Re: UEFI PC?

2019-11-17 Thread Joshua Branson
Peter Wiehe  writes:

> Hello,
> I'm new to Hurd and this mailing list.
>
> Is Debian GNU/Hurd able to be installed on an UEFI PC?

Are you certain that you want to install the Hurd on bare hardware?  The
Hurd is not as stable as GNU/Linux.  Most Hurd developers run the Hurd
in qemu.  This is probably the easiest way to run the Hurd.

That being said, I would guess that the Hurd does not have UEFI
support...but I do not know for certain.  I do know of one Hurd
developer who managed to install Debian GNU/Hurd on a Thinkpad T60.

Those old Thinkpads are pretty awesome.  I'm using a T400, and it was
only $50 on ebay.  Also, as far as I know, you have to use the
proprietary BIOS to boot the Hurd.  Coreboot or libreboot currently do
not support booting the Hurd.

That being said, you can probably still boot the Hurd in newer
laptops...You can normally boot a UEFI laptop in BIOS compatibility
mode.

I hope that helps!

>
> (I searched the archives for UEFI but only found this thread:
> https://lists.debian.org/debian-hurd/2016/05/msg00033.html
> ...which doesn't answer the question.)
>
> Greetings
> Peter
>

--
Joshua Branson
Sent from Emacs and Gnus



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Peter Wiehe, le sam. 16 nov. 2019 16:41:06 +0100, a ecrit:
> Is Debian GNU/Hurd able to be installed on an UEFI PC?

Actually there is no technical reason why it shouldn't work: it's grub
which has to care about BIOS vs UEFI.

That being said, it seems that the hurd-i386 variant of the Debian
installation CDs can't be booted on UEFI. I don't know why, it's using
grub, like the i386 variant.

So technically that should be possible with a few fixes, but nobody did
the fixes yet, so without tinkering you will not get it to work this
way, and you have to configure your BIOS to boot in legacy mode.

Samuel



Re: UEFI PC?

2019-11-17 Thread Thomas Schmitt
Hi,

Samuel Thibault wrote:
> Actually there is no technical reason why it shouldn't work: it's grub
> which has to care about BIOS vs UEFI.
>
> That being said, it seems that the hurd-i386 variant of the Debian
> installation CDs can't be booted on UEFI. I don't know why, it's using
> grub, like the i386 variant.

The immediate cause is the lack of an El Torito boot image for UEFI and of
an EFI System Partition.
(Looking at debian-sid-hurd-i386-NETINST-1.iso, 2017 06 20 12 43 44 00.)

On an amd64 system i can get UEFI equipment in a grub-mkrescue ISO
by installing binary packages grub-efi-amd64-bin and/or grub-efi-ia32-bin.
This can be further combined with grub-pc-bin for PC-BIOS.
Then i run grub-mkrescue with a dummy payload directory instead of a
directory tree with operating system files:

  mkdir minimal
  grub-mkrescue -o output.iso minimal

The result should then have EFI System Partition and El Torito image in
personal union, containing start programs for i386 and/or amd64.
Further there should be bootable MBR code and El Torito image for PC-BIOS.
Partition table is GPT. Have a look at Guix ISOs which get built by
grub-mkrescue. (There is also HFS+/APM stuff for antique x86 Macs.)

The Volume Id of the Hurd ISO from june 2017 looks like debian-cd
  Volume Id: Debian sid h-i386 n
But it has GRUB equipment for PC-BIOS rather than the usual ISOLINUX.
Well, debian-cd is flexible.
If it's that what you are using, let's talk to Steve McIntyre when he is
done with converting Jigdo to SHA256.

Whatever, it will be not a big obstacle to develop a xorriso run if
you have all GRUB equipment ready for EFI. It may also be a good
opportunity to switch to grub-mkrescue for ISO generation (and my wrapper
script for making MBR partitions rather than GPT).

--

I cannot say whether there is nothing which the Mach kernel would want from
the firmware. Maybe ask Steve McIntyre whether he had to wait for special
Linux precautions before being able to boot amd64 via EFI (long ago).


