A small self-contained patch set which centers around determining sufficient option
headroom when calculating the maximum packet size.
Patch #1: Adds documentation for the fields in struct ccid_operations.
Patch #2: More informative registration of CCIDs.
Patch #3: Introduces the CCMPS field for CCIDs. This is currently only used
by CCID4. The ccmps value needs to include network/transport header
lengths (since dccp_sync_mss subtracts these). For CCID4 this would
supposedly look like (RFC 3714/4828):
struct ccid_operations ccid4 = {
.ccid_mps = 1500,
/* ... */
};
If the field is omitted or set to 0, no CCMPS is imposed.
Patch #4: Is the core of this patch set and estimates the minimally required option
headroom for Data/DataAck packets. It addresses a FIXME in dccp_sync_mss.
I have long thought about this and come to the conclusion that a simple
solution - optimising for data transfer - is better. The complexities of
variable-length options (Ack Vector, Data Dropped, initial feature-negotiation)
need to use a different path - e.g. using a separate Ack (cf. next patch).
Rationale and background is documented on
http://www.erg.abdn.ac.uk/users/gerrit/dccp/notes/misc_notes/estimating_option_sizes_for_MPS.txt
Patch #5: Reorders statements to estimate the maximum room left for Ack Vectors and adds a Fixme + warning
message for dealing with insufficient remaining space when piggybacking on DataAck.
-
To unsubscribe from this list: send the line "unsubscribe dccp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
