wireshark/packet-pgm.c

1350 lines
38 KiB
C
Raw Normal View History

/* packet-pgm.c
* Routines for pgm packet disassembly
*
* $Id: packet-pgm.c,v 1.19 2002/08/28 21:00:25 jmayer Exp $
*
* Copyright (c) 2000 by Talarian Corp
*
* Ethereal - Network traffic analyzer
* By Gerald Combs <gerald@ethereal.com>
* Copyright 1999 Gerald Combs
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
#include <time.h>
#include <string.h>
#include <epan/packet.h>
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
#include "afn.h"
#include "ipproto.h"
#include <epan/resolv.h>
#include <epan/strutil.h>
#include <epan/conversation.h>
#include "prefs.h"
#include <epan/proto.h>
void proto_reg_handoff_pgm(void);
static void proto_rereg_pgm(void);
typedef guint8 nchar_t;
typedef guint16 nshort_t;
typedef guint32 nlong_t;
/* The PGM main header */
typedef struct {
nshort_t sport; /* source port */
nshort_t dport; /* destination port */
nchar_t type; /* PGM type */
nchar_t opts; /* options */
nshort_t cksum; /* checksum */
nchar_t gsi[6]; /* Global Source ID */
nshort_t tsdulen; /* TSDU length */
} pgm_type;
#define pgmhdr_ntoh(_p) \
(_p)->sport = g_ntohs((_p)->sport); \
(_p)->dport = g_ntohs((_p)->dport); \
(_p)->type = g_ntohs((_p)->type); \
(_p)->opts = g_ntohs((_p)->opts); \
(_p)->cksum = g_ntohs((_p)->cksum); \
(_p)->tsdulen = g_ntohs((_p)->tsdulen)
/* The PGM SPM header */
typedef struct {
nlong_t sqn; /* SPM's sequence number */
nlong_t trail; /* Trailing edge sequence number */
nlong_t lead; /* Leading edge sequence number */
nshort_t path_afi; /* NLA AFI */
nshort_t res; /* reserved */
nlong_t path; /* Path NLA */
} pgm_spm_t;
static const size_t PGM_SPM_SZ = sizeof(pgm_type)+sizeof(pgm_spm_t);
#define spm_ntoh(_p) \
(_p)->sqn = g_ntohl((_p)->sqn); \
(_p)->trail = g_ntohl((_p)->trail); \
(_p)->lead = g_ntohl((_p)->lead); \
(_p)->path_afi = g_ntohs((_p)->path_afi); \
(_p)->res = g_ntohs((_p)->res);
/* The PGM Data (ODATA/RDATA) header */
typedef struct {
nlong_t sqn; /* Data Packet sequence number */
nlong_t trail; /* Trailing edge sequence number */
} pgm_data_t;
#define data_ntoh(_p) \
(_p)->sqn = g_ntohl((_p)->sqn); \
(_p)->trail = g_ntohl((_p)->trail)
static const size_t PGM_DATA_HDR_SZ = sizeof(pgm_type)+sizeof(pgm_data_t);
/* The PGM NAK (NAK/N-NAK/NCF) header */
typedef struct {
nlong_t sqn; /* Requested sequence number */
nshort_t src_afi; /* NLA AFI for source (IPv4 is set to 1) */
nshort_t src_res; /* reserved */
nlong_t src; /* Source NLA */
nshort_t grp_afi; /* Multicast group AFI (IPv4 is set to 1) */
nshort_t grp_res; /* reserved */
nlong_t grp; /* Multicast group NLA */
} pgm_nak_t;
static const size_t PGM_NAK_SZ = sizeof(pgm_type)+sizeof(pgm_nak_t);
#define nak_ntoh(_p) \
(_p)->sqn = g_ntohl((_p)->sqn); \
(_p)->src_afi = g_ntohs((_p)->src_afi); \
(_p)->src_res = g_ntohs((_p)->src_res); \
(_p)->grp_afi = g_ntohs((_p)->grp_afi); \
(_p)->grp_res = g_ntohs((_p)->grp_res)
/* The PGM ACK header (PGMCC) */
typedef struct {
nlong_t rx_max_sqn; /* RX_MAX sequence number */
nlong_t bitmap; /* Received Packet Bitmap */
} pgm_ack_t;
static const size_t PGM_ACK_SZ = sizeof(pgm_type)+sizeof(pgm_ack_t);
#define ack_ntoh(_p) \
(_p)->rx_max_sqn = g_ntohl((_p)->rx_max_sqn); \
(_p)->bitmap = g_ntohl((_p)->bitmap)
/* constants for hdr types */
#if defined(PGM_SPEC_01_PCKTS)
/* old spec-01 types */
#define PGM_SPM_PCKT 0x00
#define PGM_ODATA_PCKT 0x10
#define PGM_RDATA_PCKT 0x11
#define PGM_NAK_PCKT 0x20
#define PGM_NNAK_PCKT 0x21
#define PGM_NCF_PCKT 0x30
#else
/* spec-02 types (as well as spec-04+) */
#define PGM_SPM_PCKT 0x00
#define PGM_ODATA_PCKT 0x04
#define PGM_RDATA_PCKT 0x05
#define PGM_NAK_PCKT 0x08
#define PGM_NNAK_PCKT 0x09
#define PGM_NCF_PCKT 0x0A
#define PGM_ACK_PCKT 0x0D
#endif /* PGM_SPEC_01_PCKTS */
/* port swapping on NAK and NNAKs or not (default is to swap) */
/* PGM_NO_PORT_SWAP */
/* option flags (main PGM header) */
#define PGM_OPT 0x01
#define PGM_OPT_NETSIG 0x02
#define PGM_OPT_VAR_PKTLEN 0x40
#define PGM_OPT_PARITY 0x80
/* option types */
#define PGM_OPT_LENGTH 0x00
#define PGM_OPT_END 0x80
#define PGM_OPT_FRAGMENT 0x01
#define PGM_OPT_NAK_LIST 0x02
#define PGM_OPT_JOIN 0x03
#define PGM_OPT_REDIRECT 0x07
#define PGM_OPT_SYN 0x0D
#define PGM_OPT_FIN 0x0E
#define PGM_OPT_RST 0x0F
#define PGM_OPT_PARITY_PRM 0x08
#define PGM_OPT_PARITY_GRP 0x09
#define PGM_OPT_CURR_TGSIZE 0x0A
#define PGM_OPT_PGMCC_DATA 0x12
#define PGM_OPT_PGMCC_FEEDBACK 0x13
static const nchar_t PGM_OPT_INVALID = 0x7F;
