svn commit: r335276 - in head/stand/i386: gptboot zfsboot

[Warner Losh]

On Wed, Jun 20, 2018 at 10:03 AM, Konstantin Belousov <kostikbel at gmail.com>
wrote:

> On Wed, Jun 20, 2018 at 09:43:31AM -0600, Ian Lepore wrote:
> > On Wed, 2018-06-20 at 12:22 +0300, Konstantin Belousov wrote:
> > > On Tue, Jun 19, 2018 at 08:34:18PM -0400, Allan Jude wrote:
> > > >
> > > > On 2018-06-17 07:32, Eugene Grosbein wrote:
> > > > >
> > > > > 17.06.2018 10:18, Allan Jude wrote:
> > > > >
> > > > > >
> > > > > > Author: allanjude
> > > > > > Date: Sun Jun 17 03:18:56 2018
> > > > > > New Revision: 335276
> > > > > > URL: https://svnweb.freebsd.org/changeset/base/335276
> > > > > >
> > > > > > Log:
> > > > > >   gptboot, zfsboot, gptzfsboot: Enable the video and serial
> consoles early
> > > > > >
> > > > > >   Normally the serial console is not enabled until /boot.config
> is read and
> > > > > >   we know how the serial console should be
> configured.  Initialize the
> > > > > >   consoles early in 'dual' mode (serial & keyboard) with a
> default serial
> > > > > >   rate of 115200. Then serial is re-initialized once the disk is
> decrypted
> > > > > >   and the /boot.config file can be read.
> > > > > >
> > > > > >   This allows the GELIBoot passphrase to be provided via the
> serial console.
> > > > > >
> > > > > >   PR:         221526
> > > > > >   Requested by:       many
> > > > > >   Reviewed by:        imp
> > > > > >   Sponsored by:       Klara Systems
> > > > > >   Differential Revision:      https://reviews.freebsd.org/D15862
> > > > > I had several cases when booting FreeBSD/amd64 with motherboard
> having no serial ports
> > > > > hang hard early at boot unless I rebuilt boot media configuring it
> to NOT try accessing
> > > > > missing serial ports. I even could reproduce that with VirtualBox
> machine configured
> > > > > with no serial ports (not same as existing bug inactive serial
> port).
> > > > >
> > > > > Should there be some way to disable this serial ports
> configuration at compile time?
> > > > >
> > > > >
> > > > >
> > > > I think what we'll do it compile it both ways, and use the non-serial
> > > > one by default, because it is safer. Then you can just use
> > > > 'gptboot-serial' if you want serial support.
> > > >
> > > > This will likely make Warner a bit sad, since we are just finally
> > > > getting around to reducing the number of different bootcode files.
> > > I think we should follow the hardware trends there and apply a policy
> > > where new features are not added to the CSM boot. All modern machines
> > > can be booted in UEFI mode, and if some modern machine cannot, then we
> > > need it fixed.  We should encourage users to make new installs boot by
> > > UEFI.
> > >
> > > The feature from the commit is only relevant for machines that require
> > > CSM boot or do not have UEFI option at all, am I right ? With the
> policy
> > > applied, an additional CSM-boot bootblock would be not shipped.
> > >
> >
> > I think it is far too early to say that the code for booting without
> > efi is abandonware. I have half a dozen x86 boxes in use here, and only
> > one of them is even able to boot efi, and its default resolution in efi
> > mode confuses the kvm switch it's connected to, so even on that I have
> > to use legacy bios boot.
> I do not propose to abandon bios boot, or even to declare it legacy
> with the proper meaning. I mean that CSM is disappearing on the newest
> platforms, and should become only used on old machines or i386.  With that
> attitude, adding a features for it, esp. by the cost of the user confusion,
> is not worth the efforts.  It still should be maintained for the foreseable
> future.
>
> If the machines where you get the trouble is newer than say 5 years,
> then they should boot with UEFI.  If not, the problem in loader.efi
> needs to be fixed.
>

There is no problem in loader.efi that's specific to geli. It already uses
the UEFI boot loader config. While there's some issues downstream (eg
kernel messages), this specific issue is a non-issue for loader.efi.

Warer