Re: [PATCH] powercap: Adjust printing the constraint name with new line

[Lukasz Luba <lukasz.luba@arm.com>]

On 11/17/20 12:51 PM, Rafael J. Wysocki wrote:
> On Tue, Nov 17, 2020 at 10:22 AM Lukasz Luba <lukasz.luba@arm.com> wrote:
>>
>> Hi Rafael,
>>
>> On 11/10/20 7:19 PM, Rafael J. Wysocki wrote:
>>> On Mon, Nov 9, 2020 at 6:25 PM Lukasz Luba <lukasz.luba@arm.com> wrote:
>>>>
>>>> The constraint name has limit of size 30, which sometimes might be hit.
>>>> When this happens the new line might be lost. Prevent this and set the
>>>> new line when the name string is too long."
>>>>
>>>> Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
>>>> ---
>>>>    drivers/powercap/powercap_sys.c | 4 ++++
>>>>    1 file changed, 4 insertions(+)
>>>>
>>>> diff --git a/drivers/powercap/powercap_sys.c b/drivers/powercap/powercap_sys.c
>>>> index f808c5fa9838..575f9fdb810e 100644
>>>> --- a/drivers/powercap/powercap_sys.c
>>>> +++ b/drivers/powercap/powercap_sys.c
>>>> @@ -174,6 +174,10 @@ static ssize_t show_constraint_name(struct device *dev,
>>>>                                                                   "%s\n", name);
>>>>                           buf[POWERCAP_CONSTRAINT_NAME_LEN] = '\0';
>>>>                           len = strlen(buf);
>>>> +
>>>> +                       /* When the 'name' is too long, don't lose new line */
>>>> +                       if (strlen(name) >= POWERCAP_CONSTRAINT_NAME_LEN)
>>>> +                               buf[POWERCAP_CONSTRAINT_NAME_LEN - 1] = '\n';
>>>
>>> Wouldn't it be better to pass POWERCAP_CONSTRAINT_NAME_LEN - 1 to
>>> snprintf() above?
>>
>> My apologies for delay, I was on holidays.
>>
>> The snprintf() overwrites the '\n' in that case also.
> 
> The snprintf() doesn't do that AFAICS, but the next assignment
> overwrites the newline with the zero character.
> 
> Anyway, it should have been POWERCAP_CONSTRAINT_NAME_LEN - 2, because
> it needs to leave space for 2 characters, the newline and the
> terminating zero.
> 
>> I've experimented with it a bit more and tried to come with a bit 'cleaner' solution.
>>
>> What we are looking is probably: "%.*s\n" semantic.
>> Another solution would be:
>> ------------------------------8<---------------------------
>>                   if (name) {
>> -                       snprintf(buf, POWERCAP_CONSTRAINT_NAME_LEN,
>> -                                                               "%s\n",
>> name);
>> -                       buf[POWERCAP_CONSTRAINT_NAME_LEN] = '\0';
>> +                       sprintf(buf, "%.*s\n",
>> POWERCAP_CONSTRAINT_NAME_LEN -1,
>> +                               name);
>>                           len = strlen(buf);
>>                   }
>>
>> ----------------------------->8----------------------------
>>
>> I've check this and it behaves very well.
>>
>> It looks cleaner and it is a used pattern in the kernel.
>> What do you think? Is it good for v2?
> 
> This works for me too.
> 
> Thanks!
> 

Thank you for the comments. I will send the v2 then.

Regards,
Lukasz