Date: Sat, 21 Apr 2007 19:02:43 -0700 From: Marcel Moolenaar <xcllnt@mac.com> To: Ivan Voras <ivoras@fer.hr> Cc: freebsd-current@freebsd.org, freebsd-geom@freebsd.org Subject: Re: GPT as default? Message-ID: <A08237C6-0F3C-4C60-B5C0-8C5AD3998EC7@mac.com> In-Reply-To: <f0eauu$97s$1@sea.gmane.org> References: <f0am4t$mmk$1@sea.gmane.org> <86wt076k7u.fsf@dwp.des.no> <619464E1-1CB4-4CFC-9ECF-7FC90DC24A20@mac.com> <863b2u18hz.fsf@dwp.des.no> <4629C2FE.9030301@samsco.org> <91BC1CF6-6C72-4263-A99A-C24FC209586E@mac.com> <f0eauu$97s$1@sea.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 21, 2007, at 5:40 PM, Ivan Voras wrote: > Marcel Moolenaar wrote: >> On Apr 21, 2007, at 12:53 AM, Scott Long wrote: > >>> And as much as it pains me to say it, DES is right here ;-) >>> geom_gpt >>> needs to implement the appropriate verbs to allow apps to >>> instruct the >>> gpt instance to modify itself, instead of forcing apps to blindly >>> overwrite it. >> >> Those verbs exist. There's no poor design There's only a long time >> to get from A to B. > > In particular, what is missing? A (generic) tool that utilizes the verbs. > Users will need at least the possibility to add and remove partitions > while the system is running and some partitions are mounted. > Looking at > the source tree, it seems that the userland utility is missing? > (There's > gpt(8) but it's for the old geom_gpt, right?) gpt(8) does not use GEOM verbs. It was written before that time. That's what the complains are mostly about. I started to convert geom_gpt to use verbs, but that grew into g_part because it's not only GPT that needs verbs and I also saw that APM (The Apple partitioning scheme) needed to be supported, which, if I were to do that, would need exactly the same thing. So, I extended the scope of the work... > Subquestion: geom_gpt names partitions "xxxp0", "xxxp1", etc. so what > happens if one of them is deleted? Will the numbers shift on next > boot? No. The partition number is the index of the partition entry in the table. That typically does not change (though nothing prevents a tool to reorder the entries in the table or compact the table). > I think I'l also include out of the box geom_label support :) GPT (like APM) has labels, which we should propagate. It's not a priority for me, but I presume that it shouldn't be too difficult to interface with geom_label. > Alternative is to use GUIDs, like some Linuxes do, so we have entries > like /dev/gpt/38a52be4-9352-453e-af97-5c3b448652f0 I've thought about that too, but again it's not a priority. It takes me long enough as it is to adopt the new world order that other features have to wait :-) -- Marcel Moolenaar xcllnt@mac.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?A08237C6-0F3C-4C60-B5C0-8C5AD3998EC7>