wireshark/epan/dfilter/semcheck.c

1469 lines
38 KiB
C
Raw Normal View History

/*
* Wireshark - Network traffic analyzer
* By Gerald Combs <gerald@wireshark.org>
* Copyright 2001 Gerald Combs
*
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include "config.h"
#define WS_LOG_DOMAIN LOG_DOMAIN_DFILTER
#include <string.h>
#include "dfilter-int.h"
#include "semcheck.h"
#include "syntax-tree.h"
#include "sttype-range.h"
#include "sttype-test.h"
#include "sttype-set.h"
#include "sttype-function.h"
#include <epan/exceptions.h>
#include <epan/packet.h>
2021-06-18 18:21:42 +00:00
#include <wsutil/ws_assert.h>
#include <wsutil/wslog.h>
2021-06-18 18:21:42 +00:00
#include <ftypes/ftypes.h>
#define FAIL(dfw, ...) \
do { \
ws_noisy("Semantic check failed here."); \
dfilter_fail_throw(dfw, TypeError, __VA_ARGS__); \
} while (0)
static void
semcheck(dfwork_t *dfw, stnode_t *st_node);
static void
check_function(dfwork_t *dfw, stnode_t *st_node);
static ftenum_t
check_bitwise_entity(dfwork_t *dfw, stnode_t *st_node, stnode_t *st_arg, ftenum_t ftype);
static ftenum_t
check_bitwise_operation(dfwork_t *dfw, stnode_t *st_node);
static
ftenum_t
check_arithmetic_operation(dfwork_t *dfw, stnode_t *st_node, ftenum_t lhs_ftype);
static fvalue_t *
mk_fvalue_from_val_string(dfwork_t *dfw, header_field_info *hfinfo, const char *s);
typedef gboolean (*FtypeCanFunc)(enum ftenum);
/* Compares to ftenum_t's and decides if they're
* compatible or not (if they're the same basic type) */
static gboolean
compatible_ftypes(ftenum_t a, ftenum_t b)
{
switch (a) {
case FT_NONE:
case FT_PROTOCOL:
case FT_FLOAT: /* XXX - should be able to compare with INT */
case FT_DOUBLE: /* XXX - should be able to compare with INT */
case FT_ABSOLUTE_TIME:
case FT_RELATIVE_TIME:
case FT_IEEE_11073_SFLOAT:
case FT_IEEE_11073_FLOAT:
case FT_IPv4:
case FT_IPv6:
case FT_IPXNET:
case FT_INT40: /* XXX - should be able to compare with INT */
case FT_UINT40: /* XXX - should be able to compare with INT */
case FT_INT48: /* XXX - should be able to compare with INT */
case FT_UINT48: /* XXX - should be able to compare with INT */
case FT_INT56: /* XXX - should be able to compare with INT */
case FT_UINT56: /* XXX - should be able to compare with INT */
case FT_INT64: /* XXX - should be able to compare with INT */
case FT_UINT64: /* XXX - should be able to compare with INT */
case FT_EUI64: /* XXX - should be able to compare with INT */
return a == b;
case FT_ETHER:
case FT_BYTES:
case FT_UINT_BYTES:
case FT_GUID:
case FT_OID:
case FT_AX25:
case FT_VINES:
case FT_FCWWN:
case FT_REL_OID:
case FT_SYSTEM_ID:
return (b == FT_ETHER || b == FT_BYTES || b == FT_UINT_BYTES || b == FT_GUID || b == FT_OID || b == FT_AX25 || b == FT_VINES || b == FT_FCWWN || b == FT_REL_OID || b == FT_SYSTEM_ID);
case FT_BOOLEAN:
case FT_FRAMENUM:
case FT_CHAR:
case FT_UINT8:
case FT_UINT16:
case FT_UINT24:
case FT_UINT32:
case FT_INT8:
case FT_INT16:
case FT_INT24:
case FT_INT32:
switch (b) {
case FT_BOOLEAN:
case FT_FRAMENUM:
case FT_CHAR:
case FT_UINT8:
case FT_UINT16:
case FT_UINT24:
case FT_UINT32:
case FT_INT8:
case FT_INT16:
case FT_INT24:
case FT_INT32:
return TRUE;
default:
return FALSE;
}
case FT_STRING:
case FT_STRINGZ:
case FT_UINT_STRING:
case FT_STRINGZPAD:
case FT_STRINGZTRUNC:
switch (b) {
case FT_STRING:
case FT_STRINGZ:
case FT_UINT_STRING:
case FT_STRINGZPAD:
case FT_STRINGZTRUNC:
return TRUE;
default:
return FALSE;
}
case FT_NUM_TYPES:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
return FALSE;
}
/* Gets an fvalue from a string, and sets the error message on failure. */
WS_RETNONNULL
static fvalue_t*
dfilter_fvalue_from_literal(dfwork_t *dfw, ftenum_t ftype, stnode_t *st,
gboolean allow_partial_value, header_field_info *hfinfo_value_string)
{
fvalue_t *fv;
const char *s = stnode_data(st);
/* Don't set the error message if it's already set. */
fv = fvalue_from_literal(ftype, s, allow_partial_value,
dfw->error_message == NULL ? &dfw->error_message : NULL);
if (fv == NULL && hfinfo_value_string) {
/* check value_string */
fv = mk_fvalue_from_val_string(dfw, hfinfo_value_string, s);
/*
* Ignore previous errors if this can be mapped
* to an item from value_string.
*/
if (fv && dfw->error_message) {
g_free(dfw->error_message);
dfw->error_message = NULL;
}
}
if (fv == NULL)
THROW(TypeError);
return fv;
}
/* Gets an fvalue from a string, and sets the error message on failure. */
WS_RETNONNULL
static fvalue_t *
dfilter_fvalue_from_string(dfwork_t *dfw, ftenum_t ftype, stnode_t *st,
header_field_info *hfinfo_value_string)
{
fvalue_t *fv;
const char *s = stnode_data(st);
fv = fvalue_from_string(ftype, s,
dfw->error_message == NULL ? &dfw->error_message : NULL);
if (fv == NULL && hfinfo_value_string) {
fv = mk_fvalue_from_val_string(dfw, hfinfo_value_string, s);
/*
* Ignore previous errors if this can be mapped
* to an item from value_string.
