tech-userlevel archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: mounting wedges by name
On Sun, Oct 17, 2010 at 01:43:40PM +0200, Michael van Elst wrote:
> On Sun, Oct 17, 2010 at 01:14:08PM +0200, Martin Husemann wrote:
> > Looks good, but could we also have a UUID=... variant to refer to the
> > GUID always (instead of the name)?
> >
> > I know that this would require a new dkioctl, or extending struct
> > dkwedge_info.
>
> What would be the advantage of always allowing a UUID? If you want
> to use UUIDs you can do that by not naming the partitions.
>
> You already named the drawback, it is incompatible and unlikely
> suited for a pullup to older NetBSD releases.
Besides UUID and name, there are other volume properties that it would
be nice for mount(8)/fstab(5) to match on, and there will probably be
more properties in the future.
What if you allow properties to be named by dotted keywords, and reserve
the wedge. space for dkwedge_info?
uint8_t dkw_wname[128]; /* wedge name (Unicode, UTF-8):
* wedge.name
*/
char dkw_parent[16]; /* parent disk device name:
* wedge.parent
*/
daddr_t dkw_offset; /* LBA offset of wedge in parent:
* wedge.offset
*/
uint64_t dkw_size; /* size of wedge in blocks: wedge.size
*/
char dkw_ptype[32]; /* partition type string: wedge.type */
Let all other keywords indicate device properties (drvctl -p dk0).
E.g., for a GPT partition[1], let us flesh out the wedge properties
with information such as partition type GUID, unique partition GUID,
and partition name.
The wedge. space will work for -current, past, and future releases. All
other keywords will work for -current and future releases.
Dave
[1]
http://en.wikipedia.org/wiki/GUID_Partition_Table#Partition_entries_.28LBA_2.E2.80.9333.29
--
David Young OJC Technologies
dyoung%ojctech.com@localhost Urbana, IL * (217) 278-3933
Home |
Main Index |
Thread Index |
Old Index