On Thu, Mar 30, 2017 at 12:02 PM, Paul Eggert [off-list ref] wrote:

quoted

I'm assuming you'd also possible want to be able to use F_SETFL to set
O_ATOMIC after the fact

Just for fun, one thread can set O_ATOMIC at the same time another thread is
doing a 'write'....

I'm sure that falls under "if you break it, you get to keep both
pieces". IOW, I don't think anybody will ever say that the concurrent
write has to have some particular semantics wrt the concurrent
O_ATOMIC. Maybe *part* of the write will be done with some semantics,
and part of the write will be done with other semantics.

My guess is that there is going to be very few O_ATOMIC users anyway,
and they'll very carefully set it once and test it (or not even test
it - just make it be a configuration flag and tell people "don't ask
for O_ATOMIC if your system doesn't support it")

                  Linus