Re: [PATCH 00/11] Introduce Simple atomic and non-atomic counters

[Shuah Khan <skhan@linuxfoundation.org>]

On 9/26/20 10:33 AM, Kees Cook wrote:
> On Fri, Sep 25, 2020 at 06:13:37PM -0600, Shuah Khan wrote:
>> On 9/25/20 5:52 PM, Kees Cook wrote:
>>> On Fri, Sep 25, 2020 at 05:47:14PM -0600, Shuah Khan wrote:
>>>> -- Addressed Kees's comments:
>>>>      1. Non-atomic counters renamed to counter_simple32 and counter_simple64
>>>>         to clearly indicate size.
>>>>      2. Added warning for counter_simple* usage and it should be used only
>>>>         when there is no need for atomicity.
>>>>      3. Renamed counter_atomic to counter_atomic32 to clearly indicate size.
>>>>      4. Renamed counter_atomic_long to counter_atomic64 and it now uses
>>>>         atomic64_t ops and indicates size.
>>>>      5. Test updated for the API renames.
>>>>      6. Added helper functions for test results printing
>>>>      7. Verified that the test module compiles in kunit env. and test
>>>>         module can be loaded to run the test.
>>>
>>> Thanks for all of this!
>>>
>>>>      8. Updated Documentation to reflect the intent to make the API
>>>>         restricted so it can never be used to guard object lifetimes
>>>>         and state management. I left _return ops for now, inc_return
>>>>         is necessary for now as per the discussion we had on this topic.
>>>
>>> I still *really* do not want dec_return() to exist. That is asking for
>>> trouble. I'd prefer inc_return() not exist either, but I can live with
>>> it. ;)
>>>
>>

I didn't read this correctly the first time around.

>> Thanks. I am equally concerned about adding anything that can be used to
>> guard object lifetimes. So I will make sure this set won't expand and
>> plan to remove dec_return() if we don't find any usages.
> 
> I would like it much stronger than "if". dec_return() needs to be just
> dec() and read(). It will not be less efficient (since they're both
> inlines), but it _will_ create a case where the atomicity cannot be used
> for ref counting. My point is that anything that _requires_ dec_return()
> (or, frankly, inc_return()) is _not_ "just" a statistical counter. It
> may not be a refcounter, but it relies on the inc/dec atomicity for some
> reason beyond counting in once place and reporting it in another.
> 

I am not thinking about efficiency rather two calls instead of one if
an decrement needs to followed by return. In any case, I agree with you
that there is no need to add dec_return now without any use-cases.

I will update the patch series to remove it.

thanks,
-- Shuah