From: James W McMechan <mcmechanjw@ju...> - 2003-06-06 20:11:22
Nice, I was aiming for more information hiding between ubd_kern/user and
the cow stuff, I would prefer to have all the cow internal
data/sizes/offsets hidden from the ubd_kern/user functions, in the
layouts I was
practicing with I was using only open/read/write/close_COW and the
devinfo structure as the interface to allow for stacking the COW files.
Also I noticed I could get away with only data_length and the C/H/S info
from devinfo everything else can be opaque
In testing to fit out for ISAM in the future I am trying to make the
devinfo structure more opaque because ISAM does not map completely to the
You inspired me to try putting each of the versions in a separate header
cow_v.h that includes both the header layout and then
This stuffed almost all of the smarts into open_COW which takes the name
string complete with the commas so it can do the stacked COW files, it
needs the entire string to parse out the backing files. This makes
ubd_kern/user much simpler since much of the parsing is removed.
The ubd-many stuff can be done separately but works much better once
ubd_ken/user no longer need to keep track of all the internal cow stuff.
> This patch restructures the ubd driver so that the COW header code is
> separate files, which are shared with uml_moo. cow_user.c in the
> UML tree
> is cow.c in uml_moo. There is also a shared cow.h.
> There is also a private cow_sys.h, which defines some interfaces
> by the COW code which are different in uml_moo and UML.
> This is a first step towards getting COW V3 implemented.
> I also released a utilities tarball which contains the corresponding
> in uml_moo.
The best thing to hit the internet in years - Juno SpeedBand!
Surf the web up to FIVE TIMES FASTER!
Only $14.95/ month - visit http://www.juno.com to sign up today!