One possible fix has been committed. Can you update the source code and try a new kernel?
Thanks, ozaki-r On Tue, Dec 26, 2017 at 1:00 AM, Ryota Ozaki <ozak...@netbsd.org> wrote: > On Mon, Dec 25, 2017 at 8:31 PM, Tom Ivar Helbekkmo > <t...@hamartun.priv.no> wrote: >> Martin Husemann <mar...@duskware.de> writes: >> >>> The sparc64 hang happened with: >>> >>> $NetBSD: ip_output.c,v 1.288 2017/12/22 11:22:37 ozaki-r Exp $ >> >> My amd64 that experienced hangs has: >> >> $NetBSD: ip_output.c,v 1.287 2017/12/15 04:03:46 ozaki-r Exp $ >> >> ...and was last updated from CVS on December 19th, which made it much >> more stable than a kernel from a few days previous to that (from the >> 12th, I think). After upgrading, I was trying to update a lot of >> installed packages from pkgsrc, with pkgsrc, distfiles, and the binary >> package repository NFS mounted, and a local pkgobj for building in. I >> gave up in the end, because of the hangs, but managed to build >> everything after updating on the 19th. Since then, I've only had the >> one hard hang. > > My recent changes: > - 12/6-7: IFNET_LOCK changes #1 > - 12/14-15: IFNET_LOCK changes #2 > - 12/19: Disable IFEF_MPSAFE by default > > So your experiences looks synchronized to the changes. > > Can you run the kernel with LOCKDEBUG and DEBUG enabled? And if once your > system hangs again, can you enter the DDB and collect threads stuck on > turnstile? > > Thanks, > ozaki-r