/* OPX bit values */
#define PGM_OPX_IGNORE 0x00
#define PGM_OPX_INVAL 0x01
#define PGM_OPX_DISCARD 0x10
/* option formats */
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
} pgm_opt_generic_t;
typedef struct {
nchar_t type;
nchar_t len;
nshort_t total_len;
} pgm_opt_length_t;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
} pgm_opt_nak_list_t;
/*
* To squeeze the whole option into 255 bytes, we
* can only have 62 in the list
*/
#define PGM_MAX_NAK_LIST_SZ (62)
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
nlong_t opt_join_min;
} pgm_opt_join_t;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t po;
nlong_t prm_tgsz;
} pgm_opt_parity_prm_t;
/* OPT_PARITY_PRM P and O bits */
static const nchar_t PGM_OPT_PARITY_PRM_PRO = 0x2;
static const nchar_t PGM_OPT_PARITY_PRM_OND = 0x1;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
nlong_t prm_grp;
} pgm_opt_parity_grp_t;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
nlong_t prm_atgsz;
} pgm_opt_curr_tgsize_t;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
nlong_t tsp;
nshort_t acker_afi;
nshort_t res2;
nlong_t acker;
} pgm_opt_pgmcc_data_t;
typedef struct {
nchar_t type;
nchar_t len;
nchar_t opx;
nchar_t res;
nlong_t tsp;
nshort_t acker_afi;
nshort_t loss_rate;
nlong_t acker;
} pgm_opt_pgmcc_feedback_t;
/*
* Udp port for UDP encapsulation
*/
#define DEFAULT_UDP_ENCAP_UCAST_PORT 3055
#define DEFAULT_UDP_ENCAP_MCAST_PORT 3056
static int udp_encap_ucast_port = 0;
static int udp_encap_mcast_port = 0;
static int old_encap_ucast_port = 0;
static int old_encap_mcast_port = 0;
static int proto_pgm = -1;
static int ett_pgm = -1;
static int ett_pgm_optbits = -1;
static int ett_pgm_opts = -1;
static int ett_pgm_spm = -1;
static int ett_pgm_data = -1;
static int ett_pgm_nak = -1;
static int ett_pgm_ack = -1;
static int ett_pgm_opts_join = -1;
static int ett_pgm_opts_parityprm = -1;
static int ett_pgm_opts_paritygrp = -1;
static int ett_pgm_opts_naklist = -1;
static int ett_pgm_opts_ccdata = -1;
static int hf_pgm_main_sport = -1;
static int hf_pgm_main_dport = -1;
static int hf_pgm_main_type = -1;
static int hf_pgm_main_opts = -1;
static int hf_pgm_main_opts_opt = -1;
static int hf_pgm_main_opts_netsig = -1;
static int hf_pgm_main_opts_varlen = -1;
static int hf_pgm_main_opts_parity = -1;
static int hf_pgm_main_cksum = -1;
static int hf_pgm_main_gsi = -1;
static int hf_pgm_main_tsdulen = -1;
static int hf_pgm_spm_sqn = -1;
static int hf_pgm_spm_lead = -1;
static int hf_pgm_spm_trail = -1;
static int hf_pgm_spm_pathafi = -1;
static int hf_pgm_spm_res = -1;
static int hf_pgm_spm_path = -1;
static int hf_pgm_data_sqn = -1;
static int hf_pgm_data_trail = -1;
static int hf_pgm_nak_sqn = -1;
static int hf_pgm_nak_srcafi = -1;
static int hf_pgm_nak_srcres = -1;
static int hf_pgm_nak_src = -1;
static int hf_pgm_nak_grpafi = -1;
static int hf_pgm_nak_grpres = -1;
static int hf_pgm_nak_grp = -1;
static int hf_pgm_ack_sqn = -1;
static int hf_pgm_ack_bitmap = -1;
static int hf_pgm_opt_type = -1;
static int hf_pgm_opt_len = -1;
static int hf_pgm_opt_tlen = -1;
static int hf_pgm_genopt_type = -1;
static int hf_pgm_genopt_len = -1;
static int hf_pgm_genopt_opx = -1;
static int hf_pgm_opt_join_res = -1;
static int hf_pgm_opt_join_minjoin = -1;
static int hf_pgm_opt_parity_prm_po = -1;
static int hf_pgm_opt_parity_prm_prmtgsz = -1;
static int hf_pgm_opt_parity_grp_res = -1;
static int hf_pgm_opt_parity_grp_prmgrp = -1;
#ifdef PGM_UNUSED_HANDLES
static int hf_pgm_opt_curr_tgsize_type = -1;
static int hf_pgm_opt_curr_tgsize_len = -1;
static int hf_pgm_opt_curr_tgsize_opx = -1;
static int hf_pgm_opt_curr_tgsize_res = -1;
static int hf_pgm_opt_curr_tgsize_prmatgsz = -1;
#endif
static int hf_pgm_opt_nak_res = -1;
static int hf_pgm_opt_nak_list = -1;
static int hf_pgm_opt_ccdata_res = -1;
static int hf_pgm_opt_ccdata_tsp = -1;
static int hf_pgm_opt_ccdata_afi = -1;
static int hf_pgm_opt_ccdata_res2 = -1;
static int hf_pgm_opt_ccdata_acker = -1;
static int hf_pgm_opt_ccfeedbk_res = -1;
static int hf_pgm_opt_ccfeedbk_tsp = -1;
static int hf_pgm_opt_ccfeedbk_afi = -1;
static int hf_pgm_opt_ccfeedbk_lossrate = -1;
static int hf_pgm_opt_ccfeedbk_acker = -1;
static dissector_table_t subdissector_table;
static heur_dissector_list_t heur_subdissector_list;
static dissector_handle_t data_handle;
/*
* As of the time this comment was typed
*
* http://search.ietf.org/internet-drafts/draft-speakman-pgm-spec-06.txt
*
* was the URL for the PGM draft.
