[eros-arch] Problem with the semantics of address translation and keeper invocation

[Charles Landau]

Now that the EROS people have been alerted, I suggest keeping this 
thread on CapROS-devel, and dropping eros-arch, since EROS is 
officially dead. The eros-arch people can subscribe to CapROS-devel 
at http://lists.sourceforge.net/mailman/listinfo/capros-devel if they 
are interested.

At 12:13 AM -0400 9/3/05, Eric Northup wrote:
>On Fri, 2005-09-02 at 18:05, Norman Hardy wrote:
>>  On Sep 2, 2005, at 1:38 PM, Eric Northup wrote:
>[...]
>>  Keykos had another fix to this problem, with some advantages and 
>>  disadvantages.
>>  I quote from <http://www.agorics.com/Library/KeyKos/Gnosis/28.html>.
>
>That is a very useful document!
>
>Out of curiosity, what does "ssc" stand for?

Slot Size Code

>Am I correct interpreting
>it to be morally equivalent to EROS' biased-log segment size?

Yes. The SSC is the log, base 16, of the size of address space 
defined by each "slot" (element of the array of 16 keys) in a segment 
node. So if a node has SSC=5, its 16 keys define 16^5 * 16 = 2^24 
addresses.

>Also, error code 4 is "path passes thru node that belongs to the
>domain".  Does that mean what I think it does?

It means that one of the nodes in the path is also one of the three 
nodes comprising the domain.

>(Because what I think it
>means [is] that each and every time I find a tricky situation that looks like
>it might have been a bug in KeyKOS, it eventually turns out to have been
>anticipated and avoided by the KeyKOS team.)

Yes, we were very careful. It helps that we weren't operating under 
any sort of deadline.

>  > Perhaps the architecture should merely say for some fixed n that 
>>  "there is a limit of at least n levels on each portion".
>
>So, a valid translation path would translate at least one
>never-before-translated bit each n steps?

What I suggested in a separate message was that a valid translation 
path would translate at least *four* never-before-translated bits 
each n steps. That's equivalent, since bits are always translated 
four at a time, but it makes it clear that the bound on translating 
20 bits is 5*n.