summaryrefslogtreecommitdiffstats
path: root/util-print.c
diff options
context:
space:
mode:
Diffstat (limited to 'util-print.c')
-rw-r--r--util-print.c959
1 files changed, 959 insertions, 0 deletions
diff --git a/util-print.c b/util-print.c
new file mode 100644
index 0000000..8b20649
--- /dev/null
+++ b/util-print.c
@@ -0,0 +1,959 @@
+/*
+ * Copyright (c) 1990, 1991, 1993, 1994, 1995, 1996, 1997
+ * The Regents of the University of California. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that: (1) source code distributions
+ * retain the above copyright notice and this paragraph in its entirety, (2)
+ * distributions including binary code include the above copyright notice and
+ * this paragraph in its entirety in the documentation or other materials
+ * provided with the distribution, and (3) all advertising materials mentioning
+ * features or use of this software display the following acknowledgement:
+ * ``This product includes software developed by the University of California,
+ * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of
+ * the University nor the names of its contributors may be used to endorse
+ * or promote products derived from this software without specific prior
+ * written permission.
+ * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
+ * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
+ */
+
+/*
+ * txtproto_print() derived from original code by Hannes Gredler
+ * (hannes@gredler.at):
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that: (1) source code
+ * distributions retain the above copyright notice and this paragraph
+ * in its entirety, and (2) distributions including binary code include
+ * the above copyright notice and this paragraph in its entirety in
+ * the documentation or other materials provided with the distribution.
+ * THIS SOFTWARE IS PROVIDED ``AS IS'' AND
+ * WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT
+ * LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE.
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "netdissect-stdinc.h"
+
+#include <sys/stat.h>
+
+#ifdef HAVE_FCNTL_H
+#include <fcntl.h>
+#endif
+#include <stdio.h>
+#include <stdarg.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "netdissect-ctype.h"
+
+#include "netdissect.h"
+#include "extract.h"
+#include "ascii_strcasecmp.h"
+#include "timeval-operations.h"
+
+#define TOKBUFSIZE 128
+
+enum date_flag { WITHOUT_DATE = 0, WITH_DATE = 1 };
+enum time_flag { UTC_TIME = 0, LOCAL_TIME = 1 };
+
+/*
+ * Print out a character, filtering out the non-printable ones
+ */
+void
+fn_print_char(netdissect_options *ndo, u_char c)
+{
+ if (!ND_ISASCII(c)) {
+ c = ND_TOASCII(c);
+ ND_PRINT("M-");
+ }
+ if (!ND_ASCII_ISPRINT(c)) {
+ c ^= 0x40; /* DEL to ?, others to alpha */
+ ND_PRINT("^");
+ }
+ ND_PRINT("%c", c);
+}
+
+/*
+ * Print a null-terminated string, filtering out non-printable characters.
+ * DON'T USE IT with a pointer on the packet buffer because there is no
+ * truncation check. For this use, see the nd_printX() functions below.
+ */
+void
+fn_print_str(netdissect_options *ndo, const u_char *s)
+{
+ while (*s != '\0') {
+ fn_print_char(ndo, *s);
+ s++;
+ }
+}
+
+/*
+ * Print out a null-terminated filename (or other ASCII string) from
+ * a fixed-length field in the packet buffer, or from what remains of
+ * the packet.
+ *
+ * n is the length of the fixed-length field, or the number of bytes
+ * remaining in the packet based on its on-the-network length.
+ *
+ * If ep is non-null, it should point just past the last captured byte
+ * of the packet, e.g. ndo->ndo_snapend. If ep is NULL, we assume no
+ * truncation check, other than the checks of the field length/remaining
+ * packet data length, is needed.
+ *
+ * Return the number of bytes of string processed, including the
+ * terminating null, if not truncated; as the terminating null is
+ * included in the count, and as there must be a terminating null,
+ * this will always be non-zero. Return 0 if truncated.
