Re: USB Install Fails, Complains about CD-ROM

["Thomas Schmitt" <scdbackup@gmx.net>]

Hi,

Mike Kupfer wrote:
> there seems to be something odd going on with the live image ISOs.
> [...]
>   alto$ sudo dd if=debian-live-9.4.0-amd64-xfce.iso of=/dev/sdb bs=32K
>   [...]
>   1951465472 bytes (2.0 GB, 1.8 GiB) copied, 455.702 s, 4.3 MB/s
>   alto$ sudo isosize -x /dev/sdb
>   sector count: 952848, sector size: 2048
> But 952848 * 2048 = 1951432704, not 1951465472.

That's indeed undesirable.
Interestingly debian-live-9.3.0-amd64-xfce.iso shows no such difference.

It looks like 9.4 was padded up to a full multiple of 64 KiB, whereas
the isosize of 9.3 was already aligned to 64 KiB.

My guess for now is that the reason is the multi-session preparation
which xorriso does by default if used with its native command set.
IIRC (after 10 years) this preparation includes session alignment to
full 32 blocks - 64 KiB.
The producer of Debian Live ISOs, live-wrapper, is one of the very few
producers of bootable ISOs which uses that command set. The others use
the mkisofs emulation, which omits the multi-session preparation by
default.

I will have to investigate deeper and probably ask the live-wrapper
maintainer to disable the multi-session preparation by command
  -compliance "no_emul_toc"

Thanks for pointing this out.


> I believe the problem is, in
> fact, the USB stick.  And not just the first one that I had problems
> with.  It looks like there's something about that entire make/model
> (MOSDART 8GB).

Let's hope they are not some of those fake capacity sticks. They pretend
to have their nominal storage capacity, offer the promised number of
logical blocks, but map them to a smaller number of physical blocks.
This makes them look as if they take all data. But in the end some of
the earlier written physical blocks get overwritten by later written
blocks with a different logical block address.


Have a nice day :)

Thomas