On Tue, May 18, 2021 at 10:02 PM George Koehler <kern...@gmail.com> wrote: > My G5 did hang in the same way, and I found the bug in the OpenBSD > kernel. I got unlucky when the system randomly reordered my kernel. > Then the kernel got an unlucky page fault while trying to acquire or > release the mp kernel's pmap's hash lock. The unlucky fault hung at > the "using parent" line. The fault can happen only on G5. > > The problem might go away by again reordering the kernel. The easiest > way (but I'm not sure) might be to boot bsd.rd and use its "upgrade" > feature to reinstall base69.tgz. I took a different way: I had been > compiling my own kernels, so when my bsd.mp froze, I booted an older > bsd.mp that still worked, and compiled a new bsd.mp. > > My diff, below, tries to fix the problem. I would like people, who > have a macppc with more than one cpu, to compile and run a GENERIC.MP > kernel with my diff, and tell me if it has trouble. The diff is for > OpenBSD-current but should also apply to 6.9. Something like,
I tried applying the patch to stable and current last night, both resulting in failed apply and reject files. I don't think I have a big enough partition for my /usr/src directory to actually compile the kernel. The auto-layout created 2G, but I don't have much leftover when bringing down the source tree. At this point, I may just re-install and give myself plenty more space for the source partition. I tried searching the mailing list archive, I didn't find an answer to why macppc is limited to 2G memory? Thanks, Tony
