -------------+-------------------+------------+-----------
> 1,600,000    |  1,600,000,000    |    1.81    |    1.80
> 1,228,800    |  1,228,800,000    |    2.36    |    2.36
> 1,000,000    |  1,000,000,000    |    2.89    |    2.89
>    819,000    |    819,200,000    |    3.55    |    3.55
>    614,400    |    614,400,000    |    4.71    |    4.70
> ----------------------------------------------------------
> 
> Signed-off-by: Shuwei Wu <[email protected]>

Hi Shuwei,

Thanks for your work.

I have tested this series on the OrangePi RV2 4GB board on top of
next-20260626, with Vincent's patch for OrangePi RV2 applied.

# uname -a
Linux orangepi-rv2 7.1.0-next-20260626-00005-g66ef697be46d #1 SMP PREEMPT Sat Jun 27 14:53:02 CST 2026 riscv64 GNU/Linux

# ls /sys/devices/system/cpu/cpufreq/
boost    policy0  policy4

# cat /sys/devices/system/cpu/cpufreq/policy*/scaling_driver
cpufreq-dt
cpufreq-dt

# cat /sys/devices/system/cpu/cpufreq/policy0/affected_cpus
0 1 2 3

# cat /sys/devices/system/cpu/cpufreq/policy4/affected_cpus
4 5 6 7

# cat /sys/devices/system/cpu/cpufreq/policy*/scaling_available_frequencies
614400 819000 1000000 1228800 1600000
614400 819000 1000000 1228800 1600000

I used the following test script to verify cpufreq functionality:

```bash
#!/bin/sh

echo userspace > /sys/devices/system/cpu/cpufreq/policy0/scaling_governor
echo userspace > /sys/devices/system/cpu/cpufreq/policy4/scaling_governor

echo "--- same frequency test ---"
for freq in 614400 819000 1000000 1228800 1600000; do
     echo $freq > /sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed
     echo $freq > /sys/devices/system/cpu/cpufreq/policy4/scaling_setspeed
     sleep 1
     t=$(taskset -c 0 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
     echo "$freq kHz: ${t}s"
done

echo "--- mixed frequency test ---"
echo 614400 > /sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed
echo 1600000 > /sys/devices/system/cpu/cpufreq/policy4/scaling_setspeed
sleep 1
t0=$(taskset -c 0 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
t4=$(taskset -c 4 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
echo "p0=614400 p4=1600000 -> cpu0=${t0}s cpu4=${t4}s"

echo 1600000 > /sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed
echo 614400 > /sys/devices/system/cpu/cpufreq/policy4/scaling_setspeed
sleep 1
t0=$(taskset -c 0 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
t4=$(taskset -c 4 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
echo "p0=1600000 p4=614400 -> cpu0=${t0}s cpu4=${t4}s"

echo 1000000 > /sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed
echo 1228800 > /sys/devices/system/cpu/cpufreq/policy4/scaling_setspeed
sleep 1
t0=$(taskset -c 0 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
t4=$(taskset -c 4 time -f "%e" dd if=/dev/zero of=/dev/null bs=1M count=1000 2>&1 | tail -1)
echo "p0=1000000 p4=1228800 -> cpu0=${t0}s cpu4=${t4}s"
```

Results:

# ./test_cpufreq_userspace.sh
--- same frequency test ---
614400 kHz: 0.90s
819000 kHz: 0.67s
1000000 kHz: 0.55s
1228800 kHz: 0.45s
1600000 kHz: 0.34s
--- mixed frequency test ---
p0=614400 p4=1600000 -> cpu0=0.90s cpu4=0.34s
p0=1600000 p4=614400 -> cpu0=0.34s cpu4=0.90s
p0=1000000 p4=1228800 -> cpu0=0.55s cpu4=0.45s


Tested-by: Gong Shuai <[email protected]> # OrangePi-RV2

Regards,
Shuai


> ---
> Changes in v4:
> - Represent K1 as two cpufreq-dt policies, one per CPU clock cluster
> - Use OPP voltage ranges for the shared CPU supply
> - Link to v3: https://lore.kernel.org/r/[email protected]
> 
> Changes in v3:
> - Add a K1-specific cpufreq driver for the shared-rail, dual-clock topology
> - Use one shared CPU OPP table and one cpufreq policy for all CPUs
> - Link to v2: https://lore.kernel.org/r/[email protected]
> 
> Changes in v2:
> - Move OPP tables to dedicated k1-opp.dtsi
> - Enable OPP only on BPI-F3 with cpu-supply present
> - Link to v1: https://lore.kernel.org/r/[email protected]
> 
> ---
> Shuwei Wu (2):
>        cpufreq: dt-platdev: Add SpacemiT K1 SoC to the allowlist
>        riscv: dts: spacemit: Add cpu scaling for K1 SoC
> 
>   arch/riscv/boot/dts/spacemit/k1-bananapi-f3.dts |  35 +++++++-
>   arch/riscv/boot/dts/spacemit/k1-opp.dtsi        | 105 ++++++++++++++++++++++++
>   arch/riscv/boot/dts/spacemit/k1.dtsi            |   8 ++
>   drivers/cpufreq/cpufreq-dt-platdev.c            |   2 +
>   4 files changed, 149 insertions(+), 1 deletion(-)
> ---
> base-commit: 5164e95565d3fd508ca8a95351323f5716dfb695
> change-id: 20260307-shadow-deps-3582a78aa756
> prerequisite-patch-id: 154bd4f720ce5065d58b988de8f273207b44572e
> prerequisite-message-id: <[email protected]>
> prerequisite-patch-id: 5da3e75b18291a5540d4f66d7a0600fb8975ef62
> prerequisite-patch-id: bcf41917414ecef8cf743095d130f6004c32f6a5
> prerequisite-patch-id: cfe3800f8c791ec4c63e070af9628e88e0fc31b9
> prerequisite-message-id: <[email protected]>
> prerequisite-patch-id: 7c7fb9f87dba019ece4c97c45750349a7cd28f3a
> 
> Best regards,