Unlike C, in Perl printf() doesn't return the number of bytes written.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Actually generate the stack frame isolinux expects.  This is not the
optimal order, but it appears to be what both earlier and later
versions of isolinux expects...

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Use ebx to hold the upper half of the partition offset, not edx.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

When we call abort_check, we're not idle even though we are polling
the keyboard.  Handle that by calling reset_idle near the top.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

We cannot put bsHidden in initializated data, as it is set before we
run the global checksumming.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

At least one system has been identified which disables interrupts when
El Torito INT 13h is executed.  Thus, save/restore IF around INT 13h.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

It turns out we *can* determine if we have the extra partition offset
information after all, by looking at the value of the stack pointer.
This depends on the internals of the old isohdpfx code, but that's
really all we need to worry about.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

The partition information is pointed to by ds:si; although ds
will *usually* be 0 it isn't guaranteed (in particular, it may
very well be 0x40 or 0x60); therefore, move the partition parsing
up before the reset of %ds and %es.  As a side benefit, we get
to use push again...

Also, fix the ordering of the two halves of the GPT LBA!

Signed-off-by: H. Peter Anvin <hpa@zytor.com>