libnetapi/HttpResult: improve Length() algorithm

`atoi()` maximum value is `LONG_MAX` which is smaller than `SIZE_MAX`.
This cause any Content-Length > LONG_MAX to be represented in an
erroneous value. This will also happen to any invalid Content-Length
header.

This change uses strtoul() for the extended range (should be the same as
size_t range), and combined with proper error checking to determine
whether the received Content-Length is a valid and/or representable
value. Returns 0 if the data is invalid or can not be represented.

Some shortcomings about the current implementation are also
noted accordingly. They will be solved in later patches.

Change-Id: If28c4c3b8430ed83dd0f600030ec8949cf7e0051
Reviewed-on: https://review.haiku-os.org/c/haiku/+/2927
Reviewed-by: waddlesplash <waddlesplash@gmail.com>
This commit is contained in:
Leorize 2020-06-17 23:49:57 -05:00 committed by waddlesplash
parent 9bfdc55864
commit d345666ea8

View File

@ -87,7 +87,27 @@ BHttpResult::Length() const
const char* length = Headers()["Content-Length"];
if (length == NULL)
return 0;
return atoi(length);
/* NOTE: Not RFC7230 compliant:
* - If Content-Length is a list, all values must be checked and verified
* to be duplicates of each other, but this is currently not supported.
*/
size_t result = 0;
/* strtoul() will ignore a prefixed sign, so we verify that there aren't
* any before continuing (RFC7230 only permits digits).
*
* We can check length[0] directly because header values are trimmed by
* HttpHeader beforehand. */
if (length[0] != '-' || length[0] != '+') {
errno = 0;
char *endptr = NULL;
result = strtoul(length, &endptr, 10);
/* ERANGE will be signalled if the result is too large (which can
* happen), in that case, return 0. */
if (errno != 0 || *endptr != '\0')
result = 0;
}
return result;
}