f2bb1cae36
Submitted by: Maksim Yevmenkin <m_evmenkin@yahoo.com> Approved by: re@
44 lines
1.2 KiB
Plaintext
44 lines
1.2 KiB
Plaintext
$Id: TODO,v 1.1 2002/11/24 19:47:06 max Exp $
|
|
$FreeBSD$
|
|
|
|
FIXME/TODO list
|
|
|
|
0) Ping itself. Should L2CAP layer loopback data?
|
|
|
|
1) Locking/SMP
|
|
|
|
External code now uses ng_send_fn to inject data into Netgraph, so
|
|
it should be fine as long as Netgraph is SMP safe. Just need to
|
|
verify it.
|
|
|
|
2) Understand and implement L2CAP QoS
|
|
|
|
Will fix later. I only have CSR based hardware and it does not
|
|
support QoS.
|
|
|
|
3) Better functions to manage CIDs and command ident's.
|
|
|
|
Resource manager is not good because it uses MTX_DEF mutexes,
|
|
(i.e. could block/sleep)
|
|
|
|
4) Implement group channels (multicast)
|
|
|
|
Will fix later
|
|
|
|
5) Add bytes/packets counters and commands to get/reset them
|
|
|
|
Will fix later. What to count?
|
|
|
|
6) Better way to get information about channels
|
|
|
|
L2CAP can support about 65000 channels. Need define some good way
|
|
to get data from kernel to user space. For example if we need to pass
|
|
1K of information for every channel, then worst case is that we need
|
|
to pass 65Mbytes of data from kernel to user space. Not good.
|
|
|
|
7) Deal properly with "shutdown"s and hook "disconnect"s
|
|
|
|
For now we destroy all channels when upstream hook is disconnected.
|
|
Is there a better way to handle this?
|
|
|