Preload an HSTS domain list from chromium

This resolves the first part of servo/servo#6105. The remaining part is
to update the list based on the STS headers from the server.

.gitignore

@@ -26,3 +26,4 @@ Servo.app
 .config.mk.last
 parser.out
 /glfw
+hsts_preload.json

components/net/resource_task.rs

@@ -17,21 +17,24 @@ use net_traits::{Metadata, ProgressMsg, ResourceTask, AsyncResponseTarget, Respo
 use net_traits::ProgressMsg::Done;
 use util::opts;
 use util::task::spawn_named;
+use util::resource_files::read_resource_file;
 
 use devtools_traits::{DevtoolsControlMsg};
 use hyper::header::{ContentType, Header, SetCookie, UserAgent};
 use hyper::mime::{Mime, TopLevel, SubLevel};
 
+use rustc_serialize::json::{decode};
+
 use std::borrow::ToOwned;
 use std::boxed::FnBox;
 use std::collections::HashMap;
 use std::env;
 use std::fs::File;
 use std::io::{BufReader, Read};
+use std::str::{FromStr, from_utf8};
 use std::sync::Arc;
 use std::sync::mpsc::{channel, Receiver, Sender};
 
-
 static mut HOST_TABLE: Option<*mut HashMap<String, String>> = None;
 
 pub fn global_init() {
@@ -152,17 +155,86 @@ pub fn start_sending_opt(start_chan: LoadConsumer, metadata: Metadata) -> Result
     }
 }
 
+fn preload_hsts_domains() -> Option<HSTSList> {
+    match read_resource_file(&["hsts_preload.json"]) {
+        Ok(bytes) => {
+            match from_utf8(&bytes) {
+                Ok(hsts_preload_content) => {
+                    HSTSList::new_from_preload(hsts_preload_content)
+                },
+                Err(_) => None
+            }
+        },
+        Err(_) => None
+    }
+}
+
 /// Create a ResourceTask
 pub fn new_resource_task(user_agent: Option<String>,
                          devtools_chan: Option<Sender<DevtoolsControlMsg>>) -> ResourceTask {
+    let hsts_preload = preload_hsts_domains();
+
     let (setup_chan, setup_port) = channel();
     let setup_chan_clone = setup_chan.clone();
     spawn_named("ResourceManager".to_owned(), move || {
-        ResourceManager::new(setup_port, user_agent, setup_chan_clone, devtools_chan).start();
+        ResourceManager::new(setup_port, user_agent, setup_chan_clone, hsts_preload, devtools_chan).start();
     });
     setup_chan
 }
 
+#[derive(RustcDecodable, RustcEncodable)]
+pub struct HSTSEntry {
+    pub host: String,
+    pub include_subdomains: bool
+}
+
+#[derive(RustcDecodable, RustcEncodable)]
+pub struct HSTSList {
+    pub entries: Vec<HSTSEntry>
+}
+
+impl HSTSList {
+    pub fn new_from_preload(preload_content: &str) -> Option<HSTSList> {
+        match decode(preload_content) {
+            Ok(list) => Some(list),
+            Err(_) => None
+        }
+    }
+
+    pub fn always_secure(&self, host: &str) -> bool {
+        // TODO - Should this be faster than O(n)? The HSTS list is only a few
+        // hundred or maybe thousand entries...
+        self.entries.iter().any(|e| {
+            if e.include_subdomains {
+                host.ends_with(&format!(".{}", e.host)) || e.host == host
+            } else {
+                e.host == host
+            }
+        })
+    }
+
+
+    pub fn make_hsts_secure(&self, load_data: LoadData) -> LoadData {
+        if let Some(h) = load_data.url.domain() {
+            if self.always_secure(h) {
+                match &*load_data.url.scheme {
+                    "http" => {
+                        let mut secure_load_data = load_data.clone();
+                        let mut secure_url = load_data.url.clone();
+                        secure_url.scheme = "https".to_string();
+                        secure_load_data.url = secure_url;
+
+                        return secure_load_data
+                    },
+                    _ => ()
+                };
+            }
+        }
+
+        load_data
+    }
+}
+
 pub fn parse_hostsfile(hostsfile_content: &str) -> Box<HashMap<String, String>> {
     let ipv4_regex = regex!(
         r"^(?:(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.){3}(?:25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)$");
@@ -204,13 +276,15 @@ struct ResourceManager {
     cookie_storage: CookieStorage,
     resource_task: Sender<ControlMsg>,
     mime_classifier: Arc<MIMEClassifier>,
-    devtools_chan: Option<Sender<DevtoolsControlMsg>>
+    devtools_chan: Option<Sender<DevtoolsControlMsg>>,
+    hsts_list: Option<HSTSList>
 }
 
