Security/Sandbox/2018-04-05

From MozillaWiki
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

« previous week | index | next week »

jld

gcp

  • bug 1434711 WebGL causes a crash with the AMDGPU-PRO video driver
    • Fixed
  • Reviews
  • bug 1129492 Firefox content process has a live connection to the X11 server.

Alex_Gaynor

  • IPC fuzzer
    • First sec finding!
    • It works!
    • Working on tuning it now (added a message-type blacklist to avoid generating messages that are known to crash)
    • bug 1450047 - updated in tree copy of libFuzzer
    • https://reviews.llvm.org/D45322 - upstream bug in libFuzzer
    • Going to get a patch up on phab later today and start breaking out all the random hacks I've layered around the codebase
    • Still a handful of design issues I need to cycle back on (e.g. some things MOZ_ASSERT because they expect to be on a worker thread)
    • Need to generate a coverage report and see what it looks like
    • There's a file descriptor leak I think

haik

  • bug 1437281 - OSX dragging image to desktop changes OSX File associations
    • Landed
  • bug 1448374 - Loading a .javascript file from a WebExtension's web_accessible_resources messing with macOS file associations
    • Landed
  • bug 1448161 - Firefox changes default .txt app from TextEdit to Console (OS X)
    • Probably a dupe of 1448374, need to confirm
  • 1433577 bug 1433577 - [Mac] Enable sandboxing for the Flash NPAPI plugin process on Nightly
    • Landed, limited to Nightly
  • bug 1395504 - Infinite hang of web content process when parent process crashes...
    • Started trying to figure out what's going on here, need to understand more about how breakpad works in Mac Firefox

tjr

  • Working on Tor Network Planning w/ mcmanus
  • bug 1434316 MinGW x64 Build
    • Working on getting a debug build with symbols that _can_ run so we can figure out why it _doesn't_ run
  • Think I've agreed to implement SameSite Cookies....
  • Other small and various bugs

bobowen

  • bug 1449480 - Crash in logging::LogMessage::~LogMessage
    • Fallout for crashing where we're supposed to in chromium sandbox code.
    • Had to change a CHECK to a DCHECK for now.
    • I've realised that the handle verifier in the chromium sandbox code isn't turned on in release for chromium.
    • Also when it is turned on they have hooks enabled to track all other handle closing.
    • I have a patch for this, should be up for review soon.
  • Bug 1447019 - Use MITIGATION_WIN32K_DISABLE flag for GMP process.
    • Landed this but have had to pref it off for now due to intermittent test failures.
  • Continuing to look into the canvas code.

handyman

  • bug 1366256 - NPAPI sandbox level 3
    • should now be able to provision win10 loaner
  • bug 1436972 - Properly maintain audio session ref count
    • uplifted
  • 2 plugin sec bugs
  • bug 1446499 - FunctionHook::HookProtectedMode should be persistent
    • in review
  • bug 1450773 - Block restricting SIDs in plugin processes except in nightly
    • Weakened sandbox so that audio device changes are recognized
    • uplifted.
    • We can restore this at any time once Adobe fixes issues.
  • bug 1450708 - Crash in FunctionBroker
    • Looks like weird ConditionVariable behavior. Docs are too spare to know.

Round Table

  • Workday goals for Q1 not required
  • waiting on guidance from hr on what to do for Q2 and beyond
  • Windows cubeb audio remoting completion pushed back to end of Q3 on Windows.
    • Wondering if this might block win32k lockdown switch
    • OSX recently turned on but backed out shortly after due to perf issues.
    • (Linux: audio remoting blocked network/socket isolation, without which we basically don't have a sandbox)
  • WebGL remoting discussion
  • Mac Nightly now using OOP Extensions on Mac
  • From Jeff Wednesday - "We can't quite bind right at webidl, since we need to work with some content-side objects like images and videos. We need a light marshalling layer there for objects anyways, I think. Most webgl webidl entry points will likely become shims though. I'll try to get you a starting point this week."