l4-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: auth handshake and rendevouz objects


From: Marcus Brinkmann
Subject: Re: auth handshake and rendevouz objects
Date: Mon, 4 Nov 2002 16:57:01 +0100
User-agent: Mutt/1.4i

On Mon, Nov 04, 2002 at 03:59:28PM +0100, Bram Meersman wrote:
> Can I find an explanation about handle passing in Hurd/L4 somewhere on 
> the web ?

Here is what I recollect:

Server S provides object, client A wants to copy or move a handle to this
object to client B.

For this, A sends a message to S that says "insert a handle (send right) for B".
The server makes a note of that, but it doesn't actually create a real
reference.  Then A sends a message to B that says "I just gave you the
handle for the object".  Then B sends a message to the server S that says:
"I accept reception of the handle from A" [1].

Now, in the case of moving a handle from A to B, A can safely remove its
reference.  However, it must be careful not to do this too early on, before
B has acknowledge receiving the handle from A to the server S.  Because
otherwise there is a race that could lead to the object being destroyed
before B acknowledges its reception.  This means some synchronization
between B and A is necessary: B could send a message back to A that says: "I
got the handle and acknowledged it."  This could be implied by the reply
message to the RPC that sent the handle, or it could be a separate RPC...

There was another variation: In this case, the default operation in the
server would be to move a reference from A to B, and for a copy operation A
would first have to allocate an additional reference.  Both approaches were
mostly identical, but one had fewer problematic details (races and such),
and I don't remember the details which one it was for what reason.

I know I am vague.  Maybe Neal can fill in the details, or we can certainly
recover them by a bit of thinking.  It took us a couple of minutes to work
it out in full detail, so it's not too hard.

Thanks,
Marcus

[1] This is very important, to avoid leaking references: Otherwise a malicious
task could donate send rights to random long lived privileged tasks in the 
system
and thus prevent object destruction beyond its lifespan.




-- 
`Rhubarb is no Egyptian god.' GNU      http://www.gnu.org    address@hidden
Marcus Brinkmann              The Hurd http://www.gnu.org/software/hurd/
address@hidden
http://www.marcus-brinkmann.de/




reply via email to

[Prev in Thread] Current Thread [Next in Thread]