GEOM architecture and the (lack of) need for foot-shooting

[Poul-Henning Kamp]

In message <20050408040035.GA34844 at VARK.MIT.EDU>, David Schultz writes:

>So I don't have time to implement an API for disk partition layout
>changes.  But before I read your message, I was looking into
>getting sysinstall to issue a DIOCSMBR ioctl to the disk instead
>of trying to write it directly.  Would that suffice to allow users
>to safely modify their partition tables, while preserving the
>anti-foot-shooting mechanism?  Of course, when someone has time
>to implement a g_ctl() to do this as you suggest, it would be easy
>to adapt sysinstall at that time.

You need to implement whatever ioctl/gctl is used for all the
kinds of partitioning sysinstall/libdisk know about that means
at least mbr, bsdlabel and gpt.  Maybe also maclabels.

>FWIW, my complaint was not about what you did or didn't manage to
>accomplish single-handedly.  My complaint was that you turned on the
>anti-footshooting knob by default before userland was ready for it.

What you call "The anti-footshooting knob" is an integral part of
geoms event driven design, it is not a feature added specifically
to annoy people, it is a necessary behaviour for the proper discovery
of on-disk metadata.

>
>> 1. Find out which partition format we migrate to instead of BSDlabel
>>    which runs out of steam around 2TB.  GPT has been proposed but
>>    seems to be a rather dead end with Itanic sinking fast.
>
>Regardless of what happens to IA-64 and EFI, GPTs probably aren't
>going away.  Within a few years, Intel and Microsoft will be
>forced to address the 2TB limit for run-of-the-mill desktops, and
>I really can't see them inventing yet another format.  But those
>are just my two cents; you're right that this should be discussed
>further.

I will just add that the release our users will be running two years
from now will be 5.x and 6.x.  We _have_ to have this resolved for
7.x.  Somebody [TM] should sit down and read the lastest pc hardware
requirement from Microsoft to see what they say about it.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.