Fwd: A request for unnested UFS implementation in MBR

[Manish Jain]

On 07/08/18 04:16, Polytropon wrote:
> On Sun, 8 Jul 2018 03:50:09 +0530, Manish Jain wrote:
>> MBR permits 3 primary partitions + 1 EBR slice.
>>
>> What I meant was one of the three partitions serving as FreeBSD / (UFS,
>> unnested), and any extra filesystems the user needs reside within the EBR.
>>
>> In other words, an option for UFS to be just like Ext2 in this respect.
> 
> Yes, that is possible if you simply format that slice. It
> does not matter if it's a DOS primary partition or a logical
> volume inside a DOS extended partition. What you basically
> do is that you get a 'c'-labeled BSD partition (where 'c'
> means "the whole thing", in this case, a partition.
> 
> Let's say you have the following construct:
> 
> 	da0 = { [da0s1] [da0s2] [da0s3] [da0s5 da0s6] }
> 
> We assume da0s1 - da0s3 are used for DOS, "Windows", maybe
> one Linux Ext2 data partition - 3 DOS primary partitions.
> And then we assume one DOS extended partition with two
> logical volumes in it. You can now use newfs to format
> them:
> 
> 	# newfs /dev/da0s5
> 	# newfs /dev/da0s6
> 
> This gives you /dev/da0s5c and /dev/da0s6c (where the 'c'
> can be omitted, and you can use /dev/da0s5 and /dev/da0s6
> instead), and you could for example do:
> 
> 	# mount -t ufs /dev/da0s5 /usr
> 	# mount -t ufs /dev/da0s6 /var
> 
> Entries in /etc/fstab will attach them at boot time.
> 
> However, there is a problem with booting, as I already said.
> FreeBSD requires an 'a'-labeled BSD partition to boot from,
> MBR or not (or its equivalent in GPT, freebsd-boot, to
> construct a "boot chain" from the freebsd-boot to a
> freebsd-ufs partition which holds /), and this partition
> will then be the / partition.
> 
> As you can see, this is not possible. You cannot have a
> 'a' partition when you can only create a 'c' partition
> on a bare slice (coverin the whole slice).
> 
> 
> 
> Resources:
> 
> https://www.freebsd.org/doc/handbook/disk-organization.html
> 
> http://www.wonkity.com/~wblock/docs/html/disksetup.html
> 
> https://wiki.pacbsd.org/index.php/Partitioning


Then how do the 'a' and 'c' partitions get accommodated in GPT ?

Poly, your thinking is wrong. The BSD schema served its purpose. Now, we 
should think about retiring it. Or at least, provide it as only an 
option to conform with past behaviour. But definitely not as a matter of 
forced usage.

Tx
MJ