+ */
+u_int
+nd_printztn(netdissect_options *ndo,
+ const u_char *s, u_int n, const u_char *ep)
+{
+ u_int bytes;
+ u_char c;
+
+ bytes = 0;
+ for (;;) {
+ if (n == 0 || (ep != NULL && s >= ep)) {
+ /*
+ * Truncated. This includes "no null before we
+ * got to the end of the fixed-length buffer or
+ * the end of the packet".
+ *
+ * XXX - BOOTP says "null-terminated", which
+ * means the maximum length of the string, in
+ * bytes, is 1 less than the size of the buffer,
+ * as there must always be a terminating null.
+ */
+ bytes = 0;
+ break;
+ }
+
+ c = GET_U_1(s);
+ s++;
+ bytes++;
+ n--;
+ if (c == '\0') {
+ /* End of string */
+ break;
+ }
+ fn_print_char(ndo, c);
+ }
+ return(bytes);
+}
+
+/*
+ * Print out a counted filename (or other ASCII string), part of
+ * the packet buffer.
+ * If ep is NULL, assume no truncation check is needed.
+ * Return true if truncated.
+ * Stop at ep (if given) or after n bytes, whichever is first.
+ */
+int
+nd_printn(netdissect_options *ndo,
+ const u_char *s, u_int n, const u_char *ep)
+{
+ u_char c;
+
+ while (n > 0 && (ep == NULL || s < ep)) {
+ n--;
+ c = GET_U_1(s);
+ s++;
+ fn_print_char(ndo, c);
+ }
+ return (n == 0) ? 0 : 1;
+}
+
+/*
+ * Print a null-padded filename (or other ASCII string), part of
+ * the packet buffer, filtering out non-printable characters.
+ * Stop if truncated (via GET_U_1/longjmp) or after n bytes or before
+ * the null char, whichever occurs first.
+ * The suffix comes from: j:longJmp, n:after N bytes, p:null-Padded.
+ */
+void
+nd_printjnp(netdissect_options *ndo, const u_char *s, u_int n)
+{
+ u_char c;
+
+ while (n > 0) {
+ c = GET_U_1(s);
+ if (c == '\0')
+ break;
+ fn_print_char(ndo, c);
+ n--;
+ s++;
+ }
+}
+
+/*
+ * Print the timestamp .FRAC part (Microseconds/nanoseconds)
+ */
+static void
+ts_frac_print(netdissect_options *ndo, long usec)
+{
+#ifdef HAVE_PCAP_SET_TSTAMP_PRECISION
+ switch (ndo->ndo_tstamp_precision) {
+
+ case PCAP_TSTAMP_PRECISION_MICRO:
+ ND_PRINT(".%06u", (unsigned)usec);
+ break;
+
+ case PCAP_TSTAMP_PRECISION_NANO:
+ ND_PRINT(".%09u", (unsigned)usec);
+ break;
+
+ default:
+ ND_PRINT(".{unknown}");
+ break;
+ }
+#else
+ ND_PRINT(".%06u", (unsigned)usec);
+#endif
+}
+
+/*
+ * Print the timestamp as [YY:MM:DD] HH:MM:SS.FRAC.
+ * if time_flag == LOCAL_TIME print local time else UTC/GMT time
+ * if date_flag == WITH_DATE print YY:MM:DD before HH:MM:SS.FRAC
+ */
+static void
+ts_date_hmsfrac_print(netdissect_options *ndo, long sec, long usec,
+ enum date_flag date_flag, enum time_flag time_flag)
+{
+ time_t Time = sec;
+ struct tm *tm;
+ char timestr[32];
+
+ if ((unsigned)sec & 0x80000000) {
+ ND_PRINT("[Error converting time]");
+ return;
+ }
+
+ if (time_flag == LOCAL_TIME)
+ tm = localtime(&Time);
+ else
+ tm = gmtime(&Time);
+
+ if (!tm) {
+ ND_PRINT("[Error converting time]");
+ return;
+ }
+ if (date_flag == WITH_DATE)
+ strftime(timestr, sizeof(timestr), "%Y-%m-%d %H:%M:%S", tm);
+ else
+ strftime(timestr, sizeof(timestr), "%H:%M:%S", tm);
+ ND_PRINT("%s", timestr);
+
+ ts_frac_print(ndo, usec);
+}
+
+/*
+ * Print the timestamp - Unix timeval style, as SECS.FRAC.
