Testing requested: Hybrid ISO/USB boot
benno at FreeBSD.org
Fri Mar 23 21:01:42 UTC 2018
> On Mar 22, 2018, at 1:48 PM, Thomas Schmitt <scdbackup at gmx.net> wrote:
> Benno Rice wrote:
>> I’ve been working on the ability to create hybrid ISO/HDD boot images for
>> x86, a la what Linux systems do with ISOHYBRID.
> Waving friendly over the fence i feel entitled to give some neighbor's
Hello! Thanks for your response.
> This does not look much like it addresses EFI.
> $ xorriso -indev hybrid-bootonly.iso -report_el_torito plain -report_system_area plain
> El Torito catalog : 19 1
> El Torito images : N Pltf B Emul Ld_seg Hdpt Ldsiz LBA
> El Torito boot img : 1 BIOS y none 0x0000 0x00 4 20
> El Torito boot img : 2 BIOS y none 0x0000 0x00 1600 24
> MBR partition table: N Status Type Start Blocks
> MBR partition : 1 0x80 0xa5 1 16
> UEFI 2.4, 22.214.171.124, says about El Torito:
> "A Platform Id of 0xEF indicates an EFI System Partition."
> But entry 2 of this catalog is in a section with Platform Id 0x00 (x86 BIOS).
> If byte 38977 (decimal) was 0xef rather than 0x00, then it would be marked
> as an El Torito boot image for EFI.
I think I’ve addressed this in this revision:
And I’ve regenerated the image here:
> Further: This section begins at byte address 38976 by a byte value 0x90.
> According to El Torito specs this means that another section follows.
> Correct would be value 0x91, which announces the end of the catalog.
> (El Torito 1.0, Figure 4, Offset 0)
I’ll look into this too.
> On HDD, EFI looks for specially marked partitions in GPT or MBR partition
> In MBR partiton table it looks for a partition of type 0xEF.
> (UEFI 2.4, 5.2.2)
> In GPT it looks for Type GUID C12A7328-F81F-11D2-BA4B-00A0C93EC93B.
> (UEFI 2.4, Table 19)
> So to mark the EFI partition, hybrid-bootonly.iso should have a MBR
> partition number 2 with type 0xEF, start at 512-byte block 24*2048/512 = 96,
> size 1600 blocks.
> (Strangely El Torito addresses by 2048-byte blocks but counts by 512-byte
> blocks. Size limit is 65535 blocks. But counts 0 and 1 mean to EFI
> "up to end of medium”.)
I’ll look in to this one too.
>> I’ve tested this image under qemu
> OVMF/SDK-II/Tianocore is too tolerant with the EFI specs and with silently
> using BIOS boot equipment. Real iron EFIs insist much more in compliance.
So it appears. Thanks again for looking at this!
More information about the freebsd-hackers