*/
if (fv && dfw->error_message) {
g_free(dfw->error_message);
dfw->error_message = NULL;
}
}
if (fv == NULL)
THROW(TypeError);
return fv;
}
static gboolean
resolve_unparsed(dfwork_t *dfw, stnode_t *st)
{
if (stnode_type_id(st) != STTYPE_UNPARSED)
return FALSE;
header_field_info *hfinfo = dfilter_resolve_unparsed(dfw, stnode_data(st));
if (hfinfo != NULL) {
stnode_replace(st, STTYPE_FIELD, hfinfo);
return TRUE;
}
stnode_replace(st, STTYPE_LITERAL, g_strdup(stnode_data(st)));
return FALSE;
}
/* Creates a FT_UINT32 fvalue with a given value. */
static fvalue_t*
mk_uint32_fvalue(guint32 val)
{
fvalue_t *fv;
fv = fvalue_new(FT_UINT32);
fvalue_set_uinteger(fv, val);
return fv;
}
/* Creates a FT_UINT64 fvalue with a given value. */
static fvalue_t*
mk_uint64_fvalue(guint64 val)
{
fvalue_t *fv;
fv = fvalue_new(FT_UINT64);
fvalue_set_uinteger64(fv, val);
return fv;
}
/* Try to make an fvalue from a string using a value_string or true_false_string.
* This works only for ftypes that are integers. Returns the created fvalue_t*
* or NULL if impossible. */
static fvalue_t*
mk_fvalue_from_val_string(dfwork_t *dfw, header_field_info *hfinfo, const char *s)
{
static const true_false_string default_tf = { "True", "False" };
const true_false_string *tf = &default_tf;
/* Early return? */
switch(hfinfo->type) {
case FT_NONE:
case FT_PROTOCOL:
case FT_FLOAT:
case FT_DOUBLE:
case FT_IEEE_11073_SFLOAT:
case FT_IEEE_11073_FLOAT:
case FT_ABSOLUTE_TIME:
case FT_RELATIVE_TIME:
case FT_IPv4:
case FT_IPv6:
case FT_IPXNET:
case FT_AX25:
case FT_VINES:
case FT_FCWWN:
case FT_ETHER:
case FT_BYTES:
case FT_UINT_BYTES:
case FT_STRING:
case FT_STRINGZ:
case FT_UINT_STRING:
case FT_STRINGZPAD:
case FT_STRINGZTRUNC:
case FT_EUI64:
case FT_GUID:
case FT_OID:
case FT_REL_OID:
case FT_SYSTEM_ID:
case FT_FRAMENUM: /* hfinfo->strings contains ft_framenum_type_t, not strings */
return NULL;
case FT_BOOLEAN:
case FT_CHAR:
case FT_UINT8:
case FT_UINT16:
case FT_UINT24:
case FT_UINT32:
case FT_UINT40:
case FT_UINT48:
case FT_UINT56:
case FT_UINT64:
case FT_INT8:
case FT_INT16:
case FT_INT24:
case FT_INT32:
case FT_INT40:
case FT_INT48:
case FT_INT56:
case FT_INT64:
break;
case FT_NUM_TYPES:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
/* TRUE/FALSE *always* exist for FT_BOOLEAN. */
if (hfinfo->type == FT_BOOLEAN) {
if (hfinfo->strings) {
tf = (const true_false_string *)hfinfo->strings;
}
if (g_ascii_strcasecmp(s, tf->true_string) == 0) {
return mk_uint64_fvalue(TRUE);
}
else if (g_ascii_strcasecmp(s, tf->false_string) == 0) {
return mk_uint64_fvalue(FALSE);
}
else {
/*
* Prefer this error message to whatever error message
* has already been set.
*/
g_free(dfw->error_message);
dfw->error_message = NULL;
dfilter_fail(dfw, "\"%s\" cannot be found among the possible values for %s.",
s, hfinfo->abbrev);
return NULL;
}
}
/* Do val_strings exist? */
if (!hfinfo->strings) {
dfilter_fail(dfw, "%s cannot accept strings as values.",
hfinfo->abbrev);
return NULL;
}
/* Reset the error message, since *something* interesting will happen,
* and the error message will be more interesting than any error message
* I happen to have now. */
g_free(dfw->error_message);
dfw->error_message = NULL;
if (hfinfo->display & BASE_RANGE_STRING) {
dfilter_fail(dfw, "\"%s\" cannot accept [range] strings as values.",
hfinfo->abbrev);
}
else if (hfinfo->display & BASE_VAL64_STRING) {
const val64_string *vals = (const val64_string *)hfinfo->strings;
while (vals->strptr != NULL) {
if (g_ascii_strcasecmp(s, vals->strptr) == 0) {
return mk_uint64_fvalue(vals->value);
}
vals++;
}
dfilter_fail(dfw, "\"%s\" cannot be found among the possible values for %s.",
s, hfinfo->abbrev);
}
else if (hfinfo->display == BASE_CUSTOM) {
/* If a user wants to match against a custom string, we would
* somehow have to have the integer value here to pass it in
* to the custom-display function. But we don't have an
* integer, we have the string they're trying to match.
* -><-
*/
dfilter_fail(dfw, "\"%s\" cannot accept [custom] strings as values.",
hfinfo->abbrev);
}
else {
const value_string *vals = (const value_string *)hfinfo->strings;
if (hfinfo->display & BASE_EXT_STRING)
vals = VALUE_STRING_EXT_VS_P((const value_string_ext *) vals);
while (vals->strptr != NULL) {
if (g_ascii_strcasecmp(s, vals->strptr) == 0) {
return mk_uint32_fvalue(vals->value);
}
vals++;
}
dfilter_fail(dfw, "\"%s\" cannot be found among the possible values for %s.",
s, hfinfo->abbrev);
}
return NULL;
}
static gboolean
is_bytes_type(enum ftenum type)
{
switch(type) {
case FT_AX25:
case FT_VINES:
case FT_FCWWN:
case FT_ETHER:
case FT_BYTES:
case FT_UINT_BYTES:
case FT_IPv6:
case FT_GUID:
case FT_OID:
case FT_REL_OID:
case FT_SYSTEM_ID:
return TRUE;
case FT_NONE:
case FT_PROTOCOL:
case FT_FLOAT:
case FT_DOUBLE:
case FT_IEEE_11073_SFLOAT:
case FT_IEEE_11073_FLOAT:
case FT_ABSOLUTE_TIME:
case FT_RELATIVE_TIME:
case FT_IPv4:
case FT_IPXNET:
case FT_STRING:
case FT_STRINGZ:
case FT_UINT_STRING:
case FT_STRINGZPAD:
case FT_STRINGZTRUNC:
case FT_BOOLEAN:
case FT_FRAMENUM:
case FT_CHAR:
case FT_UINT8:
case FT_UINT16:
case FT_UINT24:
case FT_UINT32:
case FT_UINT40:
case FT_UINT48:
case FT_UINT56:
case FT_UINT64:
case FT_INT8:
case FT_INT16:
case FT_INT24:
case FT_INT32:
case FT_INT40:
case FT_INT48:
case FT_INT56:
case FT_INT64:
case FT_EUI64:
return FALSE;
case FT_NUM_TYPES:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
return FALSE;
}
/* Check the semantics of an existence test. */
static void
check_exists(dfwork_t *dfw, stnode_t *st_arg1)
{
2022-02-27 14:11:50 +00:00
LOG_NODE(st_arg1);
resolve_unparsed(dfw, st_arg1);
switch (stnode_type_id(st_arg1)) {
case STTYPE_FIELD:
/* This is OK */
break;
case STTYPE_STRING:
case STTYPE_UNPARSED:
case STTYPE_LITERAL:
case STTYPE_CHARCONST:
case STTYPE_ARITHMETIC:
FAIL(dfw, "%s is neither a field nor a protocol name.",
stnode_todisplay(st_arg1));
break;
case STTYPE_RANGE:
/*
* XXX - why not? Shouldn't "eth[3:2]" mean
* "check whether the 'eth' field is present and
* has at least 2 bytes starting at an offset of
* 3"?
*/
FAIL(dfw, "You cannot test whether a range is present.");
break;
case STTYPE_FUNCTION:
/* XXX - Maybe we should change functions so they can return fields,
* in which case the 'exist' should be fine. */
FAIL(dfw, "You cannot test whether a function is present.");
break;
2021-11-12 19:43:17 +00:00
default:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
static void
check_drange_sanity(dfwork_t *dfw, stnode_t *st)
{
stnode_t *entity1;
header_field_info *hfinfo1;
ftenum_t ftype1;
2022-02-27 14:11:50 +00:00
LOG_NODE(st);
entity1 = sttype_range_entity(st);
2021-11-08 01:05:53 +00:00
ws_assert(entity1);
resolve_unparsed(dfw, entity1);
2021-11-08 01:05:53 +00:00
if (stnode_type_id(entity1) == STTYPE_FIELD) {
hfinfo1 = stnode_data(entity1);
ftype1 = hfinfo1->type;
if (!ftype_can_slice(ftype1)) {
FAIL(dfw, "\"%s\" is a %s and cannot be sliced into a sequence of bytes.",
hfinfo1->abbrev, ftype_pretty_name(ftype1));
}
2021-11-08 01:05:53 +00:00
} else if (stnode_type_id(entity1) == STTYPE_FUNCTION) {
df_func_def_t *funcdef = sttype_function_funcdef(entity1);
ftype1 = funcdef->retval_ftype;
if (!ftype_can_slice(ftype1)) {
FAIL(dfw, "Return value of function \"%s\" is a %s and cannot be converted into a sequence of bytes.",
funcdef->name, ftype_pretty_name(ftype1));
}
check_function(dfw, entity1);
2021-11-08 01:05:53 +00:00
} else if (stnode_type_id(entity1) == STTYPE_RANGE) {
/* Should this be rejected instead? */
check_drange_sanity(dfw, entity1);
2021-11-08 01:05:53 +00:00
} else {
FAIL(dfw, "Range is not supported for entity %s",
stnode_todisplay(entity1));
}
}
static void
convert_to_bytes(stnode_t *arg)
{
stnode_t *entity1;
drange_node *rn;
entity1 = stnode_dup(arg);
rn = drange_node_new();
drange_node_set_start_offset(rn, 0);
drange_node_set_to_the_end(rn);
stnode_replace(arg, STTYPE_RANGE, NULL);
sttype_range_set1(arg, entity1, rn);
}
static void
check_function(dfwork_t *dfw, stnode_t *st_node)
{
df_func_def_t *funcdef;
GSList *params;
guint iparam;
guint nparams;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
funcdef = sttype_function_funcdef(st_node);
params = sttype_function_params(st_node);
nparams = g_slist_length(params);
if (nparams < funcdef->min_nargs) {
FAIL(dfw, "Function %s needs at least %u arguments.",
funcdef->name, funcdef->min_nargs);
} else if (nparams > funcdef->max_nargs) {
FAIL(dfw, "Function %s can only accept %u arguments.",
funcdef->name, funcdef->max_nargs);
}
iparam = 0;
while (params) {
resolve_unparsed(dfw, params->data);
funcdef->semcheck_param_function(dfw, funcdef->name, iparam, params->data);
params = params->next;
iparam++;
}
}
WS_RETNONNULL
static fvalue_t *
dfilter_fvalue_from_charconst(dfwork_t *dfw, ftenum_t ftype, stnode_t *st)
{
fvalue_t *fvalue;
unsigned long *nump = stnode_data(st);
fvalue = fvalue_from_charconst(ftype, *nump,
dfw->error_message == NULL ? &dfw->error_message : NULL);
if (fvalue == NULL)
THROW(TypeError);
return fvalue;
}
/* If the LHS of a relation test is a FIELD, run some checks
* and possibly some modifications of syntax tree nodes. */
static void
check_relation_LHS_FIELD(dfwork_t *dfw, test_op_t st_op,
FtypeCanFunc can_func, gboolean allow_partial_value,
stnode_t *st_node,
stnode_t *st_arg1, stnode_t *st_arg2)
{
sttype_id_t type2;
header_field_info *hfinfo1, *hfinfo2;
df_func_def_t *funcdef;