*/
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
static char *
optsstr(nchar_t opts)
{
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
static char msg[256];
char *p = msg, *str;
if (opts == 0)
return("");
if (opts & PGM_OPT){
sprintf(p, "Present");
p += strlen("Present");
}
if (opts & PGM_OPT_NETSIG){
if (p != msg)
str = ",NetSig";
else
str = "NetSig";
sprintf(p, str);
p += strlen(str);
}
if (opts & PGM_OPT_VAR_PKTLEN){
if (p != msg)
str = ",VarLen";
else
str = "VarLen";
sprintf(p, str);
p += strlen(str);
}
if (opts & PGM_OPT_PARITY){
if (p != msg)
str = ",Parity";
else
str = "Parity";
sprintf(p, str);
p += strlen(str);
}
if (p == msg) {
sprintf(p, "0x%x", opts);
}
return(msg);
}
static char *
paritystr(nchar_t parity)
{
static char msg[256];
char *p = msg, *str;
if (parity == 0)
return("");
if (parity & PGM_OPT_PARITY_PRM_PRO){
sprintf(p, "Pro-active");
p += strlen("Pro-active");
}
if (parity & PGM_OPT_PARITY_PRM_OND){
if (p != msg)
str = ",On-demand";
else
str = "On-demand";
sprintf(p, str);
p += strlen(str);
}
if (p == msg) {
sprintf(p, "0x%x", parity);
}
return(msg);
}
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
static const value_string opt_vals[] = {
{ PGM_OPT_LENGTH, "Length" },
{ PGM_OPT_END, "End" },
{ PGM_OPT_FRAGMENT, "Fragment" },
{ PGM_OPT_NAK_LIST, "NakList" },
{ PGM_OPT_JOIN, "Join" },
{ PGM_OPT_REDIRECT, "ReDirect" },
{ PGM_OPT_SYN, "Syn" },
{ PGM_OPT_FIN, "Fin" },
{ PGM_OPT_RST, "Rst" },
{ PGM_OPT_PARITY_PRM, "ParityPrm" },
{ PGM_OPT_PARITY_GRP, "ParityGrp" },
{ PGM_OPT_CURR_TGSIZE, "CurrTgsiz" },
{ PGM_OPT_PGMCC_DATA, "CcData" },
{ PGM_OPT_PGMCC_FEEDBACK, "CcFeedBack" },
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ 0, NULL }
};
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
static const value_string opx_vals[] = {
{ PGM_OPX_IGNORE, "Ignore" },
{ PGM_OPX_INVAL, "Inval" },
{ PGM_OPX_DISCARD, "DisCard" },
{ 0, NULL }
};
static void
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
dissect_pgmopts(tvbuff_t *tvb, int offset, proto_tree *tree,
const char *pktname)
{
proto_item *tf;
proto_tree *opts_tree = NULL;
proto_tree *opt_tree = NULL;
pgm_opt_length_t opts;
pgm_opt_generic_t genopts;
int theend = 0, firsttime = 1;
tvb_memcpy(tvb, (guint8 *)&opts, offset, sizeof(opts));
opts.total_len = g_ntohs(opts.total_len);
tf = proto_tree_add_text(tree, tvb, offset,
opts.total_len,
"%s Options (Total Length %d)", pktname, opts.total_len);
opts_tree = proto_item_add_subtree(tf, ett_pgm_opts);
proto_tree_add_uint(opts_tree, hf_pgm_opt_type, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset, 1, opts.type);
proto_tree_add_uint(opts_tree, hf_pgm_opt_len, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+1, 1, opts.len);
proto_tree_add_uint(opts_tree, hf_pgm_opt_tlen, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+2, 2, opts.total_len);
offset += 4;
for (opts.total_len -= 4; opts.total_len > 0;){
tvb_memcpy(tvb, (guint8 *)&genopts, offset, sizeof(genopts));
if (genopts.type & PGM_OPT_END) {
genopts.type &= ~PGM_OPT_END;
theend = 1;
}
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
tf = proto_tree_add_text(opts_tree, tvb, offset, genopts.len,
"Option: %s, Length: %u",
val_to_str(genopts.type, opt_vals, "Unknown (0x%02x)"),
genopts.len);
if (genopts.len == 0)
break;
switch(genopts.type) {
case PGM_OPT_JOIN:{
pgm_opt_join_t optdata;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_join);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type,
tvb, offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx, tvb,
offset+2, 1, genopts.opx);
proto_tree_add_uint(opt_tree, hf_pgm_opt_join_res, tvb,
offset+3, 1, optdata.res);
proto_tree_add_uint(opt_tree, hf_pgm_opt_join_minjoin, tvb,
offset+4, 4, g_ntohl(optdata.opt_join_min));
break;
}
case PGM_OPT_PARITY_PRM:{
pgm_opt_parity_prm_t optdata;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_parityprm);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type,
tvb, offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx,
tvb, offset+2, 1, genopts.opx);
proto_tree_add_uint_format(opt_tree, hf_pgm_opt_parity_prm_po, tvb,
offset+3, 1, optdata.po, "Parity Parameters: %s (0x%x)",
paritystr(optdata.po), optdata.po);
proto_tree_add_uint(opt_tree, hf_pgm_opt_parity_prm_prmtgsz,
