Re: [uml-devel] COW can now moo

[Jeff D. <jd...@ka...>]

jam...@ho... said:
> 1 Is it worth the effort to make uml_moo make the output files to
> sparse when copying? as it is I have to do a cp --sparse=always to get
> back to a sparse file afterwards

I'm not sure.  The filesystems I use are (or have been at some point) full,
so there's not much point in trying to sparse them.  Other people may be
using less full filesystems though.

> 2 Does anyone still use the fakehd features of the ubd driver they add
> to the complexity and I at least don't understand who is using them or
> how?  I was thinking of continuing to work on the partition code and
> would if they are unused:
>     ignore those sections and let them break :( 

That stuff is fairly specialized (it's only useful if you're trying to run
an installation which refuses to use ubd devices), but essential in that
situation.

It is messy code and I wrote it in a hurry and haven't tried cleaning it up.

My inclination is to try to tidy it up and separate it it from the rest of
the driver somehow so it doesn't confuse things too much.

At some point, I'm going to compile the SCSI subsystem into UML (using 
a ubd-type driver as the lowest level).  That would allow installations
to use a virtual SCSI disk instead of a ubd device which is faking an hd
device.  Then we could throw that code out.

And if the IDE subsystem has a similar layering, we could do the same thing
there.

				Jeff