Have a nice day :)

Thomas
Running with C source



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Thomas Schmitt, le dim. 17 nov. 2019 16:39:28 +0100, a ecrit:
> Samuel Thibault wrote:
> > That being said, it seems that the hurd-i386 variant of the Debian
> > installation CDs can't be booted on UEFI. I don't know why, it's using
> > grub, like the i386 variant.
> 
> The immediate cause is the lack of an El Torito boot image for UEFI and of
> an EFI System Partition.

Ok, so the build scripts definitely are different, but I don't see any
reason why.

Mmm, actually I had tested with the amd64 debian iso image, not the
i386. The i386 
http://cdimage.debian.org/cdimage/release/10.2.0/i386/iso-cd/debian-10.2.0-i386-netinst.iso
has the same issue, so that must be just a small thing in the debian-cd
package which builds images differently on amd64 and i386.

Samuel



Re: UEFI PC?

2019-11-17 Thread Thomas Schmitt
Hi,

Samuel Thibault wrote:
> Mmm, actually I had tested with the amd64 debian iso image , not the
> i386. The i386
http://cdimage.debian.org/cdimage/release/10.2.0/i386/iso-cd/debian-10.2.0-i386-netinst.iso
> has the same issue,

Aren't amd64 and i386 GNU/Linux systems ?
Those have ISOLINUX equipment for PC-BIOS and GRUB for EFI. CD and USB stick.

Don't we talk about ISOs like
  
https://cdimage.debian.org/cdimage/ports/current-hurd-i386/iso-cd/debian-hurd-2019-i386-NETINST-1.iso

It has a GRUB El Torito image for CD via PC-BIOS. No recognizable MBR code
for booting via PC-BIOS from USB stick.

Can it be that hurd-i386 is prepared by
  
https://sources.debian.org/src/debian-cd/3.1.27/tools/boot/bullseye/boot-hurd-common
whereas i386 and amd64 are prepared by
  https://sources.debian.org/src/debian-cd/3.1.27/tools/boot/bullseye/boot-x86
?

Both tinker with xorrisofs boot options. boot-x86 is much more elaborate
than boot-hurd-common. In both it seems that ISOLINUX and GRUB binaries
are already prepared. Dunno where this is done.
The hurd script shows no traces of xorrisofs options for EFI, which would
be --efi-boot or -e. No option --grub2-mbr is to see for booting via PC-BIOS
from USB stick.
As said, it uses GRUB for PC-BIOS, whereas i386/amd64 use ISOLINUX.

(In case you plan to join boot-x86 with EFI:
 I have a change proposal pending for i386 and amd64 images. Motivated in
 detail at:
   https://lists.debian.org/debian-cd/2019/07/msg7.html
 It would be worth to adopt it if you copy all the boot-x86 stuff.
 I should prod Steve again. His excuse about the old xorriso version is
 gone meanwhile. The i386 ISO has
   Preparer Id  : XORRISO-1.5.0 2018.09.15.133001, ...
)

---

xorriso offers commands for inspection of boot equipment.
E.g.

  xorriso -indev debian-hurd-2019-i386-NETINST-1.iso \
  -report_system_area plain \
  -report_el_torito plain
yields
  ...
  Volume id: 'Debian 2019 h-i386 n'
  System area options: 0x0201
  System area summary: MBR protective-msdos-label cyl-align-off
  ISO image size/512 : 457764
  Partition offset   : 0
  MBR heads per cyl  : 64
  MBR secs per head  : 32
  MBR partition table:   N Status  TypeStart   Blocks
  MBR partition  :   1   0x80  0xcd1   457763
  El Torito catalog  : 804  1
  El Torito cat path : /boot/boot.cat
  El Torito images   :   N  Pltf  B   Emul  Ld_seg  Hdpt  Ldsiz LBA
  El Torito boot img :   1  BIOS  y   none  0x  0x00  4   53448
  El Torito img path :   1  /boot/grub/grub_eltorito
  El Torito img opts :   1  boot-info-table

With i386 GNU/Linux:

  Volume id: 'Debian 10.2.0 i386 n'
  System area options: 0x0102
  System area summary: MBR isohybrid cyl-align-on GPT APM
  ISO image size/512 : 872448
  Partition offset   : 0
  MBR heads per cyl  : 64
  MBR secs per head  : 32
  MBR partition table:   N Status  TypeStart   Blocks
  MBR partition  :   1   0x80  0x000   872448
  MBR partition  :   2   0x00  0xef 3856 4416
  MBR partition path :   2  /boot/grub/efi.img
  ...
  ... invalid GPT and useless APM ...
  ...
  El Torito catalog  : 963  1
  El Torito cat path : /isolinux/boot.cat
  El Torito images   :   N  Pltf  B   Emul  Ld_seg  Hdpt  Ldsiz LBA
  El Torito boot img :   1  BIOS  y   none  0x  0x00  42068
  El Torito boot img :   2  UEFI  y   none  0x  0x00   4416 964
  El Torito img path :   1  /isolinux/isolinux.bin
  El Torito img opts :   1  boot-info-table isohybrid-suitable
  El Torito img path :   2  /boot/grub/efi.img

With Guix (made by grub-mkrescue):

  Volume id: 'GUIXSD_IMAGE'
  System area options: 0x4201
  System area summary: MBR protective-msdos-label grub2-mbr cyl-align-off GPT 
APM
  ISO image size/512 : 2109436
  Partition offset   : 0
  MBR heads per cyl  : 65
  MBR secs per head  : 32
  MBR partition table:   N Status  TypeStart   Blocks
  MBR partition  :   1   0x00  0xee1  2109435
  GPT:   N  Info
  GPT disk GUID  :  38bec621914e604ea91884e71f367b5c
  GPT entry array:  20  176  separated
  GPT lba range  :  64  2109390  2109435
  GPT partition name :   1  4700610070003000
  GPT partname local :   1  Gap0
  GPT partition GUID :   1  38bec621914e604ea91984e71f367b5c
  GPT type GUID  :   1  a2a0d0ebe5b9334487c068b6b72699c7
  GPT partition flags:   1  0x1001
  GPT start and size :   1  64  34424
  GPT partition name :   2  
450046004900200062006f006f007400200070006100720074006900740069006f006e00
  GPT partname local :   2  EFI boot partition
  GPT partition GUID :   2  38bec621914e604ea91a84e71f367b5c
  GPT type GUID  :   2  28732ac11ff8d211ba4b00a0c93ec93b
  GPT partition flags:   2  0x1001
  GPT start and size :   2  34488  2880
  GPT partition path :   2  /efi.img
  GPT 

Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Thomas Schmitt, le dim. 17 nov. 2019 18:35:34 +0100, a ecrit:
> Samuel Thibault wrote:
> > Mmm, actually I had tested with the amd64 debian iso image , not the
> > i386. The i386
> http://cdimage.debian.org/cdimage/release/10.2.0/i386/iso-cd/debian-10.2.0-i386-netinst.iso
> > has the same issue,
> 
> Aren't amd64 and i386 GNU/Linux systems ?

They are

> Those have ISOLINUX equipment for PC-BIOS and GRUB for EFI. CD and USB stick.

Yes, but they use grub for EFI booting.

> Don't we talk about ISOs like
>   
> https://cdimage.debian.org/cdimage/ports/current-hurd-i386/iso-cd/debian-hurd-2019-i386-NETINST-1.iso

Depends what exactly we are talking about. The original request is for
such image, yes, but apparently scripts building it don't have EFI
enabled indeed, I'm having a look.

Samuel



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Thomas Schmitt, le dim. 17 nov. 2019 18:35:34 +0100, a ecrit:
> Can it be that hurd-i386 is prepared by
>   
> https://sources.debian.org/src/debian-cd/3.1.27/tools/boot/bullseye/boot-hurd-common

It is. I have now pushed EFI code there.

I have regenerated 
http://people.debian.org/~sthibault/hurd-i386/installer/cdimages/daily/debian-sid-hurd-i386-NETINST-1.iso