tvb, offset+4, 4, g_ntohl(optdata.prm_tgsz));
break;
}
case PGM_OPT_PARITY_GRP:{
pgm_opt_parity_grp_t optdata;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_paritygrp);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type,
tvb, offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx,
tvb, offset+2, 1, genopts.opx);
proto_tree_add_uint(opt_tree, hf_pgm_opt_parity_grp_res, tvb,
offset+3, 1, optdata.res);
proto_tree_add_uint(opt_tree, hf_pgm_opt_parity_grp_prmgrp,
tvb, offset+4, 4, g_ntohl(optdata.prm_grp));
break;
}
case PGM_OPT_NAK_LIST:{
pgm_opt_nak_list_t optdata;
nlong_t naklist[PGM_MAX_NAK_LIST_SZ+1];
char nakbuf[8192], *ptr;
int i, j, naks, soffset = 0;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_naklist);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type, tvb,
offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx,
tvb, offset+2, 1, genopts.opx);
proto_tree_add_uint(opt_tree, hf_pgm_opt_nak_res, tvb,
offset+3, 1, optdata.res);
optdata.len -= sizeof(pgm_opt_nak_list_t);
tvb_memcpy(tvb, (guint8 *)naklist, offset+4, optdata.len);
naks = (optdata.len/sizeof(nlong_t));
ptr = nakbuf;
j = 0;
/*
* Print out 8 per line
*/
for (i=0; i < naks; i++) {
sprintf(nakbuf+soffset, "0x%lx ",
(unsigned long)g_ntohl(naklist[i]));
soffset = strlen(nakbuf);
if ((++j % 8) == 0) {
if (firsttime) {
proto_tree_add_bytes_format(opt_tree,
hf_pgm_opt_nak_list, tvb, offset+4, optdata.len,
nakbuf, "List(%d): %s", naks, nakbuf);
soffset = 0;
} else {
proto_tree_add_bytes_format(opt_tree,
hf_pgm_opt_nak_list, tvb, offset+4, optdata.len,
nakbuf, "List: %s", nakbuf);
soffset = 0;
}
firsttime = 0;
}
}
if (soffset) {
if (firsttime) {
proto_tree_add_bytes_format(opt_tree,
hf_pgm_opt_nak_list, tvb, offset+4, optdata.len,
nakbuf, "List(%d): %s", naks, nakbuf);
soffset = 0;
} else {
proto_tree_add_bytes_format(opt_tree,
hf_pgm_opt_nak_list, tvb, offset+4, optdata.len,
nakbuf, "List: %s", nakbuf);
soffset = 0;
}
}
break;
}
case PGM_OPT_PGMCC_DATA:{
pgm_opt_pgmcc_data_t optdata;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_ccdata);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type,
tvb, offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx,
tvb, offset+2, 1, genopts.opx);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccdata_res, tvb,
offset+3, 1, optdata.res);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccdata_tsp, tvb,
offset+4, 4, optdata.tsp);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccdata_afi, tvb,
offset+8, 2, g_ntohs(optdata.acker_afi));
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccdata_res2, tvb,
offset+10, 2, g_ntohs(optdata.res2));
switch (g_ntohs(optdata.acker_afi)) {
case AFNUM_INET:
proto_tree_add_ipv4(opt_tree, hf_pgm_opt_ccdata_acker,
tvb, offset+12, 4, optdata.acker);
break;
default:
/*
* XXX - the header is variable-length,
* as the length of the NLA depends on
* its AFI.
*
* However, our structure for it is
* fixed-length, and assumes it's a 4-byte
* IPv4 address.
*/
break;
}
break;
}
case PGM_OPT_PGMCC_FEEDBACK:{
pgm_opt_pgmcc_feedback_t optdata;
tvb_memcpy(tvb, (guint8 *)&optdata, offset, sizeof(optdata));
opt_tree = proto_item_add_subtree(tf, ett_pgm_opts_ccdata);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_type,
tvb, offset, 1, genopts.type);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_len, tvb,
offset+1, 1, genopts.len);
proto_tree_add_uint(opt_tree, hf_pgm_genopt_opx,
tvb, offset+2, 1, genopts.opx);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccfeedbk_res, tvb,
offset+3, 1, optdata.res);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccfeedbk_tsp, tvb,
offset+4, 4, optdata.tsp);
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccfeedbk_afi, tvb,
offset+8, 2, g_ntohs(optdata.acker_afi));
proto_tree_add_uint(opt_tree, hf_pgm_opt_ccfeedbk_lossrate, tvb,
offset+10, 2, g_ntohs(optdata.loss_rate));
switch (g_ntohs(optdata.acker_afi)) {
case AFNUM_INET:
proto_tree_add_ipv4(opt_tree, hf_pgm_opt_ccfeedbk_acker,
tvb, offset+12, 4, optdata.acker);
break;
default:
/*
* XXX - the header is variable-length,
* as the length of the NLA depends on
* its AFI.
*
* However, our structure for it is
* fixed-length, and assumes it's a 4-byte
* IPv4 address.
