firefox-main/netwerk/base/rust-helper/src/lib.rs (file symbol)

Source code

Revision control

Copy as Markdown

Other Tools

/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
use std::{
fs::File,
io::{self, BufRead},
net::Ipv4Addr,
path::Path,
};
use nserror::*;
use nsstring::{nsACString, nsCString};
use thin_vec::ThinVec;
#[cfg(windows)]
use {
std::{fs::OpenOptions, os::windows::fs::OpenOptionsExt},
windows_sys::Win32::Storage::FileSystem::{
FILE_SHARE_DELETE, FILE_SHARE_READ, FILE_SHARE_WRITE,
},
};
/// HTTP leading whitespace, defined in netwerk/protocol/http/nsHttp.h
static HTTP_LWS: &[u8] = b" \t";
/// Trim leading whitespace, trailing whitespace, and quality-value
/// from a token.
fn trim_token(token: &[u8]) -> &[u8] {
// Trim left whitespace
let ltrim = token
.iter()
.take_while(|c| HTTP_LWS.iter().any(|ws| &ws == c))
.count();
// Trim right whitespace
// remove "; q=..." if present
let rtrim = token[ltrim..]
.iter()
.take_while(|c| **c != b';' && HTTP_LWS.iter().all(|ws| ws != *c))
.count();
&token[ltrim..ltrim + rtrim]
}
// Small helper that opens for reading in a fail-fast, cross-platform way.
// This is necessary because on windows it's possible that File::open will
// succeed but reading from the file would hang. See bug 1970349
fn open_read_fast_fail(path: &Path) -> io::Result<File> {
#[cfg(windows)]
{
// If another process opened the file with *no* sharing, this will fail
// immediately with a sharing violation instead of letting a later read hang.
OpenOptions::new()
.read(true)
.share_mode(FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE)
.open(path)
}
#[cfg(not(windows))]
{
File::open(path)
}
}
#[no_mangle]
/// Allocates an nsACString that contains a ISO 639 language list
/// notated with HTTP "q" values for output with an HTTP Accept-Language
/// header. Previous q values will be stripped because the order of
/// the langs implies the q value. q-values decrease by 0.1 for each subsequent language,
/// with a minimum value of 0.1.
///
/// Ex: passing: "en, ja"
/// returns: "en,ja;q=0.9"
///
/// passing: "en, ja, fr_CA"
/// returns: "en,ja;q=0.9,fr_CA;q=0.8"
pub extern "C" fn rust_prepare_accept_languages(
i_accept_languages: &nsACString,
o_accept_languages: &mut nsACString,
) -> nsresult {
if i_accept_languages.is_empty() {
return NS_OK;
}
let make_tokens = || {
i_accept_languages
.split(|c| *c == b',')
.map(trim_token)
.filter(|token| !token.is_empty())
};
for (count_n, i_token) in make_tokens().enumerate() {
// delimiter if not first item
if count_n != 0 {
o_accept_languages.append(",");
}
let token_pos = o_accept_languages.len();
o_accept_languages.append(i_token as &[u8]);
{
let o_token = o_accept_languages.to_mut();
canonicalize_language_tag(&mut o_token[token_pos..]);
}
//Since we need to emulate chrome behavior i.e languages should get q=1.0,0.9,0.8 and so on...
let q_val_max = 10;
let weight_of_decrement = 1;
let step = std::cmp::min(10, count_n); //if num_language > 10, q_val_max - curr_cnt*weight_of_decrement underflows
let q_val = std::cmp::max(q_val_max - step * weight_of_decrement, 1); //q-weight shouldn't go below 0.1
if count_n > 0 && q_val < 10 {
o_accept_languages.append(&format!(";q=0.{}", q_val));
}
}
NS_OK
}
/// Defines a consistent capitalization for a given language string.
///
/// # Arguments
/// * `token` - a narrow char slice describing a language.
///
/// Valid language tags are of the form
/// "*", "fr", "en-US", "es-419", "az-Arab", "x-pig-latin", "man-Nkoo-GN"
///
/// Language tags are defined in the
/// [rfc5646](https://tools.ietf.org/html/rfc5646) spec. According to
/// the spec:
///
/// > At all times, language tags and their subtags, including private
/// > use and extensions, are to be treated as case insensitive: there
/// > exist conventions for the capitalization of some of the subtags,
/// > but these MUST NOT be taken to carry meaning.
///
/// So why is this code even here? See bug 1108183, I guess.
fn canonicalize_language_tag(token: &mut [u8]) {
for c in token.iter_mut() {
*c = c.to_ascii_lowercase();
}
let sub_tags = token.split_mut(|c| *c == b'-');
for (i, sub_tag) in sub_tags.enumerate() {
if i == 0 {
// ISO 639-1 language code, like the "en" in "en-US"
continue;
}
match sub_tag.len() {
// Singleton tag, like "x" or "i". These signify a
// non-standard language, so we stop capitalizing after
// these.
1 => break,
// ISO 3166-1 Country code, like "US"
2 => {
