osmocom
/
wireshark
11
0
Fork 1
Code Issues Pull Requests Projects Releases Wiki Activity
wireshark/wsutil/json_dumper.c

688 lines
20 KiB
C
Raw Blame History

/* json_dumper.c
* Routines for serializing data as JSON.
*
* Copyright 2018, Peter Wu <peter@lekensteyn.nl>
* Copyright (C) 2016 Jakub Zawadzki
*
* Wireshark - Network traffic analyzer
* By Gerald Combs <gerald@wireshark.org>
* Copyright 1998 Gerald Combs
*
* SPDX-License-Identifier: GPL-2.0-or-later
*/
#include "json_dumper.h"
#define WS_LOG_DOMAIN LOG_DOMAIN_WSUTIL
#include <math.h>
#include <wsutil/wslog.h>
/*
* json_dumper.state[current_depth] describes a nested element:
* - type: none/object/array/non-base64 value/base64 value
* - has_name: Whether the object member name was set.
*
* (A base64 value isn't really a nested element, but that's a
* convenient way of handling them, with a begin call that opens
* the string with a double-quote, one or more calls to convert
* raw bytes to base64 and add them to the value, and an end call
* that finishes the base64 encoding, adds any remaining raw bytes
* in base64 encoding, and closes the string with a double-quote.)
*/
enum json_dumper_element_type {
JSON_DUMPER_TYPE_NONE = 0,
JSON_DUMPER_TYPE_VALUE = 1,
JSON_DUMPER_TYPE_OBJECT = 2,
JSON_DUMPER_TYPE_ARRAY = 3,
JSON_DUMPER_TYPE_BASE64 = 4,
};
#define JSON_DUMPER_TYPE(state) ((enum json_dumper_element_type)((state) & 7))
#define JSON_DUMPER_HAS_NAME (1 << 3)
static const char *json_dumper_element_type_names[] = {
[JSON_DUMPER_TYPE_NONE] = "none",
[JSON_DUMPER_TYPE_VALUE] = "value",
[JSON_DUMPER_TYPE_OBJECT] = "object",
[JSON_DUMPER_TYPE_ARRAY] = "array",
[JSON_DUMPER_TYPE_BASE64] = "base64"
};
#define NUM_JSON_DUMPER_ELEMENT_TYPE_NAMES (sizeof json_dumper_element_type_names / sizeof json_dumper_element_type_names[0])
#define JSON_DUMPER_FLAGS_ERROR (1 << 16) /* Output flag: an error occurred. */
enum json_dumper_change {
JSON_DUMPER_BEGIN,
JSON_DUMPER_END,
JSON_DUMPER_SET_NAME,
JSON_DUMPER_SET_VALUE,
JSON_DUMPER_WRITE_BASE64,
JSON_DUMPER_FINISH,
};
/* JSON Dumper putc */
static void
jd_putc(const json_dumper *dumper, char c)
{
if (dumper->output_file) {
fputc(c, dumper->output_file);
}
if (dumper->output_string) {
g_string_append_c(dumper->output_string, c);
}
}
/* JSON Dumper puts */
static void
jd_puts(const json_dumper *dumper, const char *s)
{
if (dumper->output_file) {
fputs(s, dumper->output_file);
}
if (dumper->output_string) {
g_string_append(dumper->output_string, s);
}
}
static void
jd_puts_len(const json_dumper *dumper, const char *s, gsize len)
{
if (dumper->output_file) {
fwrite(s, 1, len, dumper->output_file);
}
if (dumper->output_string) {
g_string_append_len(dumper->output_string, s, len);
}
}
static void
jd_vprintf(const json_dumper *dumper, const char *format, va_list args)
{
if (dumper->output_file) {
vfprintf(dumper->output_file, format, args);
}
if (dumper->output_string) {
g_string_append_vprintf(dumper->output_string, format, args);
}
}
static void
json_puts_string(const json_dumper *dumper, const char *str, gboolean dot_to_underscore)
{
if (!str) {
jd_puts(dumper, "null");
return;
}
static const char json_cntrl[0x20][6] = {
"u0000", "u0001", "u0002", "u0003", "u0004", "u0005", "u0006", "u0007", "b", "t", "n", "u000b", "f", "r", "u000e", "u000f",
"u0010", "u0011", "u0012", "u0013", "u0014", "u0015", "u0016", "u0017", "u0018", "u0019", "u001a", "u001b", "u001c", "u001d", "u001e", "u001f"
};
jd_putc(dumper, '"');
for (int i = 0; str[i]; i++) {
if ((guint)str[i] < 0x20) {
jd_putc(dumper, '\\');
jd_puts(dumper, json_cntrl[(guint)str[i]]);
} else if (i > 0 && str[i - 1] == '<' && str[i] == '/') {
// Convert </script> to <\/script> to avoid breaking web pages.