*/
break;
}
break;
}
}
offset += genopts.len;
opts.total_len -= genopts.len;
}
return ;
}
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
static const value_string type_vals[] = {
{ PGM_SPM_PCKT, "SPM" },
{ PGM_RDATA_PCKT, "RDATA" },
{ PGM_ODATA_PCKT, "ODATA" },
{ PGM_NAK_PCKT, "NAK" },
{ PGM_NNAK_PCKT, "NNAK" },
{ PGM_NCF_PCKT, "NCF" },
{ PGM_ACK_PCKT, "ACK" },
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ 0, NULL }
};
/* Determine if there is a sub-dissector and call it. This has been */
/* separated into a stand alone routine to other protocol dissectors */
/* can call to it, ie. socks */
void
decode_pgm_ports(tvbuff_t *tvb, int offset, packet_info *pinfo,
proto_tree *tree, pgm_type *pgmhdr)
{
tvbuff_t *next_tvb;
int found = 0;
next_tvb = tvb_new_subset(tvb, offset, -1, -1);
/* do lookup with the subdissector table */
found = dissector_try_port(subdissector_table, pgmhdr->sport,
next_tvb, pinfo, tree);
if (found)
return;
found = dissector_try_port(subdissector_table, pgmhdr->dport,
next_tvb, pinfo, tree);
if (found)
return;
/* do lookup with the heuristic subdissector table */
if (dissector_try_heuristic(heur_subdissector_list, next_tvb, pinfo, tree))
return;
/* Oh, well, we don't know this; dissect it as data. */
call_dissector(data_handle,next_tvb, pinfo, tree);
}
int
total_size(tvbuff_t *tvb, pgm_type *hdr)
{
int bytes = sizeof(pgm_type);
pgm_opt_length_t opts;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
switch(hdr->type) {
case PGM_SPM_PCKT:
bytes += sizeof(pgm_spm_t);
break;
case PGM_RDATA_PCKT:
case PGM_ODATA_PCKT:
bytes += sizeof(pgm_data_t);
break;
case PGM_NAK_PCKT:
case PGM_NNAK_PCKT:
case PGM_NCF_PCKT:
bytes += sizeof(pgm_nak_t);
break;
case PGM_ACK_PCKT:
bytes += sizeof(pgm_ack_t);
break;
}
if ((hdr->opts & PGM_OPT)) {
tvb_memcpy(tvb, (guint8 *)&opts, bytes, sizeof(opts));
bytes += g_ntohs(opts.total_len);
}
return(bytes);
}
/*
* dissect_pgm - The dissector for Pragmatic General Multicast
*/
static void
dissect_pgm(tvbuff_t *tvb, packet_info *pinfo, proto_tree *tree)
{
proto_tree *pgm_tree = NULL;
proto_tree *opt_tree = NULL;
proto_tree *type_tree = NULL;
proto_item *tf;
pgm_type pgmhdr;
pgm_spm_t spm;
pgm_data_t data;
pgm_nak_t nak;
pgm_ack_t ack;
int offset = 0;
guint hlen, plen;
proto_item *ti;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
const char *pktname;
char *gsi;
int isdata = 0;
if (check_col(pinfo->cinfo, COL_PROTOCOL))
col_set_str(pinfo->cinfo, COL_PROTOCOL, "PGM");
/* Clear out the Info column. */
if (check_col(pinfo->cinfo, COL_INFO))
col_clear(pinfo->cinfo, COL_INFO);
tvb_memcpy(tvb, (guint8 *)&pgmhdr, offset, sizeof(pgm_type));
hlen = sizeof(pgm_type);
pgmhdr.sport = g_ntohs(pgmhdr.sport);
pgmhdr.dport = g_ntohs(pgmhdr.dport);
pgmhdr.tsdulen = g_ntohs(pgmhdr.tsdulen);
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
pktname = val_to_str(pgmhdr.type, type_vals, "Unknown (0x%02x)");
gsi = bytes_to_str(pgmhdr.gsi, 6);
switch(pgmhdr.type) {
case PGM_SPM_PCKT:
plen = sizeof(pgm_spm_t);
tvb_memcpy(tvb, (guint8 *)&spm, sizeof(pgm_type), plen);
spm_ntoh(&spm);
if (check_col(pinfo->cinfo, COL_INFO)) {
col_add_fstr(pinfo->cinfo, COL_INFO,
"%-5s sqn 0x%x gsi %s", pktname, spm.sqn, gsi);
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
}
break;
case PGM_RDATA_PCKT:
case PGM_ODATA_PCKT:
plen = sizeof(pgm_data_t);
tvb_memcpy(tvb, (guint8 *)&data, sizeof(pgm_type), plen);
data_ntoh(&data);
if (check_col(pinfo->cinfo, COL_INFO)) {
col_add_fstr(pinfo->cinfo, COL_INFO,
"%-5s sqn 0x%x gsi %s tsdulen %d", pktname, data.sqn, gsi,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
pgmhdr.tsdulen);
}
isdata = 1;
break;
case PGM_NAK_PCKT:
case PGM_NNAK_PCKT:
case PGM_NCF_PCKT:
plen = sizeof(pgm_nak_t);
tvb_memcpy(tvb, (guint8 *)&nak, sizeof(pgm_type), plen);
nak_ntoh(&nak);
if (check_col(pinfo->cinfo, COL_INFO)) {
col_add_fstr(pinfo->cinfo, COL_INFO,
"%-5s sqn 0x%x gsi %s", pktname, nak.sqn, gsi);
}
break;
case PGM_ACK_PCKT:
plen = sizeof(pgm_ack_t);
tvb_memcpy(tvb, (guint8 *)&ack, sizeof(pgm_type), plen);
ack_ntoh(&ack);
if (check_col(pinfo->cinfo, COL_INFO)) {
col_add_fstr(pinfo->cinfo, COL_INFO,
"%-5s sqn 0x%x gsi %s", pktname, ack.rx_max_sqn, gsi);
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
}
break;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
default:
return;
}
if (tree) {
ti = proto_tree_add_protocol_format(tree, proto_pgm,
tvb, offset, total_size(tvb, &pgmhdr),
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
"Pragmatic General Multicast: Type %s"
" SrcPort %u, DstPort %u, GSI %s", pktname,
pgmhdr.sport, pgmhdr.dport,