ftenum_t ftype1, ftype2;
fvalue_t *fvalue;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
again:
type2 = stnode_type_id(st_arg2);
hfinfo1 = stnode_data(st_arg1);
ftype1 = hfinfo1->type;
if (!can_func(ftype1)) {
FAIL(dfw, "%s (type=%s) cannot participate in %s comparison.",
hfinfo1->abbrev, ftype_pretty_name(ftype1),
stnode_todisplay(st_node));
}
if (type2 == STTYPE_FIELD) {
hfinfo2 = stnode_data(st_arg2);
ftype2 = hfinfo2->type;
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
hfinfo1->abbrev, hfinfo2->abbrev);
}
/* Do this check even though you'd think that if
* they're compatible, then can_func() would pass. */
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
hfinfo2->abbrev, ftype_pretty_name(ftype2));
}
}
else if (type2 == STTYPE_STRING || type2 == STTYPE_LITERAL || type2 == STTYPE_UNPARSED) {
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
/* Skip incompatible fields */
while (hfinfo1->same_name_prev_id != -1 &&
((type2 == STTYPE_STRING && ftype1 != FT_STRING && ftype1!= FT_STRINGZ) ||
(type2 != STTYPE_STRING && (ftype1 == FT_STRING || ftype1== FT_STRINGZ)))) {
hfinfo1 = proto_registrar_get_nth(hfinfo1->same_name_prev_id);
ftype1 = hfinfo1->type;
}
if (type2 == STTYPE_UNPARSED) {
if (resolve_unparsed(dfw, st_arg2))
goto again;
fvalue = dfilter_fvalue_from_literal(dfw, ftype1, st_arg2, allow_partial_value, hfinfo1);
}
else if (type2 == STTYPE_STRING) {
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
fvalue = dfilter_fvalue_from_string(dfw, ftype1, st_arg2, hfinfo1);
}
else {
fvalue = dfilter_fvalue_from_literal(dfw, ftype1, st_arg2, allow_partial_value, hfinfo1);
}
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_CHARCONST) {
fvalue = dfilter_fvalue_from_charconst(dfw, ftype1, st_arg2);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_RANGE) {
check_drange_sanity(dfw, st_arg2);
if (!is_bytes_type(ftype1)) {
if (!ftype_can_slice(ftype1)) {
FAIL(dfw, "\"%s\" is a %s and cannot be converted into a sequence of bytes.",
hfinfo1->abbrev,
ftype_pretty_name(ftype1));
}
/* Convert entire field to bytes */
convert_to_bytes(st_arg1);
}
}
else if (type2 == STTYPE_FUNCTION) {
funcdef = sttype_function_funcdef(st_arg2);
ftype2 = funcdef->retval_ftype;
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s (type=%s) and return value of %s() (type=%s) are not of compatible types.",
hfinfo1->abbrev, ftype_pretty_name(ftype1),
funcdef->name, ftype_pretty_name(ftype2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "return value of %s() (type=%s) cannot participate in specified comparison.",
funcdef->name, ftype_pretty_name(ftype2));
}
check_function(dfw, st_arg2);
}
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
else if (type2 == STTYPE_PCRE) {
ws_assert(st_op == TEST_OP_MATCHES);
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
}
else if (type2 == STTYPE_BITWISE) {
ftype2 = check_bitwise_operation(dfw, st_arg2);
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else if (type2 == STTYPE_ARITHMETIC) {
ftype2 = check_arithmetic_operation(dfw, st_arg2, ftype1);
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else {
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
static void
check_relation_LHS_RANGE(dfwork_t *dfw, test_op_t st_op,
FtypeCanFunc can_func _U_,
gboolean allow_partial_value,
stnode_t *st_node _U_,
stnode_t *st_arg1, stnode_t *st_arg2)
{
sttype_id_t type2;
header_field_info *hfinfo2;
ftenum_t ftype2;
fvalue_t *fvalue;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
check_drange_sanity(dfw, st_arg1);
again:
type2 = stnode_type_id(st_arg2);
if (type2 == STTYPE_FIELD) {
hfinfo2 = stnode_data(st_arg2);
ftype2 = hfinfo2->type;
if (!is_bytes_type(ftype2)) {
if (!ftype_can_slice(ftype2)) {
FAIL(dfw, "\"%s\" is a %s and cannot be converted into a sequence of bytes.",
hfinfo2->abbrev,
ftype_pretty_name(ftype2));
}
/* Convert entire field to bytes */
convert_to_bytes(st_arg2);
}
}
else if (type2 == STTYPE_STRING) {
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
fvalue = dfilter_fvalue_from_string(dfw, FT_BYTES, st_arg2, NULL);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_UNPARSED) {
if (resolve_unparsed(dfw, st_arg2))
goto again;
fvalue = dfilter_fvalue_from_literal(dfw, FT_BYTES, st_arg2, allow_partial_value, NULL);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_LITERAL) {
fvalue = dfilter_fvalue_from_literal(dfw, FT_BYTES, st_arg2, allow_partial_value, NULL);
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_CHARCONST) {
fvalue = dfilter_fvalue_from_charconst(dfw, FT_BYTES, st_arg2);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_RANGE) {
check_drange_sanity(dfw, st_arg2);
}
else if (type2 == STTYPE_FUNCTION) {
df_func_def_t *funcdef = sttype_function_funcdef(st_arg2);
ftype2 = funcdef->retval_ftype;
if (!is_bytes_type(ftype2)) {
if (!ftype_can_slice(ftype2)) {
FAIL(dfw, "Return value of function \"%s\" is a %s and cannot be converted into a sequence of bytes.",
funcdef->name,
ftype_pretty_name(ftype2));
}
/* Convert function result to bytes */
convert_to_bytes(st_arg2);
}
check_function(dfw, st_arg2);
}
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
else if (type2 == STTYPE_PCRE) {
ws_assert(st_op == TEST_OP_MATCHES);
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
}
else if (type2 == STTYPE_BITWISE) {
ftype2 = check_bitwise_operation(dfw, st_arg2);
if (!compatible_ftypes(FT_BYTES, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else if (type2 == STTYPE_ARITHMETIC) {
ftype2 = check_arithmetic_operation(dfw, st_arg2, FT_BYTES);
if (!compatible_ftypes(FT_BYTES, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else {
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
/* If the LHS of a relation test is a FUNCTION, run some checks
* and possibly some modifications of syntax tree nodes. */
static void
check_relation_LHS_FUNCTION(dfwork_t *dfw, test_op_t st_op,
FtypeCanFunc can_func, gboolean allow_partial_value,
stnode_t *st_node,
stnode_t *st_arg1, stnode_t *st_arg2)
{
sttype_id_t type2;
header_field_info *hfinfo2;
ftenum_t ftype1, ftype2;
fvalue_t *fvalue;
2021-03-21 09:15:00 +00:00
df_func_def_t *funcdef;
df_func_def_t *funcdef2;
/* GSList *params; */
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
check_function(dfw, st_arg1);
funcdef = sttype_function_funcdef(st_arg1);
ftype1 = funcdef->retval_ftype;
if (!can_func(ftype1)) {
FAIL(dfw, "Function %s (type=%s) cannot participate in %s comparison.",
funcdef->name, ftype_pretty_name(ftype1),
stnode_todisplay(st_node));
}
again:
type2 = stnode_type_id(st_arg2);
if (type2 == STTYPE_FIELD) {
hfinfo2 = stnode_data(st_arg2);
ftype2 = hfinfo2->type;
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "Function %s and %s are not of compatible types.",
funcdef->name, hfinfo2->abbrev);
}
/* Do this check even though you'd think that if
* they're compatible, then can_func() would pass. */
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
hfinfo2->abbrev, ftype_pretty_name(ftype2));
}
}
else if (type2 == STTYPE_STRING) {
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
fvalue = dfilter_fvalue_from_string(dfw, ftype1, st_arg2, NULL);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_UNPARSED) {
if (resolve_unparsed(dfw, st_arg2))
goto again;
fvalue = dfilter_fvalue_from_literal(dfw, ftype1, st_arg2, allow_partial_value, NULL);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_LITERAL) {
fvalue = dfilter_fvalue_from_literal(dfw, ftype1, st_arg2, allow_partial_value, NULL);
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_CHARCONST) {
fvalue = dfilter_fvalue_from_charconst(dfw, ftype1, st_arg2);
stnode_replace(st_arg2, STTYPE_FVALUE, fvalue);
}
else if (type2 == STTYPE_RANGE) {
check_drange_sanity(dfw, st_arg2);
if (!is_bytes_type(ftype1)) {
if (!ftype_can_slice(ftype1)) {
FAIL(dfw, "Function \"%s\" is a %s and cannot be converted into a sequence of bytes.",
funcdef->name,
ftype_pretty_name(ftype1));
}
/* Convert function result to bytes */
convert_to_bytes(st_arg1);
}
}
else if (type2 == STTYPE_FUNCTION) {
funcdef2 = sttype_function_funcdef(st_arg2);
ftype2 = funcdef2->retval_ftype;
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "Return values of function %s (type=%s) and function %s (type=%s) are not of compatible types.",
funcdef->name, ftype_pretty_name(ftype1), funcdef2->name, ftype_pretty_name(ftype2));
}
/* Do this check even though you'd think that if
* they're compatible, then can_func() would pass. */
if (!can_func(ftype2)) {
FAIL(dfw, "Return value of %s (type=%s) cannot participate in specified comparison.",
funcdef2->name, ftype_pretty_name(ftype2));
}
check_function(dfw, st_arg2);
}
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
else if (type2 == STTYPE_PCRE) {
ws_assert(st_op == TEST_OP_MATCHES);
dfilter: Require double-quoted strings with "matches" Matches is a special case that looks on the RHS and tries to convert every unparsed value to a string, regardless of the LHS type. This is not how types work in the display filter. Require double-quotes to avoid ambiguity, because matches doesn't follow normal Wireshark display filter type rules. It doesn't need nor benefit from the flexibility provided by unparsed strings in the syntax. For matches the RHS is always a literal strings except if the RHS is also a field name, then it complains of an incompatible type. This is confusing. No type can be compatible because no type rules are ever considered. Every unparsed value is a text string except if it happens to coincide with a field name it also requires double-quoting or it throws a syntax error, just to be difficult. We could remove this odd quirk but requiring double-quotes for regular expressions is a better, more elegant fix. Before: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp.srcport dftest: tcp and udp.srcport are not of compatible types. Filter: tcp matches udp.srcportt Constants: 00000 PUT_PCRE udp.srcportt -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN After: Filter: tcp matches "udp" Constants: 00000 PUT_PCRE udp -> reg#1 Instructions: 00000 READ_TREE tcp -> reg#0 00001 IF-FALSE-GOTO 3 00002 ANY_MATCHES reg#0 matches reg#1 00003 RETURN Filter: tcp matches udp dftest: "udp" was unexpected in this context. Filter: tcp matches udp.srcport dftest: "udp.srcport" was unexpected in this context. Filter: tcp matches udp.srcportt dftest: "udp.srcportt" was unexpected in this context. The error message could still be improved.
2021-10-09 15:40:08 +00:00
}
else if (type2 == STTYPE_BITWISE) {
ftype2 = check_bitwise_operation(dfw, st_arg2);
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else if (type2 == STTYPE_ARITHMETIC) {
ftype2 = check_arithmetic_operation(dfw, st_arg2, ftype1);
if (!compatible_ftypes(ftype1, ftype2)) {
FAIL(dfw, "%s and %s are not of compatible types.",
stnode_todisplay(st_arg1), stnode_todisplay(st_arg2));
}
if (!can_func(ftype2)) {
