Re: [PATCH] cpufreq: Fix policy->freq_table is NULL in __cpufreq_driver_target()

["Yajun Deng" <yajun.deng@linux.dev>]

March 29, 2023 10:21 PM, "Rafael J. Wysocki" <rafael@kernel.org> wrote:

> On Wed, Mar 29, 2023 at 3:36 PM Yajun Deng <yajun.deng@linux.dev> wrote:
> 
>> __resolve_freq() may be return target_freq if policy->freq_table is
>> NULL. In this case, it should return -EINVAL before __target_index().
> 
> Even so, __target_index() itself doesn't dereference freq_table
> AFAICS, so arguably the driver should be prepared to deal with a NULL
> freq_table which comes from it after all.
> 

But there is a statement 'unsigned int newfreq = policy->freq_table[index].frequency;'
in __target_index(), if driver doesn't provide freq_table, __target_index()
will fault before the driver itself.

> Or, if you want to argue that drivers providing ->target_index() must
> also provide freq_table that is not NULL, a check to that effect needs
> to be done at the driver registration time IMO.
> 
>> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
>> ---
>> drivers/cpufreq/cpufreq.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
>> index c0e5be0fe2d6..308a3df1a940 100644
>> --- a/drivers/cpufreq/cpufreq.c
>> +++ b/drivers/cpufreq/cpufreq.c
>> @@ -2299,7 +2299,7 @@ int __cpufreq_driver_target(struct cpufreq_policy *policy,
>> return cpufreq_driver->target(policy, target_freq, relation);
>> }
>> 
>> - if (!cpufreq_driver->target_index)
>> + if (!cpufreq_driver->target_index || !policy->freq_table)
>> return -EINVAL;
>> 
>> return __target_index(policy, policy->cached_resolved_idx);
>> --