jd_puts(dumper, "\\/");
} else {
if (str[i] == '\\' || str[i] == '"') {
jd_putc(dumper, '\\');
}
if (dot_to_underscore && str[i] == '.')
jd_putc(dumper, '_');
else
jd_putc(dumper, str[i]);
}
}
jd_putc(dumper, '"');
}
static inline guint8
json_dumper_get_prev_state(json_dumper *dumper)
{
guint depth = dumper->current_depth;
return depth != 0 ? dumper->state[depth - 1] : 0;
}
static inline guint8
json_dumper_get_curr_state(json_dumper *dumper)
{
guint depth = dumper->current_depth;
return dumper->state[depth];
}
/**
* Called when a programming error is encountered where the JSON manipulation
* state got corrupted. This could happen when pairing the wrong begin/end
* calls, when writing multiple values for the same object, etc.
*/
static void
json_dumper_bad(json_dumper *dumper, const char *what)
{
dumper->flags |= JSON_DUMPER_FLAGS_ERROR;
if ((dumper->flags & JSON_DUMPER_FLAGS_NO_DEBUG)) {
/* Console output can be slow, disable log calls to speed up fuzzing. */
/*
* XXX - should this call abort()? If that flag isn't set,
* ws_error() wou;d call it; is there any point in continuing
* to do anything if we get here when fuzzing?
*/
return;
}
if (dumper->output_file) {
fflush(dumper->output_file);
}
char unknown_curr_type_name[10+1];
char unknown_prev_type_name[10+1];
const char *curr_type_name, *prev_type_name;
guint8 curr_state = json_dumper_get_curr_state(dumper);
guint8 curr_type = JSON_DUMPER_TYPE(curr_state);
if (curr_type < NUM_JSON_DUMPER_ELEMENT_TYPE_NAMES) {
curr_type_name = json_dumper_element_type_names[curr_type];
} else {
snprintf(unknown_curr_type_name, sizeof unknown_curr_type_name, "%u", curr_type);
curr_type_name = unknown_curr_type_name;
}
if (dumper->current_depth != 0) {
guint8 prev_state = json_dumper_get_prev_state(dumper);
guint8 prev_type = JSON_DUMPER_TYPE(prev_state);
if (prev_type < NUM_JSON_DUMPER_ELEMENT_TYPE_NAMES) {
prev_type_name = json_dumper_element_type_names[prev_type];
} else {
snprintf(unknown_prev_type_name, sizeof unknown_prev_type_name, "%u", prev_type);
prev_type_name = unknown_prev_type_name;
}
} else {
prev_type_name = "(none)";
}
ws_error("json_dumper error: %s: current stack depth %u, current type %s, previous_type %s",
what, dumper->current_depth, curr_type_name, prev_type_name);
/* NOTREACHED */
}
static inline gboolean
json_dumper_stack_would_overflow(json_dumper *dumper)
{
if (dumper->current_depth + 1 >= JSON_DUMPER_MAX_DEPTH) {
json_dumper_bad(dumper, "JSON dumper stack overflow");
return TRUE;
}
return FALSE;
}
static inline gboolean
json_dumper_stack_would_underflow(json_dumper *dumper)
{
if (dumper->current_depth == 0) {
json_dumper_bad(dumper, "JSON dumper stack underflow");
return TRUE;
}
return FALSE;
}
/**
* Checks that the dumper has not already had an error. Fail, and
* return FALSE, to tell our caller not to do any more work, if it
* has.
*/
static gboolean
json_dumper_check_previous_error(json_dumper *dumper)
{
if ((dumper->flags & JSON_DUMPER_FLAGS_ERROR)) {
json_dumper_bad(dumper, "previous corruption detected");
return FALSE;
}
return TRUE;
}
static void
print_newline_indent(const json_dumper *dumper, guint depth)
{
if ((dumper->flags & JSON_DUMPER_FLAGS_PRETTY_PRINT)) {
jd_putc(dumper, '\n');
for (guint i = 0; i < depth; i++) {
jd_puts(dumper, " ");
}
}
}
/**
* Prints commas, newlines and indentation (if necessary). Used for array
* values, object names and normal values (strings, etc.).
