On Sat, Jul 14, 2018 at 5:54 PM, Deepa Dinamani deepa.kernel@gmail.com wrote:
Right now we only have compat syscalls and native syscalls. Until we transition all the architectures to use the new syscalls, wouldn't it be the same sort of confusion as exists today? These structures today are still used by compat entry points. I'm trying to understand why such a cleanup would make sense today rather than at the end of the transition.
I don't think it makes a big difference whether we change it now or later, but if Christoph feels that it addresses his concern about the compat_ namespace being reused during the transition, doing it earlier would enable us to finish the remaining syscalls.
We are running a little late with the remaining syscalls for 4.19 as we now have an rc5. At this point we probably won't have time to finish the discussion about all of clock_adjtime, getrusage/waitid and getitimer(), but I'd really like to get utimensat, futex, sched_rr_get_interval, recvmmsg, sys_io_getevents, rt_sigtimedwait, ppoll and pselect6 done for 4.19 so we can finalize the ABI for 4.20 along with doing the actual conversion.
Arnd