On Wed, Sep 9, 2026, 10:21 Stuart Henderson <[email protected]> wrote:
> On 2026/09/09 16:54, Crystal Kolipe wrote: > > On Wed, Sep 09, 2026 at 06:45:13PM +0200, Alex Rnne Petersen wrote: > > > On Wed, Sep 9, 2026, at 18:27, Nick Owens wrote: > > > > On Wed, Sep 2, 2026 at 4:55???PM <[email protected]> wrote: > > > >> > > > >> >Synopsis: Network daemons sporadically fail during boot in > arm64 VM > > > >> >Category: system > > > >> >Environment: > > > >> System : OpenBSD 7.9 > > > >> Details : OpenBSD 7.9 (GENERIC.MP) #8: Fri Aug 21 > 13:35:08 MDT 2026 > > > >> [email protected]: > /usr/src/sys/arch/arm64/compile/GENERIC.MP > > > >> > > > >> Architecture: OpenBSD.arm64 > > > >> Machine : arm64 > > > >> >Description: > > > >> We (Zig) recently acquired an AmpereOne A192-32X machine to > self-host > > > >> and cover all our arm64 CI needs. OpenBSD is of course one > of the > > > >> platforms we would like to test on arm64. Unfortunately, we > are facing > > > >> an issue where, in the VM we created for OpenBSD, the basic > network > > > >> daemons will sporadically fail during boot for no apparent > reason. > > > >> > > > >> The affected daemons are (at least) dhcpleased, ntpd, > resolvd, and > > > >> slaacd. Sometimes only a subset will fail, sometimes all, > sometimes > > > >> none. All I can tell is that they seem to be getting killed > during rc, > > > >> likely by the kernel. The only indication of this that I > see is "Killed" > > > >> messages in the console, but even these don't always appear > when the > > > >> issue happens. There also does not appear to be anything > relevant in any > > > >> /var/log file. When it happens, I can rcctl start the > failed daemons > > > >> from the console, and things appear fine from there. > > > > > > > > this sounds a bit like > > > > https://marc.info/?l=openbsd-tech&m=178407511210176&w=2 - are you > sure > > > > swap is on, or is it off? > > > > > > The symptoms described there do match my experience exactly, however, > swap > > > is definitely on and defined correctly in /etc/fstab. But as noted, > it's a > > > swap *file* rather than partition, so perhaps that matters? > > > > Given the specific daemons you've listed, your swap file is probably not > > active when those are started by rc. Have a look at the ordering in > /etc/rc. > > > > indeed. using a partition rather than a file should help (and means > you're less likely to hit lesser-used codepaths; the installer creates > a swap partition by default and that's what most people use. swapping > to a file is pretty unusual on OpenBSD). > in my original tests where I hit the issue I never used a swap file, it was only swapless vms. I don't know what the ordering is of swap partition vs maybe the later swap file mounting but if it is later than daemons start it can still hit the same issue, that happens on both arm64 and riscv64. I'm not sure why there were no responses to my tech patch. >