*/
static void
prepare_token(json_dumper *dumper)
{
if (dumper->current_depth == 0) {
// not part of an array or object.
return;
}
guint8 prev_state = dumper->state[dumper->current_depth - 1];
// While processing the object value, reset the key state as it is consumed.
dumper->state[dumper->current_depth - 1] &= ~JSON_DUMPER_HAS_NAME;
switch (JSON_DUMPER_TYPE(prev_state)) {
case JSON_DUMPER_TYPE_OBJECT:
if ((prev_state & JSON_DUMPER_HAS_NAME)) {
// Object key already set, value follows. No indentation needed.
return;
}
break;
case JSON_DUMPER_TYPE_ARRAY:
break;
default:
// Initial values do not need indentation.
return;
}
guint8 curr_state = json_dumper_get_curr_state(dumper);
if (curr_state != JSON_DUMPER_TYPE_NONE) {
jd_putc(dumper, ',');
}
print_newline_indent(dumper, dumper->current_depth);
}
/**
* Common code to open an object/array/base64 value, printing
* an opening character.
*
* It also makes various correctness checks.
*/
static gboolean
json_dumper_begin_nested_element(json_dumper *dumper, enum json_dumper_element_type type)
{
if (!json_dumper_check_previous_error(dumper)) {
return FALSE;
}
/* Make sure we won't overflow the dumper stack */
if (json_dumper_stack_would_overflow(dumper)) {
return FALSE;
}
prepare_token(dumper);
switch (type) {
case JSON_DUMPER_TYPE_OBJECT:
jd_putc(dumper, '{');
break;
case JSON_DUMPER_TYPE_ARRAY:
jd_putc(dumper, '[');
break;
case JSON_DUMPER_TYPE_BASE64:
dumper->base64_state = 0;
dumper->base64_save = 0;
jd_putc(dumper, '"');
break;
default:
json_dumper_bad(dumper, "beginning unknown nested element type");
return FALSE;
}
dumper->state[dumper->current_depth] = type;
/*
* Guaranteed not to overflow, as json_dumper_stack_would_overflow()
* returned FALSE.
*/
++dumper->current_depth;
dumper->state[dumper->current_depth] = JSON_DUMPER_TYPE_NONE;
return TRUE;
}
/**
* Common code to close an object/array/base64 value, printing a
* closing character (and if necessary, it is preceded by newline
* and indentation).
*
* It also makes various correctness checks.
*/
static gboolean
json_dumper_end_nested_element(json_dumper *dumper, enum json_dumper_element_type type)
{
if (!json_dumper_check_previous_error(dumper)) {
return FALSE;
}
guint8 prev_state = json_dumper_get_prev_state(dumper);
switch (type) {
case JSON_DUMPER_TYPE_OBJECT:
if (JSON_DUMPER_TYPE(prev_state) != JSON_DUMPER_TYPE_OBJECT) {
json_dumper_bad(dumper, "ending non-object nested item type as object");
return FALSE;
}
break;
case JSON_DUMPER_TYPE_ARRAY:
if (JSON_DUMPER_TYPE(prev_state) != JSON_DUMPER_TYPE_ARRAY) {
json_dumper_bad(dumper, "ending non-array nested item type as array");
return FALSE;
}
break;
case JSON_DUMPER_TYPE_BASE64:
if (JSON_DUMPER_TYPE(prev_state) != JSON_DUMPER_TYPE_BASE64) {
json_dumper_bad(dumper, "ending non-base64 nested item type as base64");
return FALSE;
}
break;
default:
json_dumper_bad(dumper, "ending unknown nested element type");
return FALSE;
}
if (prev_state & JSON_DUMPER_HAS_NAME) {
json_dumper_bad(dumper, "finishing object with last item having name but no value");
return FALSE;
}
/* Make sure we won't underflow the dumper stack */
if (json_dumper_stack_would_underflow(dumper)) {
return FALSE;
}
// if the object/array was non-empty, add a newline and indentation.