FAIL(dfw, "%s (type=%s) cannot participate in specified comparison.",
stnode_todisplay(st_arg2), ftype_pretty_name(ftype2));
}
}
else {
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
static void
check_relation_LHS_BITWISE(dfwork_t *dfw, test_op_t st_op _U_,
FtypeCanFunc can_func _U_, gboolean allow_partial_value,
stnode_t *st_node _U_,
stnode_t *st_arg1, stnode_t *st_arg2)
{
stnode_t *bitwise_entity;
sttype_id_t bitwise_entity_type;
LOG_NODE(st_node);
/* (arg1: bitwise) <relop> (arg2: rhs) */
check_bitwise_operation(dfw, st_arg1);
sttype_test_get(st_arg1, NULL, &bitwise_entity, NULL);
bitwise_entity_type = stnode_type_id(bitwise_entity);
if (bitwise_entity_type == STTYPE_FIELD) {
check_relation_LHS_FIELD(dfw, st_op, can_func, allow_partial_value, st_node, bitwise_entity, st_arg2);
}
else if (bitwise_entity_type == STTYPE_RANGE) {
check_relation_LHS_RANGE(dfw, st_op, can_func, allow_partial_value, st_node, bitwise_entity, st_arg2);
}
else if (bitwise_entity_type == STTYPE_FUNCTION) {
check_relation_LHS_FUNCTION(dfw, st_op, can_func, allow_partial_value, st_node, bitwise_entity, st_arg2);
}
else {
ws_assert_not_reached();
}
}
static void
check_relation_LHS_ARITHMETIC(dfwork_t *dfw, test_op_t st_op _U_,
FtypeCanFunc can_func _U_, gboolean allow_partial_value,
stnode_t *st_node _U_,
stnode_t *st_arg1, stnode_t *st_arg2)
{
stnode_t *entity;
sttype_id_t entity_type;
LOG_NODE(st_node);
sttype_test_get(st_arg1, NULL, &entity, NULL);
entity_type = stnode_type_id(entity);
if (entity_type == STTYPE_FIELD) {
check_arithmetic_operation(dfw, st_arg1, FT_NONE);
check_relation_LHS_FIELD(dfw, st_op, can_func, allow_partial_value, st_node, entity, st_arg2);
}
else if (entity_type == STTYPE_FUNCTION) {
check_arithmetic_operation(dfw, st_arg1, FT_NONE);
check_relation_LHS_FUNCTION(dfw, st_op, can_func, allow_partial_value, st_node, entity, st_arg2);
}
else {
FAIL(dfw, "Left side of %s expression must be a field or function, not %s.",
stnode_todisplay(st_node), stnode_todisplay(entity));
}
}
/* Check the semantics of any relational test. */
static void
check_relation(dfwork_t *dfw, test_op_t st_op,
FtypeCanFunc can_func, gboolean allow_partial_value,
stnode_t *st_node, stnode_t *st_arg1, stnode_t *st_arg2)
{
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
resolve_unparsed(dfw, st_arg1);
switch (stnode_type_id(st_arg1)) {
case STTYPE_FIELD:
check_relation_LHS_FIELD(dfw, st_op, can_func,
allow_partial_value, st_node, st_arg1, st_arg2);
break;
case STTYPE_RANGE:
check_relation_LHS_RANGE(dfw, st_op, can_func,
allow_partial_value, st_node, st_arg1, st_arg2);
break;
case STTYPE_FUNCTION:
check_relation_LHS_FUNCTION(dfw, st_op, can_func,
allow_partial_value, st_node, st_arg1, st_arg2);
break;
case STTYPE_BITWISE:
check_relation_LHS_BITWISE(dfw, st_op, can_func,
allow_partial_value, st_node, st_arg1, st_arg2);
break;
case STTYPE_ARITHMETIC:
check_relation_LHS_ARITHMETIC(dfw, st_op, can_func,
allow_partial_value, st_node, st_arg1, st_arg2);
break;
default:
FAIL(dfw, "Left side of %s expression must be a field or function, not %s.",
stnode_todisplay(st_node), stnode_todisplay(st_arg1));
}
}
static void
check_relation_contains(dfwork_t *dfw, stnode_t *st_node,
stnode_t *st_arg1, stnode_t *st_arg2)
{
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
resolve_unparsed(dfw, st_arg1);
switch (stnode_type_id(st_arg1)) {
case STTYPE_FIELD:
check_relation_LHS_FIELD(dfw, TEST_OP_CONTAINS, ftype_can_contains,
TRUE, st_node, st_arg1, st_arg2);
break;
case STTYPE_FUNCTION:
check_relation_LHS_FUNCTION(dfw, TEST_OP_CONTAINS, ftype_can_contains,
TRUE, st_node, st_arg1, st_arg2);
break;
case STTYPE_RANGE:
check_relation_LHS_RANGE(dfw, TEST_OP_CONTAINS, ftype_can_contains,
TRUE, st_node, st_arg1, st_arg2);
break;
default:
FAIL(dfw, "Left side of %s expression must be a field or function, not %s.",
stnode_todisplay(st_node), stnode_todisplay(st_arg1));
}
}
static void
check_relation_matches(dfwork_t *dfw, stnode_t *st_node,
stnode_t *st_arg1, stnode_t *st_arg2)
{
2021-11-12 15:55:14 +00:00
ws_regex_t *pcre;
char *errmsg = NULL;
const char *patt;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
resolve_unparsed(dfw, st_arg1);
if (stnode_type_id(st_arg2) != STTYPE_STRING) {
FAIL(dfw, "Matches requires a double quoted string on the right side.");
}
patt = stnode_data(st_arg2);
ws_debug("Compile regex pattern: %s", patt);
2021-11-12 15:55:14 +00:00
pcre = ws_regex_compile(patt, &errmsg);
if (errmsg) {
dfilter_fail(dfw, "Regex compilation error: %s.", errmsg);
g_free(errmsg);
THROW(TypeError);
}
stnode_replace(st_arg2, STTYPE_PCRE, pcre);
switch (stnode_type_id(st_arg1)) {
case STTYPE_FIELD:
check_relation_LHS_FIELD(dfw, TEST_OP_MATCHES, ftype_can_matches,
TRUE, st_node, st_arg1, st_arg2);
break;
case STTYPE_FUNCTION:
check_relation_LHS_FUNCTION(dfw, TEST_OP_MATCHES, ftype_can_matches,
TRUE, st_node, st_arg1, st_arg2);
break;
case STTYPE_RANGE:
check_relation_LHS_RANGE(dfw, TEST_OP_MATCHES, ftype_can_matches,
TRUE, st_node, st_arg1, st_arg2);
break;
default:
FAIL(dfw, "Left side of %s expression must be a field or function, not %s.",
stnode_todisplay(st_node), stnode_todisplay(st_arg1));
}
}
static void
check_relation_in(dfwork_t *dfw, stnode_t *st_node _U_,
stnode_t *st_arg1, stnode_t *st_arg2)
{
GSList *nodelist;
stnode_t *node_left, *node_right;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
resolve_unparsed(dfw, st_arg1);
if (stnode_type_id(st_arg1) != STTYPE_FIELD) {
