Re: [Linux-ATM-General] PMP Drop Party sequencing

[chas w. - C. <ch...@cm...>]

In message <000801c77cac$7c7a56c0$6405a8c0@absolut>,"Russell McGuire" writes:
>In the ADDPARTY ioctl we can simply add calls by passing in a address, and a
>address. The ioctl then returns a reference number.
>
>Question 1: 
>Is the return value from ADDPARTY simply the number of leaves in the call?
>Or is the reference number returned a unique ID for this leaf? Should we
>keep track of this reference for later dropping the call?

i have asked the developer for a bit of documentation about this
interface but until then...  as i understand it, with addparty you are
returned the endpoint reference for the leaf.  it will start with
1, 0 being reserved for the initial destination of the p2mp.

>Question 2:
>When using the DROPPARTY IOCTL call I see we pass in a pointer to the call
>reference? Is this simply so the IOCTL can decrement it? Or again is the
>ioctl using that unique refernce ID to identify which leaf to drop?

you need to pass the root of the p2mp and the endpoint reference of
the leaf you want to drop.  we probably arent doing proper reference
counting on the leaves with regard to the parent socket.

>Question 3:
>Can we drop the original 'leaf' without killing the entire call, i.e. leaf
>0? 

i think so.  although given that we dont do reference counting for
the leaves this might not be the case.

>I know what the ATM switch can handle, but I am not sure on the limitations
>on the P2MP implementation for the atmsigd and kernel. I'll be testing this
>stuff, but if anyone can shed some light on these question, that would be
>great.

if something doesnt work, send mail to the list.  it can certainly be
fixed.