Intercept Pointers

Lenny -

The following is FYI..

It was stated in the review it would be useful to have a pointer from an
intercepted UCB to the intercept UCB.

The intercept I have designed places the DDT physically into the UCB of
the intercept driver. Thus apart from a bit of arithmetic, an intercepted
UCB$L_DDT DOES contain a pointer.

However, this pointer is not always unambiguous. If a second intercept
has occurred, UCB$L_DDT points to the second intercept, not the first.
My intercept defines a chain of these, so that an intecept driver can
follow a chain and find its own UCB by finding a unique magic number
(its internal DPTAB address) that will tell that it has found its own
UCB (presuming that a driver will not intercept anything more than once
in the same driver).

Thus from intercept code it is straightforward to locate the correct
intercept UCB, but because there may be multiple intercepts a single
backpointer will generally not be enough.

My intercept also defines a unique "magic number" 4 longwords ahead of
the DDT which could be used to tell other code that a driver was
intercepted. This approach has a small amount of risk though, since it is
not guaranteed that referring to 16 bytes ahead of the DDT is legal.

What would be better would be a UCB flag that could be set by an
intercept to flag that the UCB had in fact been intercepted somehow. If
my intercept scheme were used, with forward and back links between
intercepts, then finding intercept UCB addresses would be safe enough,
and intercepts could nest, be added, or removed, freely.

If your "new area" to be pointed at by UCBs is inserted, I'd suggest that
the first several longwords be reserved for flags. (I'd save at least
2 quadwords' worth to be on the safe side.) Then one of these could be
used. Since we're talking about disk UCBs here, I'm not too worried about
cloning or third party support. Given that a backpointer can't be unique
I think my present scheme is fairly reasonable. Unfortunately at the
review I forgot about that double intercept issue.
glenn everhart