FAIL(dfw, "Only a field may be tested for membership in a set.");
}
/* Checked in the grammar parser. */
ws_assert(stnode_type_id(st_arg2) == STTYPE_SET);
/* Attempt to interpret one element of the set at a time. Each
* element is represented by two items in the list, the element
* value and NULL. Both will be replaced by a lower and upper
* value if the element is a range. */
nodelist = stnode_data(st_arg2);
while (nodelist) {
node_left = nodelist->data;
/* Don't let a range on the RHS affect the LHS field. */
if (stnode_type_id(node_left) == STTYPE_RANGE) {
FAIL(dfw, "A range may not appear inside a set.");
break;
}
nodelist = g_slist_next(nodelist);
ws_assert(nodelist);
node_right = nodelist->data;
if (node_right) {
check_relation_LHS_FIELD(dfw, TEST_OP_GE, ftype_can_cmp,
FALSE, st_node, st_arg1, node_left);
check_relation_LHS_FIELD(dfw, TEST_OP_LE, ftype_can_cmp,
FALSE, st_node, st_arg1, node_right);
} else {
check_relation_LHS_FIELD(dfw, TEST_OP_ANY_EQ, ftype_can_eq,
FALSE, st_node, st_arg1, node_left);
}
nodelist = g_slist_next(nodelist);
}
}
/* Check the semantics of any type of TEST */
static void
check_test(dfwork_t *dfw, stnode_t *st_node)
{
test_op_t st_op;
stnode_t *st_arg1, *st_arg2;
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
sttype_test_get(st_node, &st_op, &st_arg1, &st_arg2);
switch (st_op) {
case TEST_OP_UNINITIALIZED:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
break;
case TEST_OP_EXISTS:
check_exists(dfw, st_arg1);
break;
case TEST_OP_NOTZERO:
ws_assert(stnode_type_id(st_arg1) == STTYPE_BITWISE);
check_bitwise_operation(dfw, st_arg1);
break;
case TEST_OP_NOT:
semcheck(dfw, st_arg1);
break;
case TEST_OP_AND:
case TEST_OP_OR:
semcheck(dfw, st_arg1);
semcheck(dfw, st_arg2);
break;
case TEST_OP_ALL_EQ:
dfilter: Fix "!=" relation to be free of contradictions Wireshark defines the relation of equality A == B as A any_eq B <=> An == Bn for at least one An, Bn. More accurately I think this is (formally) an equivalence relation, not true equality. Whichever definition for "==" we choose we must keep the definition of "!=" as !(A == B), otherwise it will lead to logical contradictions like (A == B) AND (A != B) being true. Fix the '!=' relation to match the definition of equality: A != B <=> !(A == B) <=> A all_ne B <=> An != Bn, for every n. This has been the recomended way to write "not equal" for a long time in the documentation, even to the point where != was deprecated, but it just wasn't implemented consistently in the language, which has understandably been a persistent source of confusion. Even a field that is normally well-behaved with "!=" like "ip.src" or "ip.dst" will produce unexpected results with encapsulations like IP-over-IP. The opcode ALL_NE could have been implemented in the compiler instead using NOT and ANY_EQ but I chose to implement it in bytecode. It just seemed more elegant and efficient but the difference was not very significant. Keep around "~=" for any_ne relation, in case someone depends on that, and because we don't have an operator for true equality: A strict_equal B <=> A all_eq B <=> !(A any_ne B). If there is only one value then any_ne and all_ne are the same comparison operation. Implementing this change did not require fixing any tests so it is unlikely the relation "~=" (any_ne) will be very useful. Note that the behaviour of the '<' (less than) comparison relation is a separate, more subtle issue. In the general case the definition of '<' that is used is only a partial order.
2021-10-18 20:07:06 +00:00
case TEST_OP_ANY_EQ:
case TEST_OP_ALL_NE:
case TEST_OP_ANY_NE:
check_relation(dfw, st_op, ftype_can_eq, FALSE, st_node, st_arg1, st_arg2);
dfilter: Fix "!=" relation to be free of contradictions Wireshark defines the relation of equality A == B as A any_eq B <=> An == Bn for at least one An, Bn. More accurately I think this is (formally) an equivalence relation, not true equality. Whichever definition for "==" we choose we must keep the definition of "!=" as !(A == B), otherwise it will lead to logical contradictions like (A == B) AND (A != B) being true. Fix the '!=' relation to match the definition of equality: A != B <=> !(A == B) <=> A all_ne B <=> An != Bn, for every n. This has been the recomended way to write "not equal" for a long time in the documentation, even to the point where != was deprecated, but it just wasn't implemented consistently in the language, which has understandably been a persistent source of confusion. Even a field that is normally well-behaved with "!=" like "ip.src" or "ip.dst" will produce unexpected results with encapsulations like IP-over-IP. The opcode ALL_NE could have been implemented in the compiler instead using NOT and ANY_EQ but I chose to implement it in bytecode. It just seemed more elegant and efficient but the difference was not very significant. Keep around "~=" for any_ne relation, in case someone depends on that, and because we don't have an operator for true equality: A strict_equal B <=> A all_eq B <=> !(A any_ne B). If there is only one value then any_ne and all_ne are the same comparison operation. Implementing this change did not require fixing any tests so it is unlikely the relation "~=" (any_ne) will be very useful. Note that the behaviour of the '<' (less than) comparison relation is a separate, more subtle issue. In the general case the definition of '<' that is used is only a partial order.