 impl ResourceManager {
     fn new(from_client: Receiver<ControlMsg>,
            user_agent: Option<String>,
            resource_task: Sender<ControlMsg>,
+           hsts_list: Option<HSTSList>,
            devtools_channel: Option<Sender<DevtoolsControlMsg>>) -> ResourceManager {
         ResourceManager {
             from_client: from_client,
@@ -218,7 +292,8 @@ impl ResourceManager {
             cookie_storage: CookieStorage::new(),
             resource_task: resource_task,
             mime_classifier: Arc::new(MIMEClassifier::new()),
-            devtools_chan: devtools_channel
+            devtools_chan: devtools_channel,
+            hsts_list: hsts_list
         }
     }
 }
@@ -262,6 +337,11 @@ impl ResourceManager {
             load_data.preserved_headers.set(UserAgent(ua.clone()));
         });
 
+        match self.hsts_list {
+            Some(ref l) => load_data = l.make_hsts_secure(load_data),
+            _ => ()
+        }
+
         fn from_factory(factory: fn(LoadData, LoadConsumer, Arc<MIMEClassifier>))
                         -> Box<FnBox(LoadData, LoadConsumer, Arc<MIMEClassifier>) + Send> {
             box move |load_data, senders, classifier| {

python/servo/bootstrap_commands.py

@@ -9,11 +9,15 @@
 
 from __future__ import print_function, unicode_literals
 
+import base64
+import json
 import os
 import os.path as path
+import re
 import shutil
 import subprocess
 import sys
+import StringIO
 import tarfile
 import urllib2
 from distutils.version import LooseVersion
@@ -27,27 +31,33 @@ from mach.decorators import (
 from servo.command_base import CommandBase, cd, host_triple
 
 
-def download(desc, src, dst):
+def download(desc, src, writer):
     print("Downloading %s..." % desc)
     dumb = (os.environ.get("TERM") == "dumb") or (not sys.stdout.isatty())
 
     try:
         resp = urllib2.urlopen(src)
+
+        fsize = None
+        if resp.info().getheader('Content-Length'):
             fsize = int(resp.info().getheader('Content-Length').strip())
+
         recved = 0
         chunk_size = 8192
 
-        with open(dst, 'wb') as fd:
         while True:
             chunk = resp.read(chunk_size)
-                if not chunk:
-                    break
+            if not chunk: break
             recved += len(chunk)
             if not dumb:
+                if fsize is not None:
                     pct = recved * 100.0 / fsize
                     print("\rDownloading %s: %5.1f%%" % (desc, pct), end="")
+                else:
+                    print("\rDownloading %s" % desc, end="")
+
                 sys.stdout.flush()
-                fd.write(chunk)
+            writer.write(chunk)
 
         if not dumb:
             print()
@@ -62,6 +72,14 @@ def download(desc, src, dst):
 
         sys.exit(1)
 
+def download_file(desc, src, dst):
+    with open(dst, 'wb') as fd:
+        download(desc, src, fd)
+
+def download_bytes(desc, src):
+    content_writer = StringIO.StringIO()
+    download(desc, src, content_writer)
+    return content_writer.getvalue()
 
 def extract(src, dst, movedir=None):
     tarfile.open(src).extractall(dst)
@@ -111,7 +129,7 @@ class MachCommands(CommandBase):
                         % self.rust_snapshot_path())
         tgz_file = rust_dir + '.tar.gz'
 
-        download("Rust snapshot", snapshot_url, tgz_file)
+        download_file("Rust snapshot", snapshot_url, tgz_file)
 
         print("Extracting Rust snapshot...")
         snap_dir = path.join(rust_dir,
@@ -142,7 +160,7 @@ class MachCommands(CommandBase):
                         % docs_name)
         tgz_file = path.join(hash_dir, 'doc.tar.gz')
 