+ */
+static void
+ts_unix_print(netdissect_options *ndo, long sec, long usec)
+{
+ if ((unsigned)sec & 0x80000000) {
+ ND_PRINT("[Error converting time]");
+ return;
+ }
+
+ ND_PRINT("%u", (unsigned)sec);
+ ts_frac_print(ndo, usec);
+}
+
+/*
+ * Print the timestamp
+ */
+void
+ts_print(netdissect_options *ndo,
+ const struct timeval *tvp)
+{
+ static struct timeval tv_ref;
+ struct timeval tv_result;
+ int negative_offset;
+ int nano_prec;
+
+ switch (ndo->ndo_tflag) {
+
+ case 0: /* Default */
+ ts_date_hmsfrac_print(ndo, tvp->tv_sec, tvp->tv_usec,
+ WITHOUT_DATE, LOCAL_TIME);
+ ND_PRINT(" ");
+ break;
+
+ case 1: /* No time stamp */
+ break;
+
+ case 2: /* Unix timeval style */
+ ts_unix_print(ndo, tvp->tv_sec, tvp->tv_usec);
+ ND_PRINT(" ");
+ break;
+
+ case 3: /* Microseconds/nanoseconds since previous packet */
+ case 5: /* Microseconds/nanoseconds since first packet */
+#ifdef HAVE_PCAP_SET_TSTAMP_PRECISION
+ switch (ndo->ndo_tstamp_precision) {
+ case PCAP_TSTAMP_PRECISION_MICRO:
+ nano_prec = 0;
+ break;
+ case PCAP_TSTAMP_PRECISION_NANO:
+ nano_prec = 1;
+ break;
+ default:
+ nano_prec = 0;
+ break;
+ }
+#else
+ nano_prec = 0;
+#endif
+ if (!(netdissect_timevalisset(&tv_ref)))
+ tv_ref = *tvp; /* set timestamp for first packet */
+
+ negative_offset = netdissect_timevalcmp(tvp, &tv_ref, <);
+ if (negative_offset)
+ netdissect_timevalsub(&tv_ref, tvp, &tv_result, nano_prec);
+ else
+ netdissect_timevalsub(tvp, &tv_ref, &tv_result, nano_prec);
+
+ ND_PRINT((negative_offset ? "-" : " "));
+ ts_date_hmsfrac_print(ndo, tv_result.tv_sec, tv_result.tv_usec,
+ WITHOUT_DATE, UTC_TIME);
+ ND_PRINT(" ");
+
+ if (ndo->ndo_tflag == 3)
+ tv_ref = *tvp; /* set timestamp for previous packet */
+ break;
+
+ case 4: /* Date + Default */
+ ts_date_hmsfrac_print(ndo, tvp->tv_sec, tvp->tv_usec,
+ WITH_DATE, LOCAL_TIME);
+ ND_PRINT(" ");
+ break;
+ }
+}
+
+/*
+ * Print an unsigned relative number of seconds (e.g. hold time, prune timer)
+ * in the form 5m1s. This does no truncation, so 32230861 seconds
+ * is represented as 1y1w1d1h1m1s.
