Strange booting problem

[Chris Murphy]

On Thu, Jul 2, 2015 at 1:27 PM, jd1008 <jd1008 at gmail.com> wrote:
>
>
> On 07/02/2015 12:02 PM, Chris Murphy wrote:
>>
>> On Thu, Jul 2, 2015 at 10:56 AM, Rick Stevens <ricks at alldigital.com>
>> wrote:
>>
>>> Simply nulling bytes 510 and 511 of the first sector (getting rid
>>> of the boot signature) should make the BIOS ignore the drive.
>>
>> That signature applies to the entire MBR, including the partition
>> scheme. The signature is not a "boot me" signature.
>>
>> GRUB probably shouldn't honor that MBR as valid, including its
>> partition scheme. I don't know if it does or not. The kernel won't,
>> and therefore won't see the partitions, and neither will libblkid.
>>
>>
>>
>>   fdisk
>>>
>>> will complain that the partition table is invalid because of the
>>> invalid boot signature, but that's all. The rest of the drive should
>>> be functional and usable--just not bootable.
>>
>> This is not correct.
>>
>>
>>> Tagging a partition as "bootable" only affects Microsoft OSes.
>>
>> This is not completely correct. It depends on what boot code is in the
>> MBR. GRUB boot.img doesn't use the active bit. But parted's code does,
>> as does syslinux/extlinux.
>>
>>
> What a convoluted frickin' mess!!

And it will always be that way, frozen in the time in which it was invented.

And hence there's UEFI. So try reading the UEFI spec. It's vastly more
complex, with tons more bugs in the implementations, but the spec
itself is not difficult to follow (although it's nearly 2300 pages
long so that might make it difficult, just not difficult to follow -
it's fairly well organized and hey at least it's documented unlike
BIOS).


-- 
Chris Murphy