2021-10-18 20:07:06 +00:00
break;
case TEST_OP_GT:
case TEST_OP_GE:
case TEST_OP_LT:
case TEST_OP_LE:
check_relation(dfw, st_op, ftype_can_cmp, FALSE, st_node, st_arg1, st_arg2);
break;
case TEST_OP_CONTAINS:
check_relation_contains(dfw, st_node, st_arg1, st_arg2);
break;
case TEST_OP_MATCHES:
check_relation_matches(dfw, st_node, st_arg1, st_arg2);
break;
case TEST_OP_IN:
check_relation_in(dfw, st_node, st_arg1, st_arg2);
break;
default:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
static ftenum_t
check_bitwise_entity(dfwork_t *dfw, stnode_t *st_node, stnode_t *st_arg, ftenum_t lhs_ftype)
{
header_field_info *hfinfo;
ftenum_t ftype;
fvalue_t *fvalue;
sttype_id_t type;
df_func_def_t *funcdef;
LOG_NODE(st_arg);
resolve_unparsed(dfw, st_arg);
type = stnode_type_id(st_arg);
if (type == STTYPE_FIELD) {
hfinfo = stnode_data(st_arg);
ftype = hfinfo->type;
if (!ftype_can_bitwise_and(ftype)) {
FAIL(dfw, "%s (type=%s) cannot participate in %s comparison.",
hfinfo->abbrev, ftype_pretty_name(ftype),
stnode_todisplay(st_node));
}
return ftype;
}
else if (type == STTYPE_FUNCTION) {
check_function(dfw, st_arg);
funcdef = sttype_function_funcdef(st_arg);
ftype = funcdef->retval_ftype;
if (!ftype_can_bitwise_and(ftype)) {
FAIL(dfw, "Function %s (type=%s) cannot participate in %s comparison.",
funcdef->name, ftype_pretty_name(ftype),
stnode_todisplay(st_node));
}
return ftype;
}
else if (type == STTYPE_RANGE) {
check_drange_sanity(dfw, st_arg);
ftype = FT_BYTES;
if (!ftype_can_bitwise_and(ftype)) {
FAIL(dfw, "Range %s (type=%s) cannot participate in %s comparison.",
stnode_todisplay(st_arg), ftype_pretty_name(ftype),
stnode_todisplay(st_node));
}
return ftype;
}
else if (lhs_ftype != FT_NONE) {
/* numeric constant */
ftype = lhs_ftype;
fvalue = dfilter_fvalue_from_literal(dfw, ftype, st_arg, FALSE, NULL);
stnode_replace(st_arg, STTYPE_FVALUE, fvalue);
}
else {
FAIL(dfw, "Invalid bitwise operand %s in comparison %s.",
stnode_todisplay(st_arg), stnode_todisplay(st_arg));
}
return ftype;
}
static ftenum_t
check_bitwise_operation(dfwork_t *dfw, stnode_t *st_node)
{
test_op_t st_op;
stnode_t *st_arg1, *st_arg2;
ftenum_t ftype1, ftype2;
LOG_NODE(st_node);
sttype_test_get(st_node, &st_op, &st_arg1, &st_arg2);
ftype1 = check_bitwise_entity(dfw, st_node, st_arg1, FT_NONE);
ftype2 = check_bitwise_entity(dfw, st_node, st_arg2, ftype1);
if (!ftype_can_is_zero(ftype1)) {
FAIL(dfw, "Cannot test if %s is true", stnode_todisplay(st_arg1));
}
/* XXX This could be relaxed with type promotions. */
if (ftype1 != ftype2) {
FAIL(dfw, "%s %s %s: entities have different types",
stnode_todisplay(st_arg1),
stnode_todisplay(st_node),
stnode_todisplay(st_arg2));
}
return ftype1;
}
ftenum_t
check_arithmetic_operation(dfwork_t *dfw, stnode_t *st_node, ftenum_t lhs_ftype)
{
test_op_t st_op;
stnode_t *st_arg;
sttype_id_t type;
ftenum_t ftype;
FtypeCanFunc can_func = NULL;
sttype_test_get(st_node, &st_op, &st_arg, NULL);
resolve_unparsed(dfw, st_arg);
type = stnode_type_id(st_arg);
switch (st_op) {
case OP_UNARY_MINUS:
can_func = ftype_can_unary_minus;
break;
default:
ws_assert_not_reached();
}
if (type == STTYPE_LITERAL) {
/* numeric constant */
ws_assert(lhs_ftype != FT_NONE);
fvalue_t *fvalue = dfilter_fvalue_from_literal(dfw, lhs_ftype, st_arg, FALSE, NULL);
stnode_replace(st_arg, STTYPE_FVALUE, fvalue);
ftype = fvalue_type_ftenum(fvalue);
if (!can_func(ftype)) {
FAIL(dfw, "%s (%s) cannot be negative.",
stnode_todisplay(st_arg),
fvalue_type_name(fvalue));
}
if (st_op == OP_UNARY_MINUS) {
/* Pre-compute constant unary minus result */
char *err_msg;
fvalue_t *new_fv = fvalue_unary_minus(fvalue, &err_msg);
if (new_fv == NULL) {
dfilter_fail(dfw, "%s: %s", stnode_todisplay(st_arg), err_msg);
g_free(err_msg);
THROW(TypeError);
}
/* Replaces unary operator with result */
stnode_replace(st_node, STTYPE_FVALUE, new_fv);
}
}
else if (type == STTYPE_FIELD) {
header_field_info *hfinfo = stnode_data(st_arg);
ftype = hfinfo->type;
if (!can_func(ftype)) {
FAIL(dfw, "%s (type=%s) cannot be negative.",
hfinfo->abbrev, ftype_pretty_name(ftype));
}
}
else if (type == STTYPE_FUNCTION) {
check_function(dfw, st_arg);
df_func_def_t *funcdef = sttype_function_funcdef(st_arg);
ftype = funcdef->retval_ftype;
if (!can_func(ftype)) {
FAIL(dfw, "Function %s (type=%s) cannot be negative.",
funcdef->name, ftype_pretty_name(ftype));
}
}
else {
FAIL(dfw, "%s cannot be a negative value.",
stnode_todisplay(st_arg));
}
return ftype;
}
/* Check the entire syntax tree. */
static void
semcheck(dfwork_t *dfw, stnode_t *st_node)
{
2022-02-27 14:11:50 +00:00
LOG_NODE(st_node);
/* The parser assures that the top-most syntax-tree
* node will be a TEST node, no matter what. So assert that. */
switch (stnode_type_id(st_node)) {
case STTYPE_TEST:
check_test(dfw, st_node);
break;
default:
2021-06-18 18:21:42 +00:00
ws_assert_not_reached();
}
}
/* Check the syntax tree for semantic errors, and convert
* some of the nodes into the form they need to be in order to
* later generate the DFVM bytecode. */
gboolean
dfw_semcheck(dfwork_t *dfw)
{
volatile gboolean ok_filter = TRUE;
ws_noisy("Starting semantic check (dfw = %p)", dfw);
/* Instead of having to check for errors at every stage of
* the semantic-checking, the semantic-checking code will
* throw an exception if a problem is found. */
TRY {
semcheck(dfw, dfw->st_root);
}
CATCH(TypeError) {
ok_filter = FALSE;
}
ENDTRY;
ws_noisy("Semantic check (dfw = %p) returns %s",
dfw, ok_filter ? "TRUE" : "FALSE");
return ok_filter;
}
/*
* Editor modelines - https://www.wireshark.org/tools/modelines.html
*
* Local variables:
* c-basic-offset: 8
* tab-width: 8
* indent-tabs-mode: t
* End:
*
* vi: set shiftwidth=8 tabstop=8 noexpandtab:
* :indentSize=8:tabSize=8:noTabs=false:
*/