if (dumper->state[dumper->current_depth]) {
print_newline_indent(dumper, dumper->current_depth - 1);
}
switch (type) {
case JSON_DUMPER_TYPE_OBJECT:
jd_putc(dumper, '}');
break;
case JSON_DUMPER_TYPE_ARRAY:
jd_putc(dumper, ']');
break;
case JSON_DUMPER_TYPE_BASE64:
{
gchar buf[4];
gsize wrote;
wrote = g_base64_encode_close(FALSE, buf, &dumper->base64_state, &dumper->base64_save);
jd_puts_len(dumper, buf, wrote);
jd_putc(dumper, '"');
break;
}
default:
json_dumper_bad(dumper, "endning unknown nested element type");
return FALSE;
}
/*
* Guaranteed not to underflow, as json_dumper_stack_would_underflow()
* returned FALSE.
*/
--dumper->current_depth;
return TRUE;
}
void
json_dumper_begin_object(json_dumper *dumper)
{
json_dumper_begin_nested_element(dumper, JSON_DUMPER_TYPE_OBJECT);
}
void
json_dumper_set_member_name(json_dumper *dumper, const char *name)
{
if (!json_dumper_check_previous_error(dumper)) {
return;
}
guint8 prev_state = json_dumper_get_prev_state(dumper);
/* Only object members, not array members, have names. */
if (JSON_DUMPER_TYPE(prev_state) != JSON_DUMPER_TYPE_OBJECT) {
json_dumper_bad(dumper, "setting name on non-object nested item type");
return;
}
/* An object member name can only be set once before its value is set. */
if (prev_state & JSON_DUMPER_HAS_NAME) {
json_dumper_bad(dumper, "setting name twice on an object member");
return;
}
prepare_token(dumper);
json_puts_string(dumper, name, dumper->flags & JSON_DUMPER_DOT_TO_UNDERSCORE);
jd_putc(dumper, ':');
if ((dumper->flags & JSON_DUMPER_FLAGS_PRETTY_PRINT)) {
jd_putc(dumper, ' ');
}
dumper->state[dumper->current_depth - 1] |= JSON_DUMPER_HAS_NAME;
}
void
json_dumper_end_object(json_dumper *dumper)
{
json_dumper_end_nested_element(dumper, JSON_DUMPER_TYPE_OBJECT);
}
void
json_dumper_begin_array(json_dumper *dumper)
{
json_dumper_begin_nested_element(dumper, JSON_DUMPER_TYPE_ARRAY);
}
void
json_dumper_end_array(json_dumper *dumper)
{
json_dumper_end_nested_element(dumper, JSON_DUMPER_TYPE_ARRAY);
}
static gboolean
json_dumper_setting_value_ok(json_dumper *dumper)
{
guint8 prev_state = json_dumper_get_prev_state(dumper);
switch (JSON_DUMPER_TYPE(prev_state)) {
case JSON_DUMPER_TYPE_OBJECT:
/*
* This value is part of an object. As such, it must
* have a name.
*/
if (!(prev_state & JSON_DUMPER_HAS_NAME)) {
json_dumper_bad(dumper, "setting value of object member without a name");
return FALSE;
}
break;
case JSON_DUMPER_TYPE_ARRAY:
/*
* This value is part of an array. As such, it's not
* required to have a name (and shouldn't have a name;
* that's already been checked in json_dumper_set_member_name()).
*/
break;
case JSON_DUMPER_TYPE_BASE64:
/*
* We're in the middle of constructing a base64-encoded
* value. Only json_dumper_write_base64() can be used
* for that; we can't add individual values to it.
*/
json_dumper_bad(dumper, "attempt to set value of base64 item to something not base64-encoded");
return FALSE;
case JSON_DUMPER_TYPE_NONE:
case JSON_DUMPER_TYPE_VALUE:
{
guint8 curr_state = json_dumper_get_curr_state(dumper);
switch (JSON_DUMPER_TYPE(curr_state)) {
case JSON_DUMPER_TYPE_NONE:
/*
* We haven't put a value yet, so we can put one now.
*/
break;
case JSON_DUMPER_TYPE_VALUE:
/*
* This value isn't part of an object or array,
* and we've already put one value.
