On 05/03/2018 12:03 AM, Dave Hansen wrote:
On 05/02/2018 02:08 PM, Florian Weimer wrote:
On 05/02/2018 05:28 PM, Dave Hansen wrote:
The other option here that I think we discussed in the past was to have
an*explicit* signal PKRU value. That way, we can be restrictive by
default but allow overrides for special cases like you have.
That's the patch I posted before (with PKEY_ALLOC_SETSIGNAL). I'm
afraid we are going in circles.
Could you remind us why you abandoned that approach and its relative
merits versus this new approach?
Ram argued for the PKEY_ALLOC_SIGNALINHERIT and no one else objected or
was interested at the time. I may have misread the consensus.
I'm not sure what do here. I tried to submit patches for the two
suggested approaches, and each one resulted in suggests to implement the
other semantics instead.
Thanks,
Florian
