Re: Test Proposal for [bug #46716] Protective MBR partition is not marked as bootable

[Alexander E. Patrakov]

28.12.2015 18:20, Thomas Schmitt пишет:
> Hi,
>
> i improved the test script by adding two more manipulation modes:
> - "mbr_only" produces an EFI compliant ISO without GPT, HFS+, APM.
> - "original" lets the xorriso arguments pass unchanged.
>
> The mode can be chosen by the environment variable
>    MKRESCUE_SED_MODE
> Default mode is "mjg": ESP in MBR+GPT+APM, HFS+ in APM.
>
> Other environment variables for parametrization are
>    MKRESCUE_SED_PARTNO   MKRESCUE_SED_PROTECTIVE   MKRESCUE_SED_DEBUG
>    MKRESCUE_SED_XORRISO  MKRESCUE_SED_XORRISO_ARGS
> Read the first half of the script for explanations.
>
> The script may now be stored anywhere. If it does not find its
> neighboring "../../xorriso/xorriso", then it uses "xorriso" by default.
> By MKRESCUE_SED_XORRISO one can address the xorriso binary explicitely.
>
> Download from
>
>    
> http://libburnia-project.org/browser/libisoburn/trunk/frontend/grub-mkrescue-sed.sh?format=txt
>
>
> If GNU xorriso -version says
>    Version timestamp :  2015.12.27.150821
> and one wants to test -DLibisofs_mjg_boot_for_grub2, then one should
> overwrite the file
>    xorriso-1.4.3/libisofs/system_area.c
> by the file version that is downloadable as
>    
> http://bazaar.launchpad.net/~libburnia-team/libisofs/scdbackup/download/1301/system_area.c-20081019135100-i5wqmczu54lzyabt-1/system_area.c
> and run "make".
> This keeps the results of mode "original" compliant with EFI
> even if -DLibisofs_mjg_boot_for_grub2 was in effect at compile time.
>
> The result of "original" still is influenced by the fact that
> -DLibisofs_mjg_boot_for_grub2 prevents writing of gap filling
> partition table entries.
>
> -----------------------------------------------------------------
>
> Examples:
>
>    wget 
> http://libburnia-project.org/browser/libisoburn/trunk/frontend/grub-mkrescue-sed.sh?format=txt
>    mv grub-mkrescue-sed.sh?format=txt grub-mkrescue-sed.sh
>    chmod u+x grub-mkrescue-sed.sh
>
>    export MKRESCUE_SED_XORRISO=$HOME/xorriso-1.4.3/xorriso/xorriso
>
>    export MKRESCUE_SED_MODE=mjg
>    export MKRESCUE_SED_PARTNO=2
>    grub-mkrescue -o output.iso minimal --xorriso=./grub-mkrescue-sed.sh
>    unset MKRESCUE_SED_PARTNO
>
>    export MKRESCUE_SED_MODE=mbr_only
>    export MKRESCUE_SED_PROTECTIVE=no
>    grub-mkrescue -o output.iso minimal --xorriso=./grub-mkrescue-sed.sh
>    unset MKRESCUE_SED_PROTECTIVE
>
>    export MKRESCUE_SED_MODE=original
>    grub-mkrescue -o output.iso minimal --xorriso=./grub-mkrescue-sed.sh
>
> Inspect results by:
>
>    $MKRESCUE_SED_XORRISO -hfsplus on -indev output.iso \
>                          -report_system_area plain -report_el_torito plain
>
> -----------------------------------------------------------------

I have tested this script with various xorriso builds.

I have three builds of xorriso:

1. "vanilla": with the modified system_area.c, but with no CPPFLAGS
2. "no-boot-dummy": with -DLibisofs_mjg_boot_for_grub2
3. "rich": with -DLibisofs_mjg_boot_for_grub2 
-DLibisofs_protective_msdos_plus_boot_dummY -DLibisofs_pmpbd_on_lba0

I have not tested anything with 
-DLibisofs_protective_msdos_plus_boot_dummY but without 
-DLibisofs_pmpbd_on_lba0, because my original layout (partitions 
occupying the same space) was based on exactly one now-invalid design 
consideration: ease of creation of valid geometry with dd.

I have four sets of grub-mkrescue options, with an appropriately patched 
xorriso in the PATH:

A. "plain": grub-mkrescue -o minimal-plain.iso minimal
B. "default": grub-mkrescue -o minimal-default.iso minimal 
--xorriso=/.../grub-mkrescue-sed.sh
C: "mjg-p2": MKRESCUE_SED_MODE=mjg MKRESCUE_SED_PARTNO=2 grub-mkrescue 
-o minimal-mjg-p2.iso minimal --xorriso=/.../grub-mkrescue-sed.sh
D: "mbr-only": MKRESCUE_SED_MODE=mbr_only MKRESCUE_SED_PROTECTIVE=no 
grub-mkrescue -o minimal-mjg-p2.iso minimal 
--xorriso=/.../grub-mkrescue-sed.sh

All combinations except 1B and 1C are, based on your prior emails, 
valid, so I have tested them on DG965SS from a USB flash drive.

Note: I have also applied Andrei's patch that replaces the boot code in 
the FAT filesystem. It was intended to allow older Macs to boot, but it 
also affects DG965SS in the 2B case.

So, here are the interesting results:

1A, 1D, 2A: not recognized by the motherboard at all. This is related to 
the fact that no MBR partitions are marked as bootable on these iso images.

2B: prints "this is not a bootable disk" message that is embedded in the 
new boot code. This is especially interesting because there is no 
partition marked as bootable, so my prior understanding of the BIOS 
logic is wrong.

All other images are bootable.

--
Alexander E. Patrakov