[PATCH] nptl: Provide a way to block all signals for the timer helper thread

[Carlos O'Donell]

On 5/14/20 11:46 AM, Florian Weimer wrote:
> * Adhemerval Zanella:
> 
>> On 12/05/2020 12:02, Florian Weimer wrote:
>>> timer_create needs to create threads with all signals blocked,
>>> including SIGTIMER (which happens to equal SIGCANCEL).  Add a new
>>> internal interface which provides an explicit way to achieve that.
>>>
>>> Fixes commit b3cae39dcbfa2432b3f3aa28854d8ac57f0de1b8 ("nptl: Start
>>> new threads with all signals blocked [BZ #25098]").
>>>
>>> Tested on x86_64-linux-gnu.
>>
>> The patch looks go, thanks.  As a side note, I am wondering if it would
>> be better to de-couple POSIX timer from cancellation handling, it should
>> avoid require a special pthread create symbols as this patch.
>>
>> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
> 
> Carlos, is it okay to check this in?  (With the false → NULL change, of
> course.)

No. I still want task-oriented internal comments to explain when and why
the new_sigmask parameter would be set. I gave a suggested paragraph in
the follow-on review. Please have a look and post a v2 or discuss further.

I know we are under pressure to fix this because it's a regression, but we
can take some time to write better task-oriented comments for the API so
future readers know how to use it.

-- 
Cheers,
Carlos.