Re: Getting boottime early in booting
- In reply to: Rick Macklem : "Re: Getting boottime early in booting"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sat, 04 Apr 2026 08:17:44 UTC
On Sat, Apr 04, 2026 at 12:37:52AM -0700, Rick Macklem wrote: > On Fri, Apr 3, 2026 at 8:15 PM Konstantin Belousov <kib@freebsd.org> wrote: > > > > On Fri, Apr 03, 2026 at 07:13:02PM -0600, Warner Losh wrote: > > > On Fri, Apr 3, 2026, 7:05 PM Rick Macklem <rick.macklem@gmail.com> wrote: > > > > > > > Hi, > > > > > > > > I'm fiddling with a NFSv4 diskless root fs setup and I've run > > > > into a couple of challenges. > > > > > > > > 1 - I need a value that will be different each time the machine boots. > > > > I would normally use getboottime(), but for the NFS root fs mount, > > > > it just returns 0. > > > > Any ideas w.r.t. something I can acquire early in booting that will > > > > change each time the system is booted? > > > > > > > > > > We don't have this today. And we can't have it without hardware assist... > > > though you may be ok enough with a value from the rtc. > > > > > > There's work to pass the vale from the monotonicly increasing counter uefi > > > provides. But not on all platforms. > > > > We already have mechanisms to ensure that early entropy is provided on > > boot. Do you care if the value used is indeed boot time, or should it > > only be unique per boot? > > > > With the entropy loading, I believe you would get what you need on any > > machine with the persistent storage. > Hmm, do you mean the stuff under /var/db/entropy? It is /boot/entropy I think. Also I see 'entropy_efi_seed' which implies that loader might get some entropy from the UEFI facilities. But this is a speculation based on the documentation. > If so, that doesn't work for this case, since for an NFS root fs, /etc is > usually a memory fs. > (See https://docs-archive.freebsd.org/doc/8.4-RELEASE/usr/share/doc/freebsd/handbook/network-pxe-nfs.html) > > Without a unique per boot value, things work, it's just that it > loops once more during the mount attempt, each time it is > rebooted. (Without boring you with the details, each time > a seq# gets incremented by 1 and it needs to loop until > it gets to the next seq#. I can tweak the mount code so > that it doesn't spew out printf()s when it loops.) If UEFI source works, then the machine should get enough entropy even for shared rootfs. > > Thanks, rick > > > > > > > > > > Warner > > > > > > 2 - Does anyone know how to fix "could not determine audit condition" > > > > which I get whenever I try and login when it is multiuser? > > > > (I can run single user, but since I can't login once it's multiuser??) > > > > > > > > Thanks for any help with these, rick > > > > > > > > > > >