[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#635968: installation-reports: Initial boot after installation failed

Ralf Jung wrote:
>> So far, I for one haven't seen a system which requires a
>> partition _table_ to boot, not to mention a bootable
>> partition in it.  BIOS merely loads first 512 bytes of
>> a disk into memory and jumps into that area, without
>> trying to interpret what's inside.  Unless you use some
>> recovery/diagnostic mode which is embedded into some
>> BIOSes/machines.
>Well, that was my understanding as well, but my laptop just proofed us wrong 
>;-) . Maybe this is related to the "mainboard firmware" actually being some EFI 
>or UEFI thingy with a BIOS compatibility layer. In any case, the Ubuntu 
>installer deals with this correctly, I never had any issues booting from HDD 
>before trying Debian as main OS (CDs are a different matter though).

The BIOS on all[1] IBM-PC-compatible systems validates the Master Boot
Record[2] for the MBR signature ($55, $AA) in bytes 510 and 511 in the
first block of the hard disk. This avoids jumping to random noise on an
uninitialized disk. Some[3] BIOS vendors also validate the presence of
at least one bootable partition by looking for the BOOT flag, located in
the first byte of each partition record, as an additional check.

It would seem prudent that the Debian Installer set this flag
appropriately.  --Don

[1] I know of no exceptions, for the reason stated. Even GPT includes a
legacy MBR.
[2] http://en.wikipedia.org/wiki/Master_boot_record
[3] I've encountered boot failures for this reason more than once, but
can't tell if the successful installs on various systems were due to
ignoring the BOOT flag or having one correctly set. I tend to partition
manually so haven't been concerned enough about any failures to research
a bug report.

Retro tagline:
On a clear disk you can seek forever.

Reply to: