Re: [PATCH] thermal: fix frequency table lookup bugs

[Eduardo Valentin <eduardo.valentin@ti.com>]

On 09-04-2013 13:02, Andrew Bresticker wrote:
> Hi Eduardo,
>
> On Tue, Apr 9, 2013 at 7:55 AM, Eduardo Valentin
> <eduardo.valentin@ti.com> wrote:
>> Hi Andrew,
>>
>>
>> On 08-04-2013 19:54, Andrew Bresticker wrote:
>>>
>>> The loops which are used to perform lookups in CPU frequency tables in
>>> cpu_cooling and the Exynos thermal driver do not update the loop counter
>>> if they encounter an invalid table entry, leading to an infinite loop in
>>> that case.
>>>
>>> Signed-off-by: Andrew Bresticker <abrestic@chromium.org>
>>> ---
>>>    drivers/thermal/cpu_cooling.c    | 19 ++++++++++---------
>>>    drivers/thermal/exynos_thermal.c |  8 ++++----
>>>    2 files changed, 14 insertions(+), 13 deletions(-)
>>>
>>> diff --git a/drivers/thermal/cpu_cooling.c b/drivers/thermal/cpu_cooling.c
>>> index 836828e..e6db441 100644
>>> --- a/drivers/thermal/cpu_cooling.c
>>> +++ b/drivers/thermal/cpu_cooling.c
>>> @@ -124,14 +124,14 @@ static int is_cpufreq_valid(int cpu)
>>>    static unsigned int get_cpu_frequency(unsigned int cpu, unsigned long
>>> level)
>>>    {
>>>          int ret = 0, i = 0;
>>> -       unsigned long level_index;
>>> +       unsigned long level_index = 0;
>>>          bool descend = false;
>>>          struct cpufreq_frequency_table *table =
>>>                                          cpufreq_frequency_get_table(cpu);
>>>          if (!table)
>>>                  return ret;
>>>
>>> -       while (table[i].frequency != CPUFREQ_TABLE_END) {
>>> +       for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
>>>                  if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
>>>                          continue;
>>
>> Wouldn't be easier to just increase the index i before doing a continue?
>
> I think this is cleaner.  The code is iterating through an array -- it
> should be a for loop.
>
>>
>>
>>>
>>> @@ -143,24 +143,25 @@ static unsigned int get_cpu_frequency(unsigned int
>>> cpu, unsigned long level)
>>>                  }
>>>
>>>                  /*return if level matched and table in descending order*/
>>> -               if (descend && i == level)
>>> +               if (descend && level_index == level)
>>>                          return table[i].frequency;
>>
>>
>> What this has to do with the patch description?
>
> I'm using level_index as the counter of valid frequencies, where as i
> is the index into the array.  If there are invalid entries, they are
> not necessarily equal.  The point of this function is to find the
> level-th *valid* frequency in the table.
>
>> Besides why would you be comparing level against 0 all the time (you have
>> initialized level_index to 0 at this point).
>
> Huh? level_index is clearly incremented below...
>
>>
>>> -               i++;
>>> +               level_index++;
>>
>>
>> level_index wont be updated in case of INVALID entry.
>
> That's the point.
>
>>
>>
>>>          }
>>>          i--;
>>> +       level_index--;
>>>
>>> -       if (level > i || descend)
>>> +       if (level > level_index || descend)
>>>                  return ret;
>>> -       level_index = i - level;
>>> +       level = level_index - level;
>>>
>>>          /*Scan the table in reverse order and match the level*/
>>> -       while (i >= 0) {
>>> +       for (; i >= 0; i--) {
>>>                  if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
>>>                          continue;
>>>                  /*return if level matched*/
>>> -               if (i == level_index)
>>> +               if (level_index == level)
>>>                          return table[i].frequency;
>>> -               i--;
>>> +               level_index--;
>>>          }
>>
>>
>> I believe you do more than what you have described in your intention under
>> you patch description
>
> I disagree.  I'm fixing the loop so that it properly handles invalid
> entries and thus the infinite loop problem I mention in the commit
> message.
>

In this case, I believe you should also rephrase your patch description, 
explaining that you are also fixing a role for each index.

>> Can you please split your patch into smaller changes?
>
> I don't think there is a need for separate patches to cpu_cooling.c.
>


You do two things in this change on cpu_cooling.c: (1) fix the case 
where the loop is kept running indefinitely. (2) Reserve a specific role 
for each index in this function.

For this reason, I suggested doing one thing per patch and splitting 
this change into two for better review process. Having that split with a 
good description for each change makes everyone life easier, don t you 
think?


>>
>>>          return ret;
>>>    }
>>> diff --git a/drivers/thermal/exynos_thermal.c
>>> b/drivers/thermal/exynos_thermal.c
>>> index d5e6267..524b2a0 100644
>>> --- a/drivers/thermal/exynos_thermal.c
>>> +++ b/drivers/thermal/exynos_thermal.c
>>> @@ -237,7 +237,7 @@ static int exynos_get_crit_temp(struct
>>> thermal_zone_device *thermal,
>>>
>>>    static int exynos_get_frequency_level(unsigned int cpu, unsigned int
>>> freq)
>>>    {
>>> -       int i = 0, ret = -EINVAL;
>>> +       int i, level = 0, ret = -EINVAL;
>>>          struct cpufreq_frequency_table *table = NULL;
>>>    #ifdef CONFIG_CPU_FREQ
>>>          table = cpufreq_frequency_get_table(cpu);
>>> @@ -245,12 +245,12 @@ static int exynos_get_frequency_level(unsigned int
>>> cpu, unsigned int freq)
>>>          if (!table)
>>>                  return ret;
>>>
>>> -       while (table[i].frequency != CPUFREQ_TABLE_END) {
>>> +       for (i = 0; table[i].frequency != CPUFREQ_TABLE_END; i++) {
>>>                  if (table[i].frequency == CPUFREQ_ENTRY_INVALID)
>>>                          continue;
>>>                  if (table[i].frequency == freq)
>>> -                       return i;
>>> -               i++;
>>> +                       return level;
>>> +               level++;
>>
>>
>> Can you please send a separate patch on this driver instead?
>
> Sure.
>
>>
>>
>>
>>>          }
>>>          return ret;
>>>    }
>>>
>>
>
> Thanks,
> Andrew
>
>