>   xorriso -indev debian-hurd-2019-i386-NETINST-1.iso \
>   -report_system_area plain \
>   -report_el_torito plain

now yields

Volume id: 'Debian sid h-i386 n'
System area options: 0x0201
System area summary: MBR protective-msdos-label cyl-align-off
ISO image size/512 : 366800
Partition offset   : 0
MBR heads per cyl  : 64
MBR secs per head  : 32
MBR partition table:   N Status  TypeStart   Blocks
MBR partition  :   1   0x80  0xcd1   366799
El Torito catalog  : 703  1
El Torito cat path : /boot/boot.cat
El Torito images   :   N  Pltf  B   Emul  Ld_seg  Hdpt  Ldsiz LBA
El Torito boot img :   1  BIOS  y   none  0x  0x00  4   36581
El Torito boot img :   2  UEFI  y   none  0x  0x00608   53849
El Torito img path :   1  /boot/grub/grub_eltorito
El Torito img opts :   1  boot-info-table
El Torito img path :   2  /boot/grub/efi.img

So there is efi stuff in there, and efi.img does contain bootia32.efi,
but
kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd

doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
a grub-ia32 issue (since I can't boot a linux-i386 image either), do you
have an idea at this point?

Samuel



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Samuel Thibault, le dim. 17 nov. 2019 18:58:08 +0100, a ecrit:
> Thomas Schmitt, le dim. 17 nov. 2019 18:35:34 +0100, a ecrit:
> > Can it be that hurd-i386 is prepared by
> >   
> > https://sources.debian.org/src/debian-cd/3.1.27/tools/boot/bullseye/boot-hurd-common
> 
> It is. I have now pushed EFI code there.

Ah, no, the push didn't work, you can read it from

https://salsa.debian.org/sthibault/debian-cd/blob/master/tools/boot/bullseye/boot-hurd-common

Samuel



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Samuel Thibault, le dim. 17 nov. 2019 18:58:08 +0100, a ecrit:
> kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd
> 
> doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
> a grub-ia32 issue (since I can't boot a linux-i386 image either),

Mmm, OVMF.fd is a 64bit firmware, perhaps it doesn't support loading
32bit binaries?

Samuel



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Samuel Thibault, le dim. 17 nov. 2019 19:08:20 +0100, a ecrit:
> Samuel Thibault, le dim. 17 nov. 2019 18:58:08 +0100, a ecrit:
> > kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd
> > 
> > doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
> > a grub-ia32 issue (since I can't boot a linux-i386 image either),
> 
> Mmm, OVMF.fd is a 64bit firmware, perhaps it doesn't support loading
> 32bit binaries?

My laptop doesn't seem to be able to boot it in EFI mode either.

But legacy mode boots fine.

Samuel



Processing of gnumach_1.8+git20191117-1_source.changes

2019-11-17 Thread Debian FTP Masters
gnumach_1.8+git20191117-1_source.changes uploaded successfully to localhost
along with the files:
  gnumach_1.8+git20191117-1.dsc
  gnumach_1.8+git20191117.orig.tar.xz
  gnumach_1.8+git20191117-1.debian.tar.xz

Greetings,

Your Debian queue daemon (running on host usper.debian.org)



gnumach_1.8+git20191117-1_source.changes ACCEPTED into unstable

2019-11-17 Thread Debian FTP Masters



Accepted:

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512

Format: 1.8
Date: Sun, 17 Nov 2019 18:33:09 +
Source: gnumach
Binary: gnumach-common gnumach-dev gnumach-image-1-486 gnumach-image-1.8-486 
gnumach-image-1.8-486-dbg kernel-image-1.8-486-di
Built-For-Profiles: noudeb noprof noxen
Architecture: source
Version: 2:1.8+git20191117-1
Distribution: unstable
Urgency: medium
Maintainer: GNU Hurd Maintainers 
Changed-By: Samuel Thibault 
Description:
 gnumach-common - GNU version of the Mach microkernel, common files.
 gnumach-dev - GNU version of the Mach microkernel
 gnumach-image-1-486 - GNU version of the Mach microkernel
 gnumach-image-1.8-486 - GNU version of the Mach microkernel
 gnumach-image-1.8-486-dbg - GNU version of the Mach microkernel for debugging
 kernel-image-1.8-486-di - GNU version of the Mach microkernel for the Debian 
installer (udeb)
Changes:
 gnumach (2:1.8+git20191117-1) unstable; urgency=medium
 .
   * New upstream snapshot.
 - git-xen-fix, git-spl-squash, git-picmask, git-interrupt-spl7,
   git-intpri, git-disable-irq: Merged.
   * patches/70_dde.patch: Update debugging prints.
Checksums-Sha1:
 bb24e84869c0c79fc732f1d73e60117f5301ba61 3074 gnumach_1.8+git20191117-1.dsc
 6cd5f7c7853ecc6219ea2cb3a46812e23f16679a 2792344 
gnumach_1.8+git20191117.orig.tar.xz
 6ba812b480518b0a15cec0a35f6245ca74d3f457 30684 
gnumach_1.8+git20191117-1.debian.tar.xz
Checksums-Sha256:
 812edd8a0897b13a8bd3590630bb1a7dcd7ea0cb4acda5606adc5d2acb623e93 3074 
gnumach_1.8+git20191117-1.dsc
 e28bd735248155ce4010c1eb8f84d49e38d088605a47cbe31b0c7458ba8a 2792344 
gnumach_1.8+git20191117.orig.tar.xz
 dc60ed9433dfb4505fb62d15730479ba8014d8b60192de447a13e2b1f4fdeccc 30684 
gnumach_1.8+git20191117-1.debian.tar.xz
Files:
 da20fddb38b417978fa24b70e7e12319 3074 kernel optional 
gnumach_1.8+git20191117-1.dsc
 39e5b987a8645a96db4b8d25c7e26655 2792344 kernel optional 
gnumach_1.8+git20191117.orig.tar.xz
 22ef7ccec8039658e16bb2b3bf3b0170 30684 kernel optional 
gnumach_1.8+git20191117-1.debian.tar.xz

-BEGIN PGP SIGNATURE-
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=F4v5
-END PGP SIGNATURE-


Thank you for your contribution to Debian.



Re: UEFI PC?

2019-11-17 Thread Thomas Schmitt
Hi,

Samuel Thibault wrote:
> kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd
> doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
> a grub-ia32 issue (since I can't boot a linux-i386 image either)

Indeed, this gives a "UEFI Interactive Shell" prompt and no Debian menu

  qemu-system-x86_64 -enable-kvm -m 512 \
   -cdrom debian-10.2.0-i386-netinst.iso \
   -bios /usr/share/ovmf/OVMF.fd

But it boots to a GRUB/Debian menu with

  qemu-system-i386 ...


> http://people.debian.org/~sthibault/hurd-i386/installer/cdimages/daily/debian-sid-hurd-i386-NETINST-1.iso

(Find the surplus "s" in this URL.)

  qemu-system-i386 ... debian-sid-hurd-i386-NETINST-1.iso ...

does not boot, i fear.

The most obvious difference is in the two EFI System Partitions:

  $ cd /mnt/fat_hurd
  $ find .
  .
  ./efi
  ./efi/boot
  ./efi/boot/bootia32.efi

  $ cd /mnt/fat_i386
  $ find .
  .
  ./efi
  ./efi/boot
  ./efi/boot/bootia32.efi
  ./efi/boot/grubia32.efi
  ./efi/debian
  ./efi/debian/grub.cfg

bootia32.efi in hurd has 278,528 bytes.
In i386 it's 1,065,120 for bootia32.efi, which says "Microsoft" 53 times,
and 1,156,464 for grubia32.efi.

The file grub.cfg contains only the instruction to use the grub.cfg in
the ISO filesystem:
  search --file --set=root /.disk/info
  set prefix=($root)/boot/grub
  source $prefix/i386-efi/grub.cfg


(In the hurd ISO, the EFI image is not marked in the partition table.
 Thus according to specs no booting from USB stick. But OVMF would boot
 by -hda nevertheless if it would work by -cdrom.)


Have a nice day :)

Thomas



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Thomas Schmitt, le dim. 17 nov. 2019 20:42:37 +0100, a ecrit:
> The most obvious difference is in the two EFI System Partitions:
> 
>   $ cd /mnt/fat_hurd
>   $ find .
>   .
>   ./efi
>   ./efi/boot
>   ./efi/boot/bootia32.efi
> 
>   $ cd /mnt/fat_i386
>   $ find .
>   .
>   ./efi
>   ./efi/boot
>   ./efi/boot/bootia32.efi
>   ./efi/boot/grubia32.efi
>   ./efi/debian
>   ./efi/debian/grub.cfg

Yes, but that should only be needed for signed EFI booting (at least
that's what the code comments say).

> bootia32.efi in hurd has 278,528 bytes.
> In i386 it's 1,065,120 for bootia32.efi, which says "Microsoft" 53 times,

Yes, that's the signed shim. But we should be able to boot in EFI mode
without a signature.

> The file grub.cfg contains only the instruction to use the grub.cfg in
> the ISO filesystem:
>   search --file --set=root /.disk/info
>   set prefix=($root)/boot/grub
>   source $prefix/i386-efi/grub.cfg

Yes, that is expected, to make the scripts simpler.

Samuel



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Samuel Thibault, le dim. 17 nov. 2019 19:08:20 +0100, a ecrit:
> Samuel Thibault, le dim. 17 nov. 2019 18:58:08 +0100, a ecrit:
> > kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd
> > 
> > doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
> > a grub-ia32 issue (since I can't boot a linux-i386 image either),
> 
> Mmm, OVMF.fd is a 64bit firmware, perhaps it doesn't support loading
> 32bit binaries?

Indeed, a 32bit version of OVMF is needed. I could grab one, and grub
boots fine. gnumach however doesn't. I guess that could be because in
EFI mode it won't find a BIOS-e820 table, so that's still a TODO, but
apparently at least the debian-installer boot part is now supported.

Samuel



Re: UEFI PC?

2019-11-17 Thread Thomas Schmitt
Hi,

Samuel Thibault wrote:
> Indeed, a 32bit version of OVMF is needed. I could grab one,

This explains why i could not repeat my success with qemu-system-i386.
(Now i wonder what i did to fall victim to that illusion.)


> grub boots fine. gnumach however doesn't.

Good luck with the rest of the journey. :))


Have a nice day :)

Thomas



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Samuel Thibault, le dim. 17 nov. 2019 21:31:50 +0100, a ecrit:
> Samuel Thibault, le dim. 17 nov. 2019 19:08:20 +0100, a ecrit:
> > Samuel Thibault, le dim. 17 nov. 2019 18:58:08 +0100, a ecrit:
> > > kvm -cdrom debian-sid-hurd-i386-NETINST-1.iso -bios OVMF.fd
> > > 
> > > doesn't manage to boot it. Perhaps that's a kvm issue, or perhaps that's
> > > a grub-ia32 issue (since I can't boot a linux-i386 image either),
> > 
> > Mmm, OVMF.fd is a 64bit firmware, perhaps it doesn't support loading
> > 32bit binaries?
> 
> Indeed, a 32bit version of OVMF is needed. I could grab one,

(FTR, from https://www.kraxel.org/repos/jenkins/edk2/ ,
https://www.kraxel.org/repos/jenkins/edk2/edk2.git-ovmf-ia32-0-20191016.1286.gc9af866cdd.noarch.rpm
, the usr/share/edk2.git/ovmf-ia32/OVMF-pure-efi.fd file)

Samuel



Re: UEFI PC?

2019-11-17 Thread Steve McIntyre
Thomas Schmitt wrote:
>
>On an amd64 system i can get UEFI equipment in a grub-mkrescue ISO
>by installing binary packages grub-efi-amd64-bin and/or grub-efi-ia32-bin.
>This can be further combined with grub-pc-bin for PC-BIOS.
>Then i run grub-mkrescue with a dummy payload directory instead of a
>directory tree with operating system files:
>
>  mkdir minimal
>  grub-mkrescue -o output.iso minimal
>
>The result should then have EFI System Partition and El Torito image in
>personal union, containing start programs for i386 and/or amd64.
>Further there should be bootable MBR code and El Torito image for PC-BIOS.
>Partition table is GPT. Have a look at Guix ISOs which get built by
>grub-mkrescue. (There is also HFS+/APM stuff for antique x86 Macs.)
>
>The Volume Id of the Hurd ISO from june 2017 looks like debian-cd
>  Volume Id: Debian sid h-i386 n
>But it has GRUB equipment for PC-BIOS rather than the usual ISOLINUX.
>Well, debian-cd is flexible.
>If it's that what you are using, let's talk to Steve McIntyre when he is
>done with converting Jigdo to SHA256.
>
>Whatever, it will be not a big obstacle to develop a xorriso run if
>you have all GRUB equipment ready for EFI. It may also be a good
>opportunity to switch to grub-mkrescue for ISO generation (and my wrapper
>script for making MBR partitions rather than GPT).
>
>--
>
>I cannot say whether there is nothing which the Mach kernel would want from
>the firmware. Maybe ask Steve McIntyre whether he had to wait for special
>Linux precautions before being able to boot amd64 via EFI (long ago).

There's a few pieces to think about here, from reading the thread:

 1. adding UEFI support in debian-cd for the hurd-i386 CD
build. That's easy

 2. If you're going to test-boot that in a qemu VM, you'll need a
32-bit OVMF build to test it too. Thet 64-bit OVMF in the archive
will only boot a 64-bit kernel. I have a 32-bit binary build I use
myself for local testing in git at
https://git.einval.com/cgi-bin/gitweb.cgi?p=efitest.git
in case that helps

 3. Finally (the elephant in the room) - booting via UEFI is not the
same as BIOS boot. The entry points are different (for a start),
and there are quite a lot og kernel changes needed to support UEFI
initialisation etc. I've not heard anything about people working
on this for Hurd at all.

-- 
Steve McIntyre, Cambridge, UK.st...@einval.com
  Armed with "Valor": "Centurion" represents quality of Discipline,
  Honor, Integrity and Loyalty. Now you don't have to be a Caesar to
  concord the digital world while feeling safe and proud.



Re: UEFI PC?

2019-11-17 Thread Samuel Thibault
Steve McIntyre, le dim. 17 nov. 2019 20:36:13 +, a ecrit:
>  1. adding UEFI support in debian-cd for the hurd-i386 CD
> build. That's easy

Yes, that's now on
https://salsa.debian.org/images-team/debian-cd/merge_requests/5

>  3. Finally (the elephant in the room) - booting via UEFI is not the
> same as BIOS boot.

gnumach uses multiboot, so it should be quite similar actually.

> to support UEFI initialisation etc.

What I can think of is memory layout initialization and PCI BIOS
support, do you think of anything else?

Samuel



Re: UEFI PC?

2019-11-17 Thread Peter Wiehe
Am 17. November 2019 22:14:05 MEZ schrieb Samuel Thibault 
:
>Steve McIntyre, le dim. 17 nov. 2019 20:36:13 +, a ecrit:
>> to support UEFI initialisation etc.
>
>What I can think of is memory layout initialization and PCI BIOS
>support, do you think of anything else?
>

I don't know if this belongs here, but speaking about UEFI init: Afaik there' 
s a watchdog timer which needs to be turned off shortly after startup or else 
your shiny OS will reboot after some seconds. I don't know if this is relevant 
for grub boot.


Peter



Re: UEFI PC?

2019-11-17 Thread Steve McIntyre
On Sun, Nov 17, 2019 at 10:14:05PM +0100, Samuel Thibault wrote:
>Steve McIntyre, le dim. 17 nov. 2019 20:36:13 +, a ecrit:
>>  1. adding UEFI support in debian-cd for the hurd-i386 CD
>> build. That's easy
>
>Yes, that's now on
>https://salsa.debian.org/images-team/debian-cd/merge_requests/5
>
>>  3. Finally (the elephant in the room) - booting via UEFI is not the
>> same as BIOS boot.
>
>gnumach uses multiboot, so it should be quite similar actually.
>
>> to support UEFI initialisation etc.
>
>What I can think of is memory layout initialization and PCI BIOS
>support, do you think of anything else?

I'm not the expert here, but off the top of my head you'll also need
to deal with UEFI boot-time and/or runtime services, and you'll
probably need to make your initial boot binary be a PE UEFI
executable. I'm sure there's probably more - I'd look at the Linux
UEFI boot path.

-- 
Steve McIntyre, Cambridge, UK.st...@einval.com
There's no sensation to compare with this
Suspended animation, A state of bliss



Re: UEFI PC?

2019-11-17 Thread Peter Wiehe
Am 17. November 2019 23:09:18 MEZ schrieb Steve McIntyre :
>you'll
>probably need to make your initial boot binary be a PE UEFI
>executable.

No, afaik that's not neccessary if you boot from an intermediate bootloader 
like grub.

Peter