Re: [PATCH v2] cpupower: fix use-after-free and stale limits in frequency-set
From: Shuah Khan
Date: Mon Nov 16 2026 - 18:30:26 EST
On 11/15/26 03:03, Lukas Beckmann wrote:
do_new_policy() directly fills the unset fields of new_pol from the
current CPU's policy. This is problematic because do_new_policy() is
called with the same new_pol for multiple CPUs by do_one_cpu() in the
slow path. This results in stale values in new_pol being applied
(starting from the second CPU that should be changed), and even a
use-after-free (for the governor string).
Fix this by building the target policy from new_pol and the current
policy in do_new_policy() instead of filling the missing values into
new_pol.
Fixes: 7fe2f6399a84 ("cpupowerutils - cpufrequtils extended with quite some features")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Lukas Beckmann <lbckmnn@xxxxxxxxxxx>
---
Testing and minimal reproducer:
Running the command
./cpupower frequency-set --max 2500MHz --min 2500MHz
with ASAN reports the use-after-free before, but not after the patch.
Run the following commands:
# set CPU 0 to a different maximum than CPU 1
echo 2000000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
echo 3000000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq
# Change min and governor (but not max) of CPU 0 and 1
./cpupower -c 0,1 frequency-set --min 1GHz -g performance
# See if max got changed
cat /sys/devices/system/cpu/cpu[0,1]/cpufreq/scaling_max_freq
This shows that the stale maximum of CPU0 is used for CPU1 before the
patch, but not after the patch.
---
Changes in v2:
- use a shorter changelog and testing section
- build temp_pol from new_pol and cur_pol instead of copying new_pol
- Link to v1: https://patch.msgid.link/20261112-cpupower-freq-set-use-after-free-v1-1-4e0034927b3f@xxxxxxxxxxx
---
tools/power/cpupower/utils/cpufreq-set.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/tools/power/cpupower/utils/cpufreq-set.c b/tools/power/cpupower/utils/cpufreq-set.c
index 06cd4b280132..674e3aabd31d 100644
--- a/tools/power/cpupower/utils/cpufreq-set.c
+++ b/tools/power/cpupower/utils/cpufreq-set.c
@@ -141,9 +141,10 @@ static unsigned long string_to_frequency(const char *str)
}
}
-static int do_new_policy(unsigned int cpu, struct cpufreq_policy *new_pol)
+static int do_new_policy(unsigned int cpu, const struct cpufreq_policy *new_pol)
{
struct cpufreq_policy *cur_pol = cpufreq_get_policy(cpu);
+ struct cpufreq_policy temp_pol;
int ret;
if (!cur_pol) {
@@ -151,16 +152,22 @@ static int do_new_policy(unsigned int cpu, struct cpufreq_policy *new_pol)
return -EINVAL;
}
- if (!new_pol->min)
- new_pol->min = cur_pol->min;
Why can't you leave this logic as is and set temp_pol fieds,
if (!new_pol->min)
temp_pol = cur_pol->min;
else
temp_pol.min = new_pol->min;
+ if (new_pol->min)
+ temp_pol.min = new_pol->min;
+ else
+ temp_pol.min = cur_pol->min;
- if (!new_pol->max)
- new_pol->max = cur_pol->max;
+ if (new_pol->max)
+ temp_pol.max = new_pol->max;
+ else
+ temp_pol.max = cur_pol->max;
- if (!new_pol->governor)
- new_pol->governor = cur_pol->governor;
+ if (new_pol->governor)
+ temp_pol.governor = new_pol->governor;
+ else
+ temp_pol.governor = cur_pol->governor;
Same for all of the above. Make just the minimum changes
to fix the problem instead of changing the logic
- ret = cpufreq_set_policy(cpu, new_pol);thanks,
+ ret = cpufreq_set_policy(cpu, &temp_pol);
cpufreq_put_policy(cur_pol);
-- Shuah