On Sat, Apr 16, 2011 at 2:44 PM, Anders Broman
<a.broman@xxxxxxxxxxxx> wrote:
Martin Mathieson skrev 2011-04-16 15:18:
On Sat, Apr 16, 2011 at 1:38 PM, Anders
Broman
<a.broman@xxxxxxxxxxxx>
wrote:
Martin Mathieson skrev
2011-04-16 14:03:
Hi Anders,
There is already most of the ROHC profile in
packet-pdcp-lte.c. I'd been meaning to separate it out
so that it could be used from other places (e.g. when I
create UMTS PDCP...)
Martin
I missed that, how to proceed? I think the ROHC code should
live in a separate file I have a trace
with Ethertype ROHC for instance.
Regards
Anders
ROHC should definitely be in a separate file, which I hadn't
got around to doing. My code was mostly written around 3
years ago, but to be honest hasn't been used much.
My implementation needs a few hints to decode the ROHC frame.
Here are the ROHC fields in packet-pdcp-lte.h - they are what
my company's decoder needs to have supplied to it.
|
/* RoHC settings */
|
| 83 |
gboolean rohc_compression;
|
| 84 |
unsigned short rohc_ip_version;
|
| 85 |
gboolean cid_inclusion_info;
|
| 86 |
gboolean large_cid_present;
|
| 87 |
enum rohc_mode mode;
|
| 88 |
gboolean rnd;
|
| 89 |
gboolean udp_checkum_present;
|
| 90 |
unsigned short profile; |
From a very quick look, I see that your implementation is
hard-wired to RTP/IPv4, and comments about maybe using
preferences or conversation info to work out other fields. My
preference would be to attach info similar to the above, and
let each user fill it in as they can (leaving the possibility
to override with preferences).
How about in dissect_rohc() look for private data and in the private
data supply a ROHC struct
similar to
#define ASN1_CTX_SIGNATURE 0x41435458 /* "ACTX" */
typedef struct _asn1_ctx_t {
guint32 signature;
:
with the needed decoder input in that way it can be called from
other places supplying the needed information
or if no struct is received use the preferences set. Some of the
data can be inferred it the IR/IR-DYN packets are seen
but that requires context tracking.
Could you supply the struct defenition, so it'll fit your data?
Regards
Anders
I haven't really looked into how easy it'd be to track the context and work out some of the ROHC fields.
Yes, the above is what is currently done with PDCP-LTE.
The pdcp_lte_info struct has information for both the PDCP frame and the ROHC frame, and the ROHC part was what I pasted into my earlier message in this thread. What I'd need to do is split them - fill in the PDCP half, and if there was ROHC present, fill that in and attach it to the PDCP frame. One MAC frame may contain many PDCP/ROHC frames, but rather than attach an array of structs (as I think UMTS RLC does), I just attach one struct and keep overwriting the contents just before the decoder is called.
LTE MAC and RLC also already do the same thing with structs for the context they need in order to decode. They either get the info from the log file format (in my case) or parsed from special headers that precede the PDU (see the UDP heuristic formats, but special ethertypes have also been used for MAC).
The first step is probably to work out if all of these context fields are really needed, and work out how you can start off by hard-coding them when your ethertype is seen, or maybe from some signalling that sets up the ROHC conversation.
Martin
___________________________________________________________________________
Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
Archives: http://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
