{"id":"CVE-2026-97905","published":"2026-09-25T11:17:17.403","lastModified":"2026-09-25T11:17:17.403","description":"In the Linux kernel, the following vulnerability has been resolved:\n\ncpufreq: zero-initialize policy cpumask before sysfs publication\n\ncpufreq_policy_alloc() allocates policy->cpus with alloc_cpumask_var(),\ni.e. without __GFP_ZERO, unlike the sibling related_cpus and real_cpus\nmasks. With CONFIG_CPUMASK_OFFSTACK=y the mask is a separate\nkmalloc_node() allocation, so its bitmap holds whatever the slab allocator\nleft behind:\n\n  cpufreq_online()\n    cpufreq_policy_alloc()\n      alloc_cpumask_var(&policy->cpus)    /* bitmap is uninitialized */\n      kobject_init_and_add()              /* policy%u/ appears in sysfs */\n    cpufreq_policy_online()\n      cpumask_copy(policy->cpus, cpumask_of(cpu))  /* first valid value */\n\nThis leaves a window in which the sysfs attributes are already reachable\nwhile policy->cpus is still garbage. show()/store() gate on\npolicy_is_inactive(), i.e. cpumask_empty(policy->cpus), so a non-zero\nbitmap makes them run the attribute callbacks on a policy that is not\ninitialized yet.\n\nFix this by using zalloc_cpumask_var() for policy->cpus.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/0de2f3918fbbbb767e7e716178194b66560dd12a","tags":[]},{"url":"https://git.kernel.org/stable/c/54d37bcf2f497140b9207968557ddb484058e749","tags":[]},{"url":"https://git.kernel.org/stable/c/6e166b9281dec98aed19213a84235250e5fdb081","tags":[]},{"url":"https://git.kernel.org/stable/c/bbc0472d2270bf732142ca71579d6ede636174ed","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}