+ */
+void
+unsigned_relts_print(netdissect_options *ndo,
+ uint32_t secs)
+{
+ static const char *lengths[] = {"y", "w", "d", "h", "m", "s"};
+ static const u_int seconds[] = {31536000, 604800, 86400, 3600, 60, 1};
+ const char **l = lengths;
+ const u_int *s = seconds;
+
+ if (secs == 0) {
+ ND_PRINT("0s");
+ return;
+ }
+ while (secs > 0) {
+ if (secs >= *s) {
+ ND_PRINT("%u%s", secs / *s, *l);
+ secs -= (secs / *s) * *s;
+ }
+ s++;
+ l++;
+ }
+}
+
+/*
+ * Print a signed relative number of seconds (e.g. hold time, prune timer)
+ * in the form 5m1s. This does no truncation, so 32230861 seconds
+ * is represented as 1y1w1d1h1m1s.
+ */
+void
+signed_relts_print(netdissect_options *ndo,
+ int32_t secs)
+{
+ if (secs < 0) {
+ ND_PRINT("-");
+ if (secs == INT32_MIN) {
+ /*
+ * -2^31; you can't fit its absolute value into
+ * a 32-bit signed integer.
+ *
+ * Just directly pass said absolute value to
+ * unsigned_relts_print() directly.
+ *
+ * (XXX - does ISO C guarantee that -(-2^n),
+ * when calculated and cast to an n-bit unsigned
+ * integer type, will have the value 2^n?)
+ */
+ unsigned_relts_print(ndo, 2147483648U);
+ } else {
+ /*
+ * We now know -secs will fit into an int32_t;
+ * negate it and pass that to unsigned_relts_print().
+ */
+ unsigned_relts_print(ndo, -secs);
+ }
+ return;
+ }
+ unsigned_relts_print(ndo, secs);
+}
+
+/* Print the truncated string */
+void nd_print_trunc(netdissect_options *ndo)
+{
+ ND_PRINT(" [|%s]", ndo->ndo_protocol);
+}
+
+/* Print the protocol name */
+void nd_print_protocol(netdissect_options *ndo)
+{
+ ND_PRINT("%s", ndo->ndo_protocol);
+}
+
+/* Print the protocol name in caps (uppercases) */
+void nd_print_protocol_caps(netdissect_options *ndo)
+{
+ const char *p;
+ for (p = ndo->ndo_protocol; *p != '\0'; p++)
+ ND_PRINT("%c", ND_ASCII_TOUPPER(*p));
+}
+
+/* Print the invalid string */
+void nd_print_invalid(netdissect_options *ndo)
+{
+ ND_PRINT(" (invalid)");
+}
+
+/*
+ * this is a generic routine for printing unknown data;
+ * we pass on the linefeed plus indentation string to
+ * get a proper output - returns 0 on error
+ */
+
+int
+print_unknown_data(netdissect_options *ndo, const u_char *cp,
+ const char *ident, u_int len)
+{
+ u_int len_to_print;
+
+ len_to_print = len;
+ if (!ND_TTEST_LEN(cp, 0)) {
+ ND_PRINT("%sDissector error: print_unknown_data called with pointer past end of packet",
+ ident);
+ return(0);
+ }
+ if (ND_BYTES_AVAILABLE_AFTER(cp) < len_to_print)
+ len_to_print = ND_BYTES_AVAILABLE_AFTER(cp);
+ hex_print(ndo, ident, cp, len_to_print);
+ return(1); /* everything is ok */
+}
+
+/*
+ * Convert a token value to a string; use "fmt" if not found.
+ */
+static const char *
+tok2strbuf(const struct tok *lp, const char *fmt,
+ u_int v, char *buf, size_t bufsize)
+{
+ if (lp != NULL) {
+ while (lp->s != NULL) {
+ if (lp->v == v)
+ return (lp->s);
+ ++lp;
+ }
+ }
+ if (fmt == NULL)
+ fmt = "#%d";
+
+ (void)snprintf(buf, bufsize, fmt, v);
+ return (const char *)buf;
+}
+
+/*
+ * Convert a token value to a string; use "fmt" if not found.
+ * Uses tok2strbuf() on one of four local static buffers of size TOKBUFSIZE
+ * in round-robin fashion.
+ */
+const char *
+tok2str(const struct tok *lp, const char *fmt,
+ u_int v)
+{
+ static char buf[4][TOKBUFSIZE];
+ static int idx = 0;
+ char *ret;
+
+ ret = buf[idx];
+ idx = (idx+1) & 3;
+ return tok2strbuf(lp, fmt, v, ret, sizeof(buf[0]));
+}
+
+/*
+ * Convert a bit token value to a string; use "fmt" if not found.
+ * this is useful for parsing bitfields, the output strings are separated
+ * if the s field is positive.
+ *
+ * A token matches iff it has one or more bits set and every bit that is set
+ * in the token is set in v. Consequently, a 0 token never matches.
+ */
+static char *
+bittok2str_internal(const struct tok *lp, const char *fmt,
+ u_int v, const char *sep)
+{
+ static char buf[1024+1]; /* our string buffer */
+ char *bufp = buf;
+ size_t space_left = sizeof(buf), string_size;
+ const char * sepstr = "";
+
+ while (lp != NULL && lp->s != NULL) {
+ if (lp->v && (v & lp->v) == lp->v) {
+ /* ok we have found something */
+ if (space_left <= 1)
+ return (buf); /* only enough room left for NUL, if that */
+ string_size = strlcpy(bufp, sepstr, space_left);
+ if (string_size >= space_left)
+ return (buf); /* we ran out of room */
+ bufp += string_size;
+ space_left -= string_size;
+ if (space_left <= 1)
+ return (buf); /* only enough room left for NUL, if that */
+ string_size = strlcpy(bufp, lp->s, space_left);
+ if (string_size >= space_left)
+ return (buf); /* we ran out of room */
+ bufp += string_size;
+ space_left -= string_size;
+ sepstr = sep;
+ }
+ lp++;
+ }
+
+ if (bufp == buf)
+ /* bummer - lets print the "unknown" message as advised in the fmt string if we got one */
+ (void)snprintf(buf, sizeof(buf), fmt == NULL ? "#%08x" : fmt, v);
+ return (buf);
+}
+
+/*
+ * Convert a bit token value to a string; use "fmt" if not found.
+ * this is useful for parsing bitfields, the output strings are not separated.
+ */
+char *
+bittok2str_nosep(const struct tok *lp, const char *fmt,
+ u_int v)
+{
+ return (bittok2str_internal(lp, fmt, v, ""));
+}
+
+/*
+ * Convert a bit token value to a string; use "fmt" if not found.
+ * this is useful for parsing bitfields, the output strings are comma separated.
+ */
+char *
+bittok2str(const struct tok *lp, const char *fmt,
+ u_int v)
+{
+ return (bittok2str_internal(lp, fmt, v, ", "));
+}
+
+/*
+ * Convert a value to a string using an array; the macro
+ * tok2strary() in <netdissect.h> is the public interface to
+ * this function and ensures that the second argument is
+ * correct for bounds-checking.