bytes_to_str(pgmhdr.gsi, 6));
pgm_tree = proto_item_add_subtree(ti, ett_pgm);
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
proto_tree_add_uint(pgm_tree, hf_pgm_main_sport, tvb, offset, 2,
pgmhdr.sport);
proto_tree_add_uint(pgm_tree, hf_pgm_main_dport, tvb, offset+2,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
2, pgmhdr.dport);
proto_tree_add_uint(pgm_tree, hf_pgm_main_type, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+4, 1, pgmhdr.type);
tf = proto_tree_add_uint_format(pgm_tree, hf_pgm_main_opts, tvb,
offset+5, 1, pgmhdr.opts, "Options: %s (0x%x)",
optsstr(pgmhdr.opts), pgmhdr.opts);
opt_tree = proto_item_add_subtree(tf, ett_pgm_optbits);
proto_tree_add_boolean(opt_tree, hf_pgm_main_opts_opt, tvb,
offset+5, 1, (pgmhdr.opts & PGM_OPT));
proto_tree_add_boolean(opt_tree, hf_pgm_main_opts_netsig, tvb,
offset+5, 1, (pgmhdr.opts & PGM_OPT_NETSIG));
proto_tree_add_boolean(opt_tree, hf_pgm_main_opts_varlen, tvb,
offset+5, 1, (pgmhdr.opts & PGM_OPT_VAR_PKTLEN));
proto_tree_add_boolean(opt_tree, hf_pgm_main_opts_parity, tvb,
offset+5, 1, (pgmhdr.opts & PGM_OPT_PARITY));
proto_tree_add_uint(pgm_tree, hf_pgm_main_cksum, tvb, offset+6,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
2, pgmhdr.cksum);
proto_tree_add_bytes(pgm_tree, hf_pgm_main_gsi, tvb, offset+8,
6, pgmhdr.gsi);
proto_tree_add_uint(pgm_tree, hf_pgm_main_tsdulen, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+14, 2, pgmhdr.tsdulen);
offset = sizeof(pgm_type);
tf = proto_tree_add_text(pgm_tree, tvb, offset, plen, "%s Packet",
pktname);
switch(pgmhdr.type) {
case PGM_SPM_PCKT:
type_tree = proto_item_add_subtree(tf, ett_pgm_spm);
proto_tree_add_uint(type_tree, hf_pgm_spm_sqn, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset, 4, spm.sqn);
proto_tree_add_uint(type_tree, hf_pgm_spm_trail, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+4, 4, spm.trail);
proto_tree_add_uint(type_tree, hf_pgm_spm_lead, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+8, 4, spm.lead);
proto_tree_add_uint(type_tree, hf_pgm_spm_pathafi, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+10, 2, spm.path_afi);
proto_tree_add_uint(type_tree, hf_pgm_spm_res, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+12, 2, spm.res);
switch (spm.path_afi) {
case AFNUM_INET:
proto_tree_add_ipv4(type_tree, hf_pgm_spm_path,
tvb, offset+14, 4, spm.path);
break;
default:
/*
* XXX - the header is variable-length,
* as the length of the NLA depends on
* its AFI.
*
* However, our structure for it is
* fixed-length, and assumes it's a 4-byte
* IPv4 address.
*/
return;
}
if ((pgmhdr.opts & PGM_OPT) == FALSE)
break;
offset += plen;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
dissect_pgmopts(tvb, offset, type_tree, pktname);
break;
case PGM_RDATA_PCKT:
case PGM_ODATA_PCKT: {
type_tree = proto_item_add_subtree(tf, ett_pgm_data);
proto_tree_add_uint(type_tree, hf_pgm_spm_sqn, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset, 4, data.sqn);
proto_tree_add_uint(type_tree, hf_pgm_spm_trail, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+4, 4, data.trail);
if ((pgmhdr.opts & PGM_OPT) == FALSE)
break;
offset += plen;
dissect_pgmopts(tvb, offset, type_tree, pktname);
break;
}
case PGM_NAK_PCKT:
case PGM_NNAK_PCKT:
case PGM_NCF_PCKT:
type_tree = proto_item_add_subtree(tf, ett_pgm_nak);
proto_tree_add_uint(type_tree, hf_pgm_nak_sqn, tvb,
offset, 4, nak.sqn);
proto_tree_add_uint(type_tree, hf_pgm_nak_srcafi, tvb,
offset+4, 2, nak.src_afi);
proto_tree_add_uint(type_tree, hf_pgm_nak_srcres, tvb,
offset+6, 2, nak.src_res);
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
switch (nak.src_afi) {
case AFNUM_INET:
proto_tree_add_ipv4(type_tree, hf_pgm_nak_src,
tvb, offset+8, 4, nak.src);
break;
default:
/*
* XXX - the header is variable-length,
* as the length of the NLA depends on
* its AFI.
*
* However, our structure for it is
* fixed-length, and assumes it's a 4-byte
* IPv4 address.
*/
break;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
}
proto_tree_add_uint(type_tree, hf_pgm_nak_grpafi, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+12, 2, nak.grp_afi);
proto_tree_add_uint(type_tree, hf_pgm_nak_grpres, tvb,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
offset+14, 2, nak.grp_res);
switch (nak.grp_afi) {
case AFNUM_INET:
proto_tree_add_ipv4(type_tree, hf_pgm_nak_grp,
tvb, offset+16, 4, nak.grp);
break;
default:
/*
* XXX - the header is variable-length,
* as the length of the NLA depends on
* its AFI.
*
* However, our structure for it is
* fixed-length, and assumes it's a 4-byte
* IPv4 address.