*/
json_dumper_bad(dumper, "value not in object or array immediately follows another value");
return FALSE;
case JSON_DUMPER_TYPE_OBJECT:
case JSON_DUMPER_TYPE_ARRAY:
case JSON_DUMPER_TYPE_BASE64:
/*
* This should never be the case, no matter what
* our callers do:
*
* JSON_DUMPER_TYPE_OBJECT can be the previous
* type, meaning we're in the process of adding
* elements to an object, but it should never be
* the current type;
*
* JSON_DUMPER_TYPE_ARRAY can be the previous
* type, meaning we're in the process of adding
* elements to an array, but it should never be
* the current type;
*
* JSON_DUMPER_TYPE_BASE64 should only be the
* current type if we're in the middle of
* building a base64 value, in which case the
* previous type should also be JSON_DUMPER_TYPE_BASE64,
* but that's not the previous type.
*/
json_dumper_bad(dumper, "internal error setting value - should not happen");
return FALSE;
default:
json_dumper_bad(dumper, "internal error setting value, bad current state - should not happen");
return FALSE;
}
break;
}
default:
json_dumper_bad(dumper, "internal error setting value, bad previous state - should not happen");
return FALSE;
}
return TRUE;
}
void
json_dumper_value_string(json_dumper *dumper, const char *value)
{
if (!json_dumper_check_previous_error(dumper)) {
return;
}
if (!json_dumper_setting_value_ok(dumper)) {
return;
}
prepare_token(dumper);
json_puts_string(dumper, value, FALSE);
dumper->state[dumper->current_depth] = JSON_DUMPER_TYPE_VALUE;
}
void
json_dumper_value_double(json_dumper *dumper, double value)
{
if (!json_dumper_check_previous_error(dumper)) {
return;
}
if (!json_dumper_setting_value_ok(dumper)) {
return;
}
prepare_token(dumper);
gchar buffer[G_ASCII_DTOSTR_BUF_SIZE] = { 0 };
if (isfinite(value) && g_ascii_dtostr(buffer, G_ASCII_DTOSTR_BUF_SIZE, value) && buffer[0]) {
jd_puts(dumper, buffer);
} else {
jd_puts(dumper, "null");
}
dumper->state[dumper->current_depth] = JSON_DUMPER_TYPE_VALUE;
}
void
json_dumper_value_va_list(json_dumper *dumper, const char *format, va_list ap)
{
if (!json_dumper_check_previous_error(dumper)) {
return;
}
if (!json_dumper_setting_value_ok(dumper)) {
return;
}
prepare_token(dumper);
jd_vprintf(dumper, format, ap);
dumper->state[dumper->current_depth] = JSON_DUMPER_TYPE_VALUE;
}
void
json_dumper_value_anyf(json_dumper *dumper, const char *format, ...)
{
va_list ap;
va_start(ap, format);
json_dumper_value_va_list(dumper, format, ap);
va_end(ap);
}
gboolean
json_dumper_finish(json_dumper *dumper)
{
if (!json_dumper_check_previous_error(dumper)) {
return FALSE;
}
if (dumper->current_depth != 0) {
json_dumper_bad(dumper, "JSON dumper stack not empty at finish");
return FALSE;
}
jd_putc(dumper, '\n');
dumper->state[0] = JSON_DUMPER_TYPE_NONE;
return TRUE;
}
void
json_dumper_begin_base64(json_dumper *dumper)
{
json_dumper_begin_nested_element(dumper, JSON_DUMPER_TYPE_BASE64);
}
void
json_dumper_write_base64(json_dumper* dumper, const guchar *data, size_t len)
{
if (!json_dumper_check_previous_error(dumper)) {
return;
}
guint8 prev_state = json_dumper_get_prev_state(dumper);
if (JSON_DUMPER_TYPE(prev_state) != JSON_DUMPER_TYPE_BASE64) {
json_dumper_bad(dumper, "writing base64 data to a non-base64 value");
return;
}
#define CHUNK_SIZE 1024
gchar buf[(CHUNK_SIZE / 3 + 1) * 4 + 4];
while (len > 0) {
gsize chunk_size = len < CHUNK_SIZE ? len : CHUNK_SIZE;
gsize output_size = g_base64_encode_step(data, chunk_size, FALSE, buf, &dumper->base64_state, &dumper->base64_save);
jd_puts_len(dumper, buf, output_size);
data += chunk_size;
len -= chunk_size;
}
dumper->state[dumper->current_depth] = JSON_DUMPER_TYPE_BASE64;
}
void
json_dumper_end_base64(json_dumper *dumper)
{
json_dumper_end_nested_element(dumper, JSON_DUMPER_TYPE_BASE64);
}
Reference in New Issue View Git Blame Copy Permalink