sub_tag[0] = sub_tag[0].to_ascii_uppercase();
sub_tag[1] = sub_tag[1].to_ascii_uppercase();
}
// ISO 15924 script code, like "Nkoo"
4 => {
sub_tag[0] = sub_tag[0].to_ascii_uppercase();
}
_ => {}
};
}
}
#[no_mangle]
pub extern "C" fn rust_net_is_valid_ipv4_addr(addr: &nsACString) -> bool {
is_valid_ipv4_addr(addr)
}
#[inline]
fn try_apply_digit(current_octet: u8, digit_to_apply: u8) -> Option<u8> {
current_octet.checked_mul(10)?.checked_add(digit_to_apply)
}
pub fn is_valid_ipv4_addr(addr: &[u8]) -> bool {
let mut current_octet: Option<u8> = None;
let mut dots: u8 = 0;
for c in addr {
let c = *c as char;
match c {
'.' => {
match current_octet {
None => {
// starting an octet with a . is not allowed
return false;
}
Some(_) => {
dots += 1;
current_octet = None;
}
}
}
// The character is not a digit
no_digit if no_digit.to_digit(10).is_none() => {
return false;
}
digit => {
match current_octet {
None => {
// Unwrap is sound because it has been checked in the previous arm
current_octet = Some(digit.to_digit(10).unwrap() as u8);
}
Some(octet) => {
if let Some(0) = current_octet {
// Leading 0 is not allowed
return false;
}
if let Some(applied) =
try_apply_digit(octet, digit.to_digit(10).unwrap() as u8)
{
current_octet = Some(applied);
} else {
// Multiplication or Addition overflowed
return false;
}
}
}
}
}
}
dots == 3 && current_octet.is_some()
}
#[no_mangle]
pub extern "C" fn rust_net_is_valid_ipv6_addr(addr: &nsACString) -> bool {
is_valid_ipv6_addr(addr)
}
pub fn is_valid_ipv6_addr(addr: &[u8]) -> bool {
let mut double_colon = false;
let mut colon_before = false;
let mut digits: u8 = 0;
let mut blocks: u8 = 0;
// The smallest ipv6 is unspecified (::)
// The IP starts with a single colon
if addr.len() < 2 || addr[0] == b':' && addr[1] != b':' {
return false;
}
//Enumerate with an u8 for cache locality
for (i, c) in (0u8..).zip(addr) {
match c {
maybe_digit if maybe_digit.is_ascii_hexdigit() => {
// Too many digits in the block
if digits == 4 {
return false;
}
colon_before = false;
digits += 1;
}
b':' => {
// Too many columns
if double_colon && colon_before || blocks == 8 {
return false;
}
if !colon_before {
if digits != 0 {
blocks += 1;
}
digits = 0;
colon_before = true;
} else if !double_colon {
double_colon = true;
}
}
b'.' => {
// IPv4 from the last block
if is_valid_ipv4_addr(&addr[(i - digits) as usize..]) {
return double_colon && blocks < 6 || !double_colon && blocks == 6;
}
return false;
}
_ => {
// Invalid character
return false;
}
}
}
if colon_before && !double_colon {
// The IP ends with a single colon
return false;
}
if digits != 0 {
blocks += 1;
}
double_colon && blocks < 8 || !double_colon && blocks == 8
}
#[no_mangle]
pub extern "C" fn rust_net_is_valid_scheme_char(a_char: u8) -> bool {
is_valid_scheme_char(a_char)
}
#[no_mangle]
pub extern "C" fn rust_net_is_valid_scheme(scheme: &nsACString) -> bool {
if scheme.is_empty() {
return false;
}
// first char must be alpha
if !scheme[0].is_ascii_alphabetic() {
return false;
}
scheme[1..]
.iter()
.all(|a_char| is_valid_scheme_char(*a_char))
}
fn is_valid_scheme_char(a_char: u8) -> bool {
a_char.is_ascii_alphanumeric() || a_char == b'+' || a_char == b'.' || a_char == b'-'
}
pub type ParsingCallback = extern "C" fn(&ThinVec<nsCString>) -> bool;
#[no_mangle]
pub extern "C" fn rust_parse_etc_hosts(path: &nsACString, callback: ParsingCallback) {
let path_str = path.to_utf8();
let path = Path::new(&*path_str);
// Try to open in a way that fails immediately if locked (on Windows).
let file = match open_read_fast_fail(path) {
Ok(f) => io::BufReader::new(f),
Err(..) => return, // Not readable right now; bail out quietly like before.
};
let mut array = ThinVec::new();
for line in file.lines() {
let line = match line {
Ok(l) => l,
Err(..) => continue,
};
let mut iter = line.split('#').next().unwrap().split_whitespace();
iter.next(); // skip the IP
array.extend(
iter.filter(|host| {
// Make sure it's a valid domain
let invalid = [
'\0', '\t', '\n', '\r', ' ', '#', '%', '/', ':', '?', '@', '[', '\\', ']',
];
host.parse::<Ipv4Addr>().is_err() && !host.contains(&invalid[..])
})
.map(nsCString::from),
);
// /etc/hosts files can be huge. To make sure we don't block shutdown
// for every 100 domains that we parse we call the callback passing the
// domains and see if we should keep parsing.
if array.len() > 100 {
let keep_going = callback(&array);
array.clear();
if !keep_going {
break;
}
}
}
if !array.is_empty() {
callback(&array);
}
}