On Jan 21, 2010, at 11:59 AM, Brian Oleksa wrote:
> But how I start the initial counting process is I do the following:
>
> guint8 * ptr = (guint8*) tvb->real_data;
Don't do that.
First of all, that isn't guaranteed to work for all tvbuffs; we currently aren't using composite tvbuffs (and there are apparently some bugs in them), but, if we do, there is no guarantee that the "real_data" field of a tvbuff will always be valid.
Second of all, you should not just extract fields from the packet data yourself:
doing so means that no bounds checking is done, so you might run past the end of the packet data (do *NOT* assume either that all packets for your protocol will be valid or that the capture wasn't cut short by setting the snapshot length when capturing);
doing so means that if a field isn't aligned on an appropriate boundary in memory, attempting to fetch the data could fail (SPARC processors, for example, do not support unaligned accesses, and we *do* support Wireshark on SPARC);
doing so means means that you have to do the byte swapping yourself.
Instead, use the tvb_get_ routines to fetch fields individually, using the offset variable.
Speaking of byte swapping, this:
https://www.darkcornersoftware.com/confluence/display/open/Packet+Structure
says "All values are in network byte order", so if you're running on a machine with the most popular family of desktop and notebook processors - i.e., an x86 or x86-64 processor - you *would* have to byte-swap values if you fetch them yourself. That also means that the tvb_get_ntoh routines should be used to fetch numerical field values.
> Actually..... maybe you can see your answer better in the code. I have attached the packet-helen.c file.
Please don't use hf_helen_length for anything other than an actual length field. Each field in the packet should have its *own* hf_ value.
Once you've started using the tvb_get_ routines, and the offset variable, to process the fields, and have given each field its own hf_ value, then:
The hf_ item for the time stamp should be something like
{ &hf_helen_time,
{ "Time", "helen.time", FT_ABSOLUTE_TIME, BASE_NONE, NULL, 0x0, "Time", HFILL}},
If you want to display the times as UTC, then the way you'd do this depends on the version of Wireshark you're using.
Wireshark 1.0.x or 1.2.x:
You would add that field to the packet by doing something such as
nstime_t t;
guint64 msecs_since_the_epoch;
struct tm *tmp;
static const char *mon_names[12] = {
"Jan",
"Feb",
"Mar",
"Apr",
"May",
"Jun",
"Jul",
"Aug",
"Sep",
"Oct",
"Nov",
"Dec"
};
msecs_since_the_epoch = tvb_get_ntoh64(tvb, offset);
t.secs = msecs_since_the_epoch/1000;
t.nsecs = (msecs_since_the_epoch%1000)*1000000; /* milliseconds to nanoseconds */
tmp = gmtime(&t.secs);
proto_tree_add_time_format_value(helen_sub_tree, hf_helen_time, tvb, offset, 8, &t,
"%s %2d, %d %02d:%02d:%02d.%09ld UTC",
mon_names[tmp->tm_mon],
tmp->tm_mday,
tmp->tm_year + 1900,
tmp->tm_hour,
tmp->tm_min,
tmp->tm_sec,
(long)t.nsecs);
Current top-of-tree Wireshark:
You would have the hf_ item be something like
{ &hf_helen_time,
{ "Time", "helen.time", FT_ABSOLUTE_TIME, ABSOLUTE_TIME_UTC, NULL, 0x0, "Time", HFILL}},
and do something like
nstime_t t;
guint64 msecs_since_the_epoch;
msecs_since_the_epoch = tvb_get_ntoh64(tvb, offset);
t.secs = msecs_since_the_epoch/1000;
t.nsecs = (msecs_since_the_epoch%1000)*1000000; /* milliseconds to nanoseconds */
proto_tree_add_time(helen_sub_tree, hf_helen_time, tvb, offset, 8, &t);
If you want to display the time as local time (in the time zone of the machine running Wireshark), that's a bit easier.
