Re: [RFC PATCH v1 2/2] chnet: Implement `get_thread()` and `put_thread()` function

[Alviro Iskandar Setiawan <alviro.iskandar@gnuweeb.org>]

On Mon, Aug 29, 2022 at 12:48 PM Ammar Nofan Faizi wrote:
> On 8/29/22 12:38 PM, Alviro Iskandar Setiawan wrote:
>> On Mon, Aug 29, 2022 at 12:29 PM Ammar Nofan Faizi wrote:
>>> On 8/29/22 12:24 PM, Alviro Iskandar Setiawan wrote:
>>>> On Mon, Aug 29, 2022 at 12:17 PM Alviro Iskandar Setiawan wrote:
>>>>> On Mon, Aug 29, 2022 at 11:54 AM Ammar Nofan Faizi wrote:
>>>>>> On 8/29/22 11:41 AM, Alviro Iskandar Setiawan wrote:
>>>>>>> On Mon, Aug 29, 2022 at 8:11 AM Ammar Faizi wrote:
>>>>>>>> +static base::Thread *get_thread(void)
>>>>>>>> +{
>>>>>>>> +       const uint32_t max_ch_thpool = g_max_ch_thpool;
>>>>>>>> +       const uint32_t nr_ref_split = 2048;
>>>>>>>> +       struct ch_thpool **thp;
>>>>>>>> +       struct ch_thpool *ret = nullptr;
>>>>>>>> +       struct ch_thpool *tmp;
>>>>>>>> +       uint32_t min_ref_idx;
>>>>>>>> +       uint32_t min_ref;
>>>>>>>> +       uint32_t i;
>>>>>>>> +
>>>>>>>> +       g_thpool_lock_.lock();
>>>>>>>> +       thp = g_thpool;
>>>>>>>> +       if (!thp) {
>>>>>>>> +               g_thpool_lock_.unlock();
>>>>>>>> +               return nullptr;
>>>>>>>> +       }
>>>>>>>
>>>>>>> in what situation @thp can be nullptr?
>>>>>>
>>>>>> When the chnet_global_destroy() is called.
>>>>
>>>> btw, i missed a case on the chnet construction, you do
>>>>
>>>>      thread_(*get_thread()),
>>>>
>>>> if get_thread() returns a nullptr, the @thread_ will become a
>>>> reference to a nullptr, it will definitely break, how do you handle
>>>> this?
>>>
>>> I don't think we need to handle that. If @thp ever be a nullptr in a
>>> get_thread() call, then it's a misuse of the API. We must never call
>>> get_thread() after chnet_global_destroy() is called. It doesn't make
>>> much sense to have this situation.
>>
>> if this should never happen, then why do you do a null check there?
>> your null check doesn't make much sense too! why are we returning
>> nullptr if we exactly know that the caller will randomly fault if that
>> happens, i can't accept your reasoning
>
> OK.
>
> Fine, let's omit that nullptr return. To assert that path, we perform a
> CHECK() assertion, so it will just crash the whole program with a full
> call traces if the contract is violated. This way also makes debugging
> process easier. Are you happy with this change?
>
> Incremental is:
>
> diff --git a/chnet/chnet.cc b/chnet/chnet.cc
> index 782e4fe..fc536d2 100644
> --- a/chnet/chnet.cc
> +++ b/chnet/chnet.cc
> @@ -647,10 +647,7 @@ static base::Thread *get_thread(void)
>
>          g_thpool_lock_.lock();
>          thp = g_thpool;
> -        if (!thp) {
> -                g_thpool_lock_.unlock();
> -                return nullptr;
> -        }
> +        CHECK(thp);
>
>          tmp = thp[0];
>          if (!tmp) {

oc, ack, it looks fine to me now

tq

-- Viro