> From: Paul Irofti <[email protected]> > Date: Fri, 12 Jun 2020 12:30:22 +0300 > > On 12.06.2020 10:48, Robert Nagy wrote: > > On 11/06/20 20:10 +0200, Mark Kettenis wrote: > >>> Date: Thu, 11 Jun 2020 19:38:48 +0200 > >>> From: Christian Weisgerber <[email protected]> > >>> > >>> Theo de Raadt: > >>> > >>>> The diff is growing complexity to support a future which wouldn't > >>>> exist if attempts at *supporting all* architectures received priority. > >>> > >>> Adding support for more archs is very simple, since you just need > >>> to copy the corresponding get_timecounter function from the kernel. > >>> > >>> Here's arm64. I'm running a kernel and libc with this. > >>> > >>> I can also provide alpha, powerpc, and sparc64, but I don't have > >>> such machines. > >> > >> Hope you didn't spend too much time on that, because I already > >> mentioned that I had arm64 working earlier in the thread. > >> > >> I've just fired up one of my sparc64 machines such that I can check > >> how well the approach works for an architecture with two exported > >> timecounters. > > > > Then please share the patches so that it can be integrated into the > > main diff so that when the time comes it can go in at one shot. > > > > Also it would help to avoid duplicate work. > > I will respond to deraadt@'s question about where are the clocks that I > mentioned we already have support for and to your message also: > > - there are diffs in this thread providing diffs for those archs > - I did not integrate them in the big diff because I did not want to > get into an argument about who commits what (some people care about > their commits number apparently) > > I can integrate those clocks easily of course if I am allowed to commit > them too at the end, if not, I don't want to get involved in that as I > will probably forget some and drama will happen. > > So from kettenis's comments I gather we have: amd64, macppc, sparc64 and > arm64?
I now have a working implementation on sparc64, where I can switch between %tick and %sys_tick while a process doing repeated gettimeofday() calls keeps running. However, on this machine both counters are synchronized so that isn't the most significant test of this functionality.