+ */
+const char *
+tok2strary_internal(const char **lp, int n, const char *fmt,
+ int v)
+{
+ static char buf[TOKBUFSIZE];
+
+ if (v >= 0 && v < n && lp[v] != NULL)
+ return lp[v];
+ if (fmt == NULL)
+ fmt = "#%d";
+ (void)snprintf(buf, sizeof(buf), fmt, v);
+ return (buf);
+}
+
+const struct tok *
+uint2tokary_internal(const struct uint_tokary dict[], const size_t size,
+ const u_int val)
+{
+ size_t i;
+ /* Try a direct lookup before the full scan. */
+ if (val < size && dict[val].uintval == val)
+ return dict[val].tokary; /* OK if NULL */
+ for (i = 0; i < size; i++)
+ if (dict[i].uintval == val)
+ return dict[i].tokary; /* OK if NULL */
+ return NULL;
+}
+
+/*
+ * Convert a 32-bit netmask to prefixlen if possible
+ * the function returns the prefix-len; if plen == -1
+ * then conversion was not possible;
+ */
+
+int
+mask2plen(uint32_t mask)
+{
+ const uint32_t bitmasks[33] = {
+ 0x00000000,
+ 0x80000000, 0xc0000000, 0xe0000000, 0xf0000000,
+ 0xf8000000, 0xfc000000, 0xfe000000, 0xff000000,
+ 0xff800000, 0xffc00000, 0xffe00000, 0xfff00000,
+ 0xfff80000, 0xfffc0000, 0xfffe0000, 0xffff0000,
+ 0xffff8000, 0xffffc000, 0xffffe000, 0xfffff000,
+ 0xfffff800, 0xfffffc00, 0xfffffe00, 0xffffff00,
+ 0xffffff80, 0xffffffc0, 0xffffffe0, 0xfffffff0,
+ 0xfffffff8, 0xfffffffc, 0xfffffffe, 0xffffffff
+ };
+ int prefix_len = 32;
+
+ /* let's see if we can transform the mask into a prefixlen */
+ while (prefix_len >= 0) {
+ if (bitmasks[prefix_len] == mask)
+ break;
+ prefix_len--;
+ }
+ return (prefix_len);
+}
+
+int
+mask62plen(const u_char *mask)
+{
+ u_char bitmasks[9] = {
+ 0x00,
+ 0x80, 0xc0, 0xe0, 0xf0,
+ 0xf8, 0xfc, 0xfe, 0xff
+ };
+ int byte;
+ int cidr_len = 0;
+
+ for (byte = 0; byte < 16; byte++) {
+ u_int bits;
+
+ for (bits = 0; bits < (sizeof (bitmasks) / sizeof (bitmasks[0])); bits++) {
+ if (mask[byte] == bitmasks[bits]) {
+ cidr_len += bits;
+ break;
+ }
+ }
+
+ if (mask[byte] != 0xff)
+ break;
+ }
+ return (cidr_len);
+}
+
+/*
+ * Routine to print out information for text-based protocols such as FTP,
+ * HTTP, SMTP, RTSP, SIP, ....
+ */
+#define MAX_TOKEN 128
+
+/*
+ * Fetch a token from a packet, starting at the specified index,
+ * and return the length of the token.
+ *
+ * Returns 0 on error; yes, this is indistinguishable from an empty
+ * token, but an "empty token" isn't a valid token - it just means
+ * either a space character at the beginning of the line (this
+ * includes a blank line) or no more tokens remaining on the line.
+ */
+static int
+fetch_token(netdissect_options *ndo, const u_char *pptr, u_int idx, u_int len,
+ u_char *tbuf, size_t tbuflen)
+{
+ size_t toklen = 0;
+ u_char c;
+
+ for (; idx < len; idx++) {
+ if (!ND_TTEST_1(pptr + idx)) {
+ /* ran past end of captured data */
+ return (0);
+ }
+ c = GET_U_1(pptr + idx);
+ if (!ND_ISASCII(c)) {
+ /* not an ASCII character */
+ return (0);
+ }
+ if (c == ' ' || c == '\t' || c == '\r' || c == '\n') {
+ /* end of token */
+ break;
+ }
+ if (!ND_ASCII_ISPRINT(c)) {
+ /* not part of a command token or response code */
+ return (0);
+ }
+ if (toklen + 2 > tbuflen) {
+ /* no room for this character and terminating '\0' */
+ return (0);
+ }
+ tbuf[toklen] = c;
+ toklen++;
+ }
+ if (toklen == 0) {
+ /* no token */
+ return (0);
+ }
+ tbuf[toklen] = '\0';
+
+ /*
+ * Skip past any white space after the token, until we see
+ * an end-of-line (CR or LF).
