On Tue, 16 Jan 2007, Darren J Moffat wrote:
W. Wayne Liauh wrote:
AFAIK this is what the project aims. An Extended
partition is
basically a container that can accommodate one or
more logical
partitions and we need the ability to have a
logical partition of type
Solaris2 (0xbf) and have OpenSolaris booting from
it.
This IMHO removes a barrier to OpenSolaris
adoption since people
already having multiple OSes would not be forced to
free up a Primary
partition to have OpenSolaris installed.
egards,
Moinak.
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
Another MAJOR advantage is that, from then on, Solaris will be able to
mount extended partitions. From my own experience, this is probably more
significant.
Pavan's project will _NOT_ make this happen.
To mount, you need either:
1. a device node (entry in /dev/dsk/....) for the "partition to
mount"
2. code in the filesystem driver to allow "subdividing" device
nodes into units that the disk target driver knows nothing
about (that's PCFS' "...:[c-z]" scheme)
Now since the project will _NOT_ deliver 1. and not change any
filesystem's behaviour wrt. to 2. what you talk about will not happen.
I.e. if you have, for example, a FreeBSD UFS filesystem in a non-primary
PC partition, Solaris UFS will not suddenly gain the ability to mount that
(even if the format were compatible enough). There will be no device nodes
for it.
What Pavan's project will allow is to put the solaris partition (what used
to be ID 0x83 just as linux swap, and now is ID 0xBF) somewhere else but
into one of the four MBR slots. Aka into (and I've had long flamewars
about what term to use, I'm not into agreement with everyone there on what
to call "extended" and what to call "logical") a "logical partition". For
Linuxers, into something Linux would number /dev/hda6 or above. Still,
you're only going to see p0...p4 on Solaris/x86 (and still none at all on
Solaris/sparc) as far as target driver device nodes are concerned.
It's a band-aid, to allow making Solaris installation in a multi-OS
environment easier. To reduce the barriers of entry.
Solving the problem of "generic partitioning support" goes beyond this.
The cmlb driver is a first step to that goal, abstract partitioning
support into a separate layer instead of burdening the target drivers (sd,
cmdk) with it. cmlb has been delivered, cmdk uses it and the work to
convert sd to use it as well is ongoing.
But to add device nodes where we have none today is more difficult,
because we run out of minor number space ...
You can't mount an extended partition you can mount a file system inside an
extended partition but not the partition itself.
On Solaris, you can't mount a filesystem inside a logical partition unless
that filesystem is FAT. Pavan's project will not change that. It will only
change the places on disk where the Solaris VTOC can reside, and the
Solaris "slices" (s0,,s7 device nodes) will be found after Pavan's project
even if the Solaris VTOC is _NOT_ in one of ...p[1-4].
If the extended partition contains a FAT file system then the pcfs filesystem
driver can do this already today (for the most part).
Where did _that_ fail for you ? Always interested to hear about
compatibility problems people encounter with PCFS.
FrankH.
==========================================================================
No good can come from selling your freedom, not for all gold of the world,
for the value of this heavenly gift exceeds that of any fortune on earth.
==========================================================================
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