-        download("Rust docs", snapshot_url, tgz_file)
+        download_file("Rust docs", snapshot_url, tgz_file)
 
         print("Extracting Rust docs...")
         temp_dir = path.join(hash_dir, "temp_docs")
@@ -166,7 +184,7 @@ class MachCommands(CommandBase):
                               self.cargo_build_id())
         if not force and path.exists(path.join(cargo_dir, "bin", "cargo")):
             print("Cargo already downloaded.", end=" ")
-            print("Use |bootstrap_cargo --force| to download again.")
+            print("Use |bootstrap-cargo --force| to download again.")
             return
 
         if path.isdir(cargo_dir):
@@ -177,7 +195,7 @@ class MachCommands(CommandBase):
         nightly_url = "https://static-rust-lang-org.s3.amazonaws.com/cargo-dist/%s/%s" % \
             (self.cargo_build_id(), tgz_file)
 
-        download("Cargo nightly", nightly_url, tgz_file)
+        download_file("Cargo nightly", nightly_url, tgz_file)
 
         print("Extracting Cargo nightly...")
         nightly_dir = path.join(cargo_dir,
@@ -185,6 +203,50 @@ class MachCommands(CommandBase):
         extract(tgz_file, cargo_dir, movedir=nightly_dir)
         print("Cargo ready.")
 
+    @Command('bootstrap-hsts-preload',
+             description='Download the HSTS preload list',
+             category='bootstrap')
+    @CommandArgument('--force', '-f',
+                     action='store_true',
+                     help='Force download even if HSTS list already exist')
+    def bootstrap_hsts_preload(self, force=False):
+        preload_filename = "hsts_preload.json"
+        preload_path = path.join(self.context.topdir, "resources")
+
+        if not force and path.exists(path.join(preload_path, preload_filename)):
+            print("HSTS preload list already downloaded.", end=" ")
+            print("Use |bootstrap-hsts-preload --force| to download again.")
+            return
+
+        chromium_hsts_url = "https://chromium.googlesource.com/chromium/src/net/+/master/http/transport_security_state_static.json?format=TEXT"
+
+        try:
+            content_base64 = download_bytes("Chromium HSTS preload list", chromium_hsts_url)
+        except URLError, e:
+            print("Unable to download chromium HSTS preload list, are you connected to the internet?")
+            sys.exit(1)
+
+        content_decoded = base64.b64decode(content_base64)
+        content_json = re.sub(r'//.*$', '', content_decoded, flags=re.MULTILINE)
+
+        try:
+            pins_and_static_preloads = json.loads(content_json)
+            entries = {
+                "entries": [
+                    {
+                        "host": e["name"],
+                        "include_subdomains": e.get("include_subdomains", False)
+                    }
+                    for e in pins_and_static_preloads["entries"]
+                ]
+            }
+
+            with open(path.join(preload_path, preload_filename), 'w') as fd:
+                json.dump(entries, fd, indent=4)
+        except ValueError, e:
+            print("Unable to parse chromium HSTS preload list, has the format changed?")
+            sys.exit(1)
+
     @Command('update-submodules',
              description='Update submodules',
              category='bootstrap')

python/servo/command_base.py

@@ -324,5 +324,7 @@ class CommandBase(object):
            not path.exists(path.join(
                 self.config["tools"]["cargo-root"], "cargo", "bin", "cargo")):
             Registrar.dispatch("bootstrap-cargo", context=self.context)
+        if not path.exists(path.join("resources", "hsts_preload.json")):
+            Registrar.dispatch("bootstrap-hsts-preload", context=self.context)
 
         self.context.bootstrapped = True

tests/unit/net/resource_task.rs

@@ -2,7 +2,9 @@
  * 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 net::resource_task::{new_resource_task, parse_hostsfile, replace_hosts};
+use net::resource_task::{
+    new_resource_task, parse_hostsfile, replace_hosts, HSTSList, HSTSEntry
+};
 use net_traits::{ControlMsg, LoadData, LoadConsumer};
 use net_traits::ProgressMsg;
 use std::borrow::ToOwned;
@@ -17,6 +19,126 @@ fn test_exit() {
     resource_task.send(ControlMsg::Exit).unwrap();
 }
 