+ */
+ for (; idx < len; idx++) {
+ if (!ND_TTEST_1(pptr + idx)) {
+ /* ran past end of captured data */
+ break;
+ }
+ c = GET_U_1(pptr + idx);
+ if (c == '\r' || c == '\n') {
+ /* end of line */
+ break;
+ }
+ if (!ND_ASCII_ISPRINT(c)) {
+ /* not a printable ASCII character */
+ break;
+ }
+ if (c != ' ' && c != '\t' && c != '\r' && c != '\n') {
+ /* beginning of next token */
+ break;
+ }
+ }
+ return (idx);
+}
+
+/*
+ * Scan a buffer looking for a line ending - LF or CR-LF.
+ * Return the index of the character after the line ending or 0 if
+ * we encounter a non-ASCII or non-printable character or don't find
+ * the line ending.
+ */
+static u_int
+print_txt_line(netdissect_options *ndo, const char *prefix,
+ const u_char *pptr, u_int idx, u_int len)
+{
+ u_int startidx;
+ u_int linelen;
+ u_char c;
+
+ startidx = idx;
+ while (idx < len) {
+ c = GET_U_1(pptr + idx);
+ if (c == '\n') {
+ /*
+ * LF without CR; end of line.
+ * Skip the LF and print the line, with the
+ * exception of the LF.
+ */
+ linelen = idx - startidx;
+ idx++;
+ goto print;
+ } else if (c == '\r') {
+ /* CR - any LF? */
+ if ((idx+1) >= len) {
+ /* not in this packet */
+ return (0);
+ }
+ if (GET_U_1(pptr + idx + 1) == '\n') {
+ /*
+ * CR-LF; end of line.
+ * Skip the CR-LF and print the line, with
+ * the exception of the CR-LF.
+ */
+ linelen = idx - startidx;
+ idx += 2;
+ goto print;
+ }
+
+ /*
+ * CR followed by something else; treat this
+ * as if it were binary data, and don't print
+ * it.
+ */
+ return (0);
+ } else if (!ND_ASCII_ISPRINT(c) && c != '\t') {
+ /*
+ * Not a printable ASCII character and not a tab;
+ * treat this as if it were binary data, and
+ * don't print it.
+ */
+ return (0);
+ }
+ idx++;
+ }
+
+ /*
+ * All printable ASCII, but no line ending after that point
+ * in the buffer; treat this as if it were truncated.
+ */
+ linelen = idx - startidx;
+ ND_PRINT("%s%.*s", prefix, (int)linelen, pptr + startidx);
+ nd_print_trunc(ndo);
+ return (0);
+
+print:
+ ND_PRINT("%s%.*s", prefix, (int)linelen, pptr + startidx);
+ return (idx);
+}
+
+/* Assign needed before calling txtproto_print(): ndo->ndo_protocol = "proto" */
+void
+txtproto_print(netdissect_options *ndo, const u_char *pptr, u_int len,
+ const char **cmds, u_int flags)
+{
+ u_int idx, eol;
+ u_char token[MAX_TOKEN+1];
+ const char *cmd;
+ int print_this = 0;
+
+ if (cmds != NULL) {
+ /*
+ * This protocol has more than just request and
+ * response lines; see whether this looks like a
+ * request or response and, if so, print it and,
+ * in verbose mode, print everything after it.
+ *
+ * This is for HTTP-like protocols, where we
+ * want to print requests and responses, but
+ * don't want to print continuations of request
+ * or response bodies in packets that don't
+ * contain the request or response line.
+ */
+ idx = fetch_token(ndo, pptr, 0, len, token, sizeof(token));
+ if (idx != 0) {
+ /* Is this a valid request name? */
+ while ((cmd = *cmds++) != NULL) {
+ if (ascii_strcasecmp((const char *)token, cmd) == 0) {
+ /* Yes. */
+ print_this = 1;
+ break;
+ }
+ }
+
+ /*
+ * No - is this a valid response code (3 digits)?
+ *
+ * Is this token the response code, or is the next
+ * token the response code?
