jsmn_validate puts a maximum number of tokens allowed. Add a function to call jsmn_parse without calling strlen (so it both works without a null terminated string and is more efficient when we already know the length) and call that instead. Add a comment that we might want to change jsmn_validate to just set a maximum length, instead of allocating memory for an array of tokens that are filled out and then freed. Even if jsmn_validate is changed, the new function would still be useful for parsing partial JSON, whether truncated files or if the heuristics decided to make a decision after less than an entire file. Fix #17589
116 lines
3.1 KiB
C
116 lines
3.1 KiB
C
/* json.c
|
|
*
|
|
* Copyright 2015, Dario Lombardo <lomato@gmail.com>
|
|
*
|
|
* SPDX-License-Identifier: GPL-2.0-or-later
|
|
*/
|
|
|
|
#include "config.h"
|
|
#include "json.h"
|
|
|
|
#include <string.h>
|
|
|
|
#include "wtap-int.h"
|
|
#include "file_wrappers.h"
|
|
|
|
#include <wsutil/wsjson.h>
|
|
|
|
/* Maximum size of json file. */
|
|
#define MAX_FILE_SIZE (50*1024*1024)
|
|
|
|
static int json_file_type_subtype = -1;
|
|
|
|
void register_json(void);
|
|
|
|
wtap_open_return_val json_open(wtap *wth, int *err, char **err_info)
|
|
{
|
|
uint8_t* filebuf;
|
|
int bytes_read;
|
|
|
|
/* XXX checking the full file contents might be a bit expensive, maybe
|
|
* resort to simpler heuristics like '{' or '[' (with some other chars)? */
|
|
filebuf = (uint8_t*)g_malloc0(MAX_FILE_SIZE);
|
|
if (!filebuf)
|
|
return WTAP_OPEN_ERROR;
|
|
|
|
bytes_read = file_read(filebuf, MAX_FILE_SIZE, wth->fh);
|
|
if (bytes_read < 0) {
|
|
/* Read error. */
|
|
*err = file_error(wth->fh, err_info);
|
|
g_free(filebuf);
|
|
return WTAP_OPEN_ERROR;
|
|
}
|
|
if (bytes_read == 0) {
|
|
/* empty file, not *anybody's* */
|
|
g_free(filebuf);
|
|
return WTAP_OPEN_NOT_MINE;
|
|
}
|
|
|
|
/* We could reduce the maximum size to read and accept if the parser
|
|
* returns JSMN_ERROR_PART (i.e., only fail on JSMN_ERROR_INVAL as we
|
|
* shouldn't get JSMN_ERROR_NOMEM if tokens is NULL.) That way we
|
|
* could handle bigger files without testing the entire file.
|
|
* packet-json shows excess unparsed data at the end with the
|
|
* data-text-lines dissector.
|
|
*/
|
|
if (json_parse_len(filebuf, bytes_read, NULL, 0) < 0) {
|
|
g_free(filebuf);
|
|
return WTAP_OPEN_NOT_MINE;
|
|
}
|
|
|
|
if (file_seek(wth->fh, 0, SEEK_SET, err) == -1) {
|
|
g_free(filebuf);
|
|
return WTAP_OPEN_ERROR;
|
|
}
|
|
|
|
wth->file_type_subtype = json_file_type_subtype;
|
|
wth->file_encap = WTAP_ENCAP_JSON;
|
|
wth->file_tsprec = WTAP_TSPREC_SEC;
|
|
wth->subtype_read = wtap_full_file_read;
|
|
wth->subtype_seek_read = wtap_full_file_seek_read;
|
|
wth->snapshot_length = 0;
|
|
|
|
g_free(filebuf);
|
|
return WTAP_OPEN_MINE;
|
|
}
|
|
|
|
static const struct supported_block_type json_blocks_supported[] = {
|
|
/*
|
|
* This is a file format that we dissect, so we provide only one
|
|
* "packet" with the file's contents, and don't support any
|
|
* options.
|
|
*/
|
|
{ WTAP_BLOCK_PACKET, ONE_BLOCK_SUPPORTED, NO_OPTIONS_SUPPORTED }
|
|
};
|
|
|
|
static const struct file_type_subtype_info json_info = {
|
|
"JavaScript Object Notation", "json", "json", NULL,
|
|
false, BLOCKS_SUPPORTED(json_blocks_supported),
|
|
NULL, NULL, NULL
|
|
};
|
|
|
|
void register_json(void)
|
|
{
|
|
json_file_type_subtype = wtap_register_file_type_subtype(&json_info);
|
|
|
|
/*
|
|
* Register name for backwards compatibility with the
|
|
* wtap_filetypes table in Lua.
|
|
*/
|
|
wtap_register_backwards_compatibility_lua_name("JSON",
|
|
json_file_type_subtype);
|
|
}
|
|
|
|
/*
|
|
* Editor modelines - https://www.wireshark.org/tools/modelines.html
|
|
*
|
|
* Local variables:
|
|
* c-basic-offset: 4
|
|
* tab-width: 8
|
|
* indent-tabs-mode: nil
|
|
* End:
|
|
*
|
|
* vi: set shiftwidth=4 tabstop=8 expandtab:
|
|
* :indentSize=4:tabSize=8:noTabs=true:
|
|
*/
|