Re: efibootmgr: Cannot translate unix loader path xxx\xxx\xxx to UEFI: No error: 0

On Fri, Dec 28, 2018, 5:45 AM O. Hartmann <ohartmann_at_walstatt.org wrote:

> On Thu, 27 Dec 2018 08:38:20 -0700
> Warner Losh <imp_at_bsdimp.com> wrote:
>
> > On Dec 27, 2018 7:42 AM, "O. Hartmann" <ohartmann_at_walstatt.org> wrote:
> >
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA256
> >
> > Am Thu, 27 Dec 2018 13:29:44 +0100
> > "Hartmann, O." <ohartmann_at_walstatt.org> schrieb:
> >
> >
> > > Updated Fujitsu Celsius M740 to its lates UEFI firmware today.
> > > After this update, the box won't boot FreeBSD 12-STABLE anymore! With
> > > disabled CSM, the firmware doesn't recognise the boot SSD's freebsd-efi
> > > partition for UEFI boot anymore - which was no problem before.
> > >
> > > When trying FreeBSD 13-CURRENT (USB image from 26.12.2018 as of the
> > > snapshot site) I receive a malloc arena error when trying to set boot
> > > vars via efibootmgr utility. So I tried the recent 12-STABLE snapshot
> > > as of 26th December 2018, the same as CURRENT USB Image, and I receive
> > > a weird error:
> > >
> > > efibootmgr -c -l /mnt/EFI/BOOT/BOOTX64.EFI -L FreeBSD
> > >
> > > efibootmgr: Cannot translate unix loader path
> > > '\mnt\EFI\BOOT\BOOTX64.EFI' to UEFI: No error: 0
> > >
> > > What the heck is that?
> > >
> > > What does this error mean? No error: 0?
> > >
> > > The box is unusable.
> > >
> > >
> > > Kind regards,
> > >
> > > O. Hartmann
> > >
> > >
> > >
> >
> > I found this PR, Bug 229191, from June, 2018:
> >
> > https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=229191
> >
> > It seems the problem has not been fixed. Indeed did I mount the ESP
> > via a GEOM label, /dev/gpt/efiboot0.
> >
> >
> > There is some code that tries to cope, but I ran out of time before it
> was
> > bulletproof.  You can use nda0p9:/path/in/fs instead. Assuming the esp is
> > on /dev/nda0p9.
> >
> > Warner
> [...]
>
> The workaround (alternative mounting) given in the mentioned PR solved the
> problem.
>
> This is the second time I face crude problems with Fujitsu
> hardware/firmware
> and if I wouldn't solved a similar problem this summer for an Esprimo
> Q956, the
> problem would have cost me valuable time. So, my thinking is: couldn't
> there be
> a short paragraph in the handbook?
>

Until it's fixed, sure. Labels and mirrors are not as well supported as I'd
like... knowing the device + path workaround is the way we should document
in the meantime.

Warner

Warner

>