+ */
+ if (flags & RESP_CODE_SECOND_TOKEN) {
+ /*
+ * Next token - get it.
+ */
+ idx = fetch_token(ndo, pptr, idx, len, token,
+ sizeof(token));
+ }
+ if (idx != 0) {
+ if (ND_ASCII_ISDIGIT(token[0]) && ND_ASCII_ISDIGIT(token[1]) &&
+ ND_ASCII_ISDIGIT(token[2]) && token[3] == '\0') {
+ /* Yes. */
+ print_this = 1;
+ }
+ }
+ }
+ } else {
+ /*
+ * Either:
+ *
+ * 1) This protocol has only request and response lines
+ * (e.g., FTP, where all the data goes over a different
+ * connection); assume the payload is a request or
+ * response.
+ *
+ * or
+ *
+ * 2) This protocol is just text, so that we should
+ * always, at minimum, print the first line and,
+ * in verbose mode, print all lines.
+ */
+ print_this = 1;
+ }
+
+ nd_print_protocol_caps(ndo);
+
+ if (print_this) {
+ /*
+ * In non-verbose mode, just print the protocol, followed
+ * by the first line.
+ *
+ * In verbose mode, print lines as text until we run out
+ * of characters or see something that's not a
+ * printable-ASCII line.
+ */
+ if (ndo->ndo_vflag) {
+ /*
+ * We're going to print all the text lines in the
+ * request or response; just print the length
+ * on the first line of the output.
+ */
+ ND_PRINT(", length: %u", len);
+ for (idx = 0;
+ idx < len && (eol = print_txt_line(ndo, "\n\t", pptr, idx, len)) != 0;
+ idx = eol)
+ ;
+ } else {
+ /*
+ * Just print the first text line.
+ */
+ print_txt_line(ndo, ": ", pptr, 0, len);
+ }
+ }
+}
+
+#if (defined(__i386__) || defined(_M_IX86) || defined(__X86__) || defined(__x86_64__) || defined(_M_X64)) || \
+ (defined(__arm__) || defined(_M_ARM) || defined(__aarch64__)) || \
+ (defined(__m68k__) && (!defined(__mc68000__) && !defined(__mc68010__))) || \
+ (defined(__ppc__) || defined(__ppc64__) || defined(_M_PPC) || defined(_ARCH_PPC) || defined(_ARCH_PPC64)) || \
+ (defined(__s390__) || defined(__s390x__) || defined(__zarch__)) || \
+ defined(__vax__)
+/*
+ * The processor natively handles unaligned loads, so just use memcpy()
+ * and memcmp(), to enable those optimizations.
+ *
+ * XXX - are those all the x86 tests we need?
+ * XXX - do we need to worry about ARMv1 through ARMv5, which didn't
+ * support unaligned loads, and, if so, do we need to worry about all
+ * of them, or just some of them, e.g. ARMv5?
+ * XXX - are those the only 68k tests we need not to generated
+ * unaligned accesses if the target is the 68000 or 68010?
+ * XXX - are there any tests we don't need, because some definitions are for
+ * compilers that also predefine the GCC symbols?
+ * XXX - do we need to test for both 32-bit and 64-bit versions of those
+ * architectures in all cases?
+ */
+#else
+/*
+ * The processor doesn't natively handle unaligned loads,
+ * and the compiler might "helpfully" optimize memcpy()
+ * and memcmp(), when handed pointers that would normally
+ * be properly aligned, into sequences that assume proper
+ * alignment.
+ *
+ * Do copies and compares of possibly-unaligned data by
+ * calling routines that wrap memcpy() and memcmp(), to
+ * prevent that optimization.
+ */
+void
+unaligned_memcpy(void *p, const void *q, size_t l)
+{
+ memcpy(p, q, l);
+}
+
+/* As with memcpy(), so with memcmp(). */
+int
+unaligned_memcmp(const void *p, const void *q, size_t l)
+{
+ return (memcmp(p, q, l));
+}
+#endif
+