forked from osmocom/wireshark
83 lines
3.4 KiB
Plaintext
83 lines
3.4 KiB
Plaintext
$Id$
|
|
|
|
This is a very quick and very dirty guide to adding support for new
|
|
capture file formats. If you see any errors or have any improvements,
|
|
submit patches - free software is a community effort....
|
|
|
|
To add the ability to read a new capture file format, you have to:
|
|
|
|
add a new WTAP_FILE_ value for the file type to
|
|
"wiretap/wtap.h", and increase WTAP_NUM_FILE_TYPES by 1;
|
|
|
|
write an "open" routine that can read the beginning of the
|
|
capture file and figure out if it's in that format or not,
|
|
either by looking at a magic number at the beginning or by using
|
|
some form of heuristic to determine if it's a file of that type
|
|
(if the file format has a magic number, that's what should be
|
|
used);
|
|
|
|
write a "read" routine that can read a packet from the file and
|
|
supply the packet length, captured data length, and time stamp,
|
|
and have the "open" routine set the "subtype_read" member of the
|
|
"wtap" structure supplied to it to point to that routine;
|
|
|
|
write a "seek and read" routine, if necessary, and have the
|
|
"open" routine set the "subtype_seek_read" member of the "wtap"
|
|
structure to point to that routine.
|
|
|
|
write a "close" routine, if necessary (if, for example, the
|
|
"open" routine allocates any memory), and set the
|
|
"subtype_close" member of the "wtap" structure to point to it,
|
|
otherwise leave it set to NULL;
|
|
|
|
add a pointer to the "open" routine to the "open_routines[]"
|
|
table in "file_access.c" - if it uses a magic number, put it in
|
|
the first section of that list, and, if it uses a heuristic, put
|
|
it in the second section, preferably putting the heuristic
|
|
routines for binary files before the heuristic routines for text
|
|
files;
|
|
|
|
add an entry for that file type in the "dump_open_table[]" in
|
|
"file_access.c", giving a descriptive name, a short name that's
|
|
convenient to type on a command line (no blanks or capital
|
|
letters, please), and pointers to the "can_write_encap" and
|
|
"dump_open" routines if writing that file is supported (see
|
|
below), otherwise just null pointers.
|
|
|
|
Wiretap applications typically first perform sequential reads through
|
|
the capture file and may later do "seek and read" for individual frames.
|
|
The "read" routine should set the variable data_offset to the byte
|
|
offset within the capture file from which the "seek and read" routine
|
|
will read. If the capture records consist of:
|
|
|
|
capture record header
|
|
pseudo-header (e.g., for ATM)
|
|
frame data
|
|
|
|
then data_offset should point to the pseudo-header. The first
|
|
sequential read pass will process and store the capture record header
|
|
data, but it will not store the pseudo-header.
|
|
|
|
To add the ability to write a new capture file format, you have to:
|
|
|
|
add a "can_write_encap" routine that returns an indication of
|
|
whether a given packet encapsulation format is supported by the
|
|
new capture file format;
|
|
|
|
add a "dump_open" routine that starts writing a file (writing
|
|
headers, allocating data structures, etc.);
|
|
|
|
add a "dump" routine to write a packet to a file, and have the
|
|
"dump_open" routine set the "subtype_write" member of the
|
|
"wtap_dumper" structure passed to it to point to it;
|
|
|
|
add a "close" routine, if necessary (if, for example, the
|
|
"dump_open" routine allocates any memory, or if some of the file
|
|
header can be written only after all the packets have been
|
|
written), and have the "dump_open" routine set the
|
|
"subtype_close" member of the "wtap_dumper" structure to point
|
|
to it;
|
|
|
|
put pointers to the "can_write_encap" and "dump_open" routines
|
|
in the "dump_open_table[]" entry for that file type.
|