*/
return;
}
if ((pgmhdr.opts & PGM_OPT) == FALSE)
break;
offset += plen;
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
dissect_pgmopts(tvb, offset, type_tree, pktname);
break;
case PGM_ACK_PCKT:
type_tree = proto_item_add_subtree(tf, ett_pgm_ack);
proto_tree_add_uint(type_tree, hf_pgm_ack_sqn, tvb,
offset, 4, ack.rx_max_sqn);
proto_tree_add_uint(type_tree, hf_pgm_ack_bitmap, tvb,
offset+4, 4, ack.bitmap);
if ((pgmhdr.opts & PGM_OPT) == FALSE)
break;
offset += plen;
dissect_pgmopts(tvb, offset, type_tree, pktname);
break;
}
}
if (isdata) {
/*
* Now see if there are any sub-dissectors, if so call them
*/
offset = total_size(tvb, &pgmhdr);
decode_pgm_ports(tvb, offset, pinfo, tree, &pgmhdr);
}
pktname = NULL;
}
static const true_false_string opts_present = {
"Present",
"Not Present"
};
/* Register all the bits needed with the filtering engine */
void
proto_register_pgm(void)
{
static hf_register_info hf[] = {
{ &hf_pgm_main_sport,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Source Port", "pgm.hdr.sport", FT_UINT16, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_main_dport,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Destination Port", "pgm.hdr.dport", FT_UINT16, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_main_type,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Type", "pgm.hdr.type", FT_UINT8, BASE_HEX,
VALS(type_vals), 0x0, "", HFILL }},
{ &hf_pgm_main_opts,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Options", "pgm.hdr.opts", FT_UINT8, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_main_opts_opt,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Options", "pgm.hdr.opts.opt", FT_BOOLEAN, BASE_NONE,
TFS(&opts_present), PGM_OPT, "", HFILL }},
{ &hf_pgm_main_opts_netsig,
{ "Network Significant Options", "pgm.hdr.opts.netsig",
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
FT_BOOLEAN, BASE_NONE,
TFS(&opts_present), PGM_OPT_NETSIG, "", HFILL }},
{ &hf_pgm_main_opts_varlen,
{ "Variable length Parity Packet Option", "pgm.hdr.opts.varlen",
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
FT_BOOLEAN, BASE_NONE,
TFS(&opts_present), PGM_OPT_VAR_PKTLEN, "", HFILL }},
{ &hf_pgm_main_opts_parity,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Parity", "pgm.hdr.opts.parity", FT_BOOLEAN, BASE_NONE,
TFS(&opts_present), PGM_OPT_PARITY, "", HFILL }},
{ &hf_pgm_main_cksum,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Checksum", "pgm.hdr.cksum", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_main_gsi,
{ "Global Source Identifier", "pgm.hdr.gsi", FT_BYTES, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_main_tsdulen,
{ "Transport Service Data Unit Length", "pgm.hdr.tsdulen", FT_UINT16,
BASE_DEC, NULL, 0x0, "", HFILL }},
{ &hf_pgm_spm_sqn,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Sequence number", "pgm.spm.sqn", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_spm_trail,
{ "Trailing Edge Sequence Number", "pgm.spm.trail", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_spm_lead,
{ "Leading Edge Sequence Number", "pgm.spm.lead", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_spm_pathafi,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Path NLA AFI", "pgm.spm.pathafi", FT_UINT16, BASE_DEC,
VALS(afn_vals), 0x0, "", HFILL }},
{ &hf_pgm_spm_res,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.spm.res", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_spm_path,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Path NLA", "pgm.spm.path", FT_IPv4, BASE_NONE,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_data_sqn,
{ "Data Packet Sequence Number", "pgm.data.sqn", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_data_trail,
{ "Trailing Edge Sequence Number", "pgm.data.trail", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_nak_sqn,
{ "Requested Sequence Number", "pgm.nak.sqn", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_nak_srcafi,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Source NLA AFI", "pgm.nak.srcafi", FT_UINT16, BASE_DEC,
VALS(afn_vals), 0x0, "", HFILL }},
{ &hf_pgm_nak_srcres,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.nak.srcres", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_nak_src,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Source NLA", "pgm.nak.src", FT_IPv4, BASE_NONE,
NULL, 0x0, "", HFILL }},
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ &hf_pgm_nak_grpafi,
{ "Multicast Group AFI", "pgm.nak.grpafi", FT_UINT16, BASE_DEC,
VALS(afn_vals), 0x0, "", HFILL }},
{ &hf_pgm_nak_grpres,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.nak.grpres", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_nak_grp,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Multicast Group NLA", "pgm.nak.grp", FT_IPv4, BASE_NONE,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_ack_sqn,
{ "Maximum Received Sequence Number", "pgm.ack.maxsqn", FT_UINT32,
BASE_HEX, NULL, 0x0, "", HFILL }},
{ &hf_pgm_ack_bitmap,
{ "Packet Bitmap", "pgm.ack.bitmap", FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_type,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Type", "pgm.opts.type", FT_UINT8, BASE_HEX,
VALS(opt_vals), 0x0, "", HFILL }},
{ &hf_pgm_opt_len,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Length", "pgm.opts.len", FT_UINT8, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_tlen,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Total Length", "pgm.opts.tlen", FT_UINT16, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_genopt_type,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Type", "pgm.genopts.type", FT_UINT8, BASE_HEX,
VALS(opt_vals), 0x0, "", HFILL }},
{ &hf_pgm_genopt_len,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Length", "pgm.genopts.len", FT_UINT8, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_genopt_opx,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Option Extensibility Bits", "pgm.genopts.opx", FT_UINT8, BASE_HEX,
VALS(opx_vals), 0x0, "", HFILL }},
{ &hf_pgm_opt_parity_prm_po,
{ "Parity Parameters", "pgm.opts.parity_prm.op", FT_UINT8, BASE_HEX,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_parity_prm_prmtgsz,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Transmission Group Size", "pgm.opts.parity_prm.prm_grp",
FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_join_res,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.opts.join.res", FT_UINT8, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_join_minjoin,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Minimum Sequence Number", "pgm.opts.join.min_join",
FT_UINT32, BASE_HEX, NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_parity_grp_res,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.opts.parity_prm.op", FT_UINT8, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_parity_grp_prmgrp,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Transmission Group Size", "pgm.opts.parity_prm.prm_grp",
FT_UINT32, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_nak_res,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "Reserved", "pgm.opts.nak.op", FT_UINT8, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_nak_list,
There is really no need to have the BGP dissector and the LDP dissector have two independent "value_string" tables mapping RFC 1700 address family numbers to names, nor is there any need to have the BGP dissector and the PIM dissector have two independent sets of #defines for RFC 1700 address family numbers; put a single "value_string" table in "afn.c" and put a declaration of it, and #defines for the address family numbers, into "afn.h", and have the dissectors use that. Move the #define for PGM into "ipproto.h", and add an entry for it in the "value_string" table in "ipproto.c". Have the PGM dissector use the standard Ethereal mechanisms for resolving addresses, and have it use "value_string" tables for mapping option types, the OPX bits, and packet types to strings. Use "bytes_to_str()" to turn byte arrays into strings of hex digits. Pass the packet type string to "dissect_pgmopts()" as an argument, rather than making it a global. Don't use "proto_tree_add_XXX_format" routines if you can possibly just use "proto_tree_add_XXX"; give various fields the correct radix and type, and VALS() strings if necessary, to make that happen (and to make filtering on them more pleasant). Put the type, length, and total length of the options into the protocol tree as separate fields. Don't have separate type, length, and OPX fields for every type of option; one field will suffice. Don't format a string with "sprintf()" and then pass that string to "col_add_fstr()" with a format of "%s" and the string as an argument - "col_add_fstr()" can format strings itself (that's what the "f" stands for). Don't byte-swap and then un-byte-swap IPv4 address fields in the header, just leave them network byte order to start with. Use the correct fields for "proto_tree_add_XXX", rather than using the same field multiple times. Quit early if an address family identifier isn't AFNUM_INET, as that means the structure we use to dissect the header doesn't match the actual header. svn path=/trunk/; revision=3761
2001-07-21 10:27:13 +00:00
{ "List", "pgm.opts.nak.list", FT_BYTES, BASE_NONE,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccdata_res,
{ "Reserved", "pgm.opts.ccdata.res", FT_UINT8, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccdata_tsp,
{ "Time Stamp", "pgm.opts.ccdata.tstamp", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccdata_afi,
{ "Acker AFI", "pgm.opts.ccdata.afi", FT_UINT16, BASE_DEC,
VALS(afn_vals), 0x0, "", HFILL }},
{ &hf_pgm_opt_ccdata_res2,
{ "Reserved", "pgm.opts.ccdata.res2", FT_UINT16, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccdata_acker,
{ "Acker", "pgm.opts.ccdata.acker", FT_IPv4, BASE_NONE,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccfeedbk_res,
{ "Reserved", "pgm.opts.ccdata.res", FT_UINT8, BASE_DEC,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccfeedbk_tsp,
{ "Time Stamp", "pgm.opts.ccdata.tstamp", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccfeedbk_afi,
{ "Acker AFI", "pgm.opts.ccdata.afi", FT_UINT16, BASE_DEC,
VALS(afn_vals), 0x0, "", HFILL }},
{ &hf_pgm_opt_ccfeedbk_lossrate,
{ "Loss Rate", "pgm.opts.ccdata.lossrate", FT_UINT16, BASE_HEX,
NULL, 0x0, "", HFILL }},
{ &hf_pgm_opt_ccfeedbk_acker,
{ "Acker", "pgm.opts.ccdata.acker", FT_IPv4, BASE_NONE,
NULL, 0x0, "", HFILL }},
};
static gint *ett[] = {
&ett_pgm,
&ett_pgm_optbits,
&ett_pgm_spm,
&ett_pgm_data,
&ett_pgm_nak,
&ett_pgm_ack,
&ett_pgm_opts,
&ett_pgm_opts_join,
&ett_pgm_opts_parityprm,
&ett_pgm_opts_paritygrp,
&ett_pgm_opts_naklist,
&ett_pgm_opts_ccdata,
};
module_t *pgm_module;
proto_pgm = proto_register_protocol("Pragmatic General Multicast",
"PGM", "pgm");
proto_register_field_array(proto_pgm, hf, array_length(hf));
proto_register_subtree_array(ett, array_length(ett));
/* subdissector code */
subdissector_table = register_dissector_table("pgm.port",
"PGM port", FT_UINT16, BASE_DEC);
register_heur_dissector_list("pgm", &heur_subdissector_list);
/*
* Register configuration preferences for UDP encapsulation
* (Note: Initially the ports are set to zero so the
* dissecting of PGM encapsulated in UPD packets
* is off by default)
*/
pgm_module = prefs_register_protocol(proto_pgm, proto_rereg_pgm);
prefs_register_uint_preference(pgm_module, "udp.encap_ucast_port",
"PGM Encap Unicast Port (standard is 3055)",
"PGM Encap is PGM packets encapsulated in UDP packets"
" (Note: This option is off, i.e. port is 0, by default)",
10, &udp_encap_ucast_port);
old_encap_ucast_port = udp_encap_ucast_port;
prefs_register_uint_preference(pgm_module, "udp.encap_mcast_port",
"PGM Encap Multicast Port (standard is 3056)",
"PGM Encap is PGM packets encapsulated in UDP packets"
" (Note: This option is off, i.e. port is 0, by default)",
10, &udp_encap_mcast_port);
old_encap_mcast_port = udp_encap_mcast_port;
}
static dissector_handle_t pgm_handle;
/* The registration hand-off routine */
void
proto_reg_handoff_pgm(void)
{
pgm_handle = create_dissector_handle(dissect_pgm, proto_pgm);
/*
* Set up PGM Encap dissecting, which is off by default
*/
dissector_add("udp.port", udp_encap_ucast_port, pgm_handle);
dissector_add("udp.port", udp_encap_mcast_port, pgm_handle);
dissector_add("ip.proto", IP_PROTO_PGM, pgm_handle);
data_handle = find_dissector("data");
}
static void
proto_rereg_pgm(void)
{
/*
* Remove the old ones
*/
dissector_delete("udp.port", old_encap_ucast_port, pgm_handle);
dissector_delete("udp.port", old_encap_mcast_port, pgm_handle);
/*
* Set the new ones
*/
dissector_add("udp.port", udp_encap_ucast_port, pgm_handle);
dissector_add("udp.port", udp_encap_mcast_port, pgm_handle);
}