Date: Fri, 18 Jul 2008 23:34:02 +0200 (CEST) From: "Didrik Madheden" <didrik@kth.se> To: "Didrik Madheden" <didrik@kth.se> Cc: freebsd-sparc64@freebsd.org Subject: Re: Sparc64 partitions compatible with PC? Message-ID: <49814.91.95.8.243.1216416842.squirrel@webmail.sys.kth.se> In-Reply-To: <64373.91.95.8.243.1216323299.squirrel@webmail.sys.kth.se> References: <57797.91.95.8.243.1216313772.squirrel@webmail.sys.kth.se> <oqbq0w8ipv.fsf@castrovalva.Ivy.NET> <64373.91.95.8.243.1216323299.squirrel@webmail.sys.kth.se>
next in thread | previous in thread | raw e-mail | index | archive | help
I wish to apologoize for me previous mail. For some reason I pressed some key and managed to send it before I was done with it. Lousy custom university webmail. They're getting rid of it just now in favour for SquirrelMail... Not a day too early. Anuway... > I did disklabel -e ad0 on the Sparc and got this: > > # size offset > # ---------- ---------- > a: 20972448 0 > b: 2097648 41612 > c: 54757584 0 > d: 20972448 20806 > e: 737378208 43693 > This was the output disklabel -e ad0 (On the Sparc, running the existing installation of 6.2) This is where things are starting to get fishy... I noticed the file opened in vi, and to quit, I accidently did :wq instead of :q. The program then complained that the slices reached outside of the partition. (Or something to that effect; don't remember exactly now) Is that a warning sign? I also noticed there's a partition too much. From the sizes, I recognice a and d as my system partitions, b as my swap and e as my big storage drive. But what the heck is c? It's not the sume of the sizes of the other labels either. It just doesn't make sense from any angle. I guess I messed around in the label editor when oing the Sparc installation, but I don't see how I could've ended up with the final result as shown above. I'm currently in 7.0 installation mode on Mr Pee-C. What I've done so far is create a partition covering the whole disk. (DD mode, then no, which covered the same range as teh previously existing partition) I've tried recreating slices a, b, d and then pressing C to see what size the editor would suggest. What it suggested was 737380224, which is exactly 2016 blocks too long. So my dilemma now is this: Where does label really begin? *Is it perhaps so that the FreeBSD style partition used a little more overhead, so I should make the previos partitions 2016 blocks longer to compensate for the offset, and make the label the same length it used to be? *Or is it perhaps so that my PC thinks that the disk is 2 sectors/cylinders (Which unit?) well, 2 somethings longer than the Sparc, and that the spot i hit after creating the three other labels is in fact the right one. *Does it really matter? will FreeBSD perhaps compensate for the offset and still work? How fuzzy is it? And a question of importance! How to make sure the area I want to save doesn't get overwritten? I suppose T for toggle newfs=N will do the job, but I'm paranoid, so I want a second opinion. Anything more I need to do? >> BSD has the foundations of a better architecture. there are ioctls in >> bsd to load the in-core disklabel without touching the on-disk label. >> There are userland programs to read bsd labels, bsd slices, smd >> labels, and mbr labels in BSD, which could be built on non-native >> architectures, to read the label and load it into the kernel without >> touching the disk. There is a whole geom framework for doing things >> more complicated than simple disklabels. but in my own experience, >> the right command line flags to do what you want just don't exist. If I interpret this correctly: There's no good way to pop the disk into a box with an existing installation and type the magic command to mount blocks X to Y on the disk as an imaginary, ro file system? /Didrik Madheden
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?49814.91.95.8.243.1216416842.squirrel>