+#[test]
+fn test_parse_hsts_preload_should_return_none_when_json_invalid() {
+    let mock_preload_content = "derp";
+    match HSTSList::new_from_preload(mock_preload_content) {
+        Some(_) => assert!(false, "preload list should not have parsed"),
+        None => assert!(true)
+    }
+}
+
+#[test]
+fn test_parse_hsts_preload_should_return_none_when_json_contains_no_entries_key() {
+    let mock_preload_content = "{\"nothing\": \"to see here\"}";
+    match HSTSList::new_from_preload(mock_preload_content) {
+        Some(_) => assert!(false, "preload list should not have parsed"),
+        None => assert!(true)
+    }
+}
+
+#[test]
+fn test_parse_hsts_preload_should_decode_host_and_includes_subdomains() {
+    let mock_preload_content = "{\
+                                     \"entries\": [\
+                                        {\"host\": \"mozilla.org\",\
+                                         \"include_subdomains\": false}\
+                                     ]\
+                                 }";
+    let hsts_list = HSTSList::new_from_preload(mock_preload_content);
+    let entries = hsts_list.unwrap().entries;
+
+    assert!(entries.get(0).unwrap().host == "mozilla.org");
+    assert!(entries.get(0).unwrap().include_subdomains == false);
+}
+
+#[test]
+fn test_hsts_list_with_no_entries_does_not_always_secure() {
+    let hsts_list = HSTSList {
+        entries: Vec::new()
+    };
+
+    assert!(hsts_list.always_secure("mozilla.org") == false);
+}
+
+#[test]
+fn test_hsts_list_with_exact_domain_entry_is_always_secure() {
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: false}]
+    };
+
+    assert!(hsts_list.always_secure("mozilla.org") == true);
+}
+
+#[test]
+fn test_hsts_list_with_subdomain_when_include_subdomains_is_true_is_always_secure() {
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: true}]
+    };
+
+    assert!(hsts_list.always_secure("servo.mozilla.org") == true);
+}
+
+#[test]
+fn test_hsts_list_with_subdomain_when_include_subdomains_is_false_is_not_always_secure() {
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: false}]
+    };
+
+    assert!(hsts_list.always_secure("servo.mozilla.org") == false);
+}
+
+#[test]
+fn test_hsts_list_with_subdomain_when_host_is_not_a_subdomain_is_not_always_secure() {
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: true}]
+    };
+
+    assert!(hsts_list.always_secure("servo-mozilla.org") == false);
+}
+
+#[test]
+fn test_hsts_list_with_subdomain_when_host_is_exact_match_is_always_secure() {
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: true}]
+    };
+
+    assert!(hsts_list.always_secure("mozilla.org") == true);
+}
+
+#[test]
+fn test_make_hsts_secure_doesnt_affect_non_http_schemas() {
+    let load_data = LoadData::new(Url::parse("file://mozilla.org").unwrap(), None);
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: false}]
+    };
+    let secure_load_data = hsts_list.make_hsts_secure(load_data);
+
+    assert!(&secure_load_data.url.scheme == "file");
+}
+
+#[test]
+fn test_make_hsts_secure_sets_secure_schema_on_subdomains_when_include_subdomains_is_true() {
+    let load_data = LoadData::new(Url::parse("http://servo.mozilla.org").unwrap(), None);
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: true}]
+    };
+    let secure_load_data = hsts_list.make_hsts_secure(load_data);
+
+    assert!(&secure_load_data.url.scheme == "https");
+}
+
+#[test]
+fn test_make_hsts_secure_forces_an_http_host_in_list_to_https() {
+    let load_data = LoadData::new(Url::parse("http://mozilla.org").unwrap(), None);
+    let hsts_list = HSTSList {
+        entries: vec![HSTSEntry { host: "mozilla.org".to_string(), include_subdomains: false}]
+    };
+    let secure_load_data = hsts_list.make_hsts_secure(load_data);
+
+    assert!(&secure_load_data.url.scheme == "https");
+}
+
 #[test]
 fn test_bad_scheme() {
     let resource_task = new_resource_task(None, None);