[syslinux] isohybrid --uefi - resulting EF00 partition too small

[Thomas Schmitt]

Hi,

> I did not yet look on the code of isohybrid.c, but it would seem to make
> sense to change the calculation to use a higher level library like e.g.
> libisofs to retrieve the correct size inforamtion.

libisofs is not very lean in terms of code size and runtime
memory consumption. It loads the whole ISO directory tree.
libcdio would probably be leaner at runtime.

But it would be more attractive then to let isohybrid.c traverse the
ISO directory tree by own means and to look for the start address
of the image file which was obtained from El Torito. El Torito
does not tell the name or path of the image file.

Nevertheless, the simpler solution would be to give isohybrid.c
an option by which you can override the questionable size info
from the El Torito catalog. (Quite everybody has more clue about
that size than El Torito has.)


> However, I think there are quite some people creating bootable ISOs sticking
> to genisoimage or mkisofs combined with isohybrid because it is proven and
> they do not want to change a running system and they fear breaking
> bootability especially with older buggy BIOSes.

isohybrid.c seems quite unmaintained currently. At least for the
GPT part. genisoimage is nearly unmaintained, too. I could help
with both, but will not take over responsibility for them.

xorriso offers a migration path by its mkisofs emulation.
There is not much room for individual solutions with the oldfashioned
ways of booting via ISO 9660. So in that aspect it should be equivalent
to mkisofs and genisoimage.
The only boot related mkisofs feature that xorriso cannot do for
Debian is option -hfs for old PowerPC machines. In exchange it can
offer -hfsplus for newer Macs (contribution of Vladimir Serbinenko).


> The thumb drive now correctly boots from BIOS and UEFI (at least up to
> kernel and initramfs on UEFI, trouble in userland should be my problem for
> now,

I live in happy ignorance towards the further boot stages.
All my tests are done with existing bootable images which i repack
like i did with yours yesterday. Regrettably i have no machine
that boots via EFI.


> trouble on ISO/MBR with not found com32 files in syslinux 5.10pre1 is
> another story).

If USB stick and CD show different bugs, then i cannot rule out a
xorriso bug. But on the other hand, the typical problem indication
is no booting at all. I just provide pointers to the first step.
Give me a note if particular suspicion arises towards xorriso.


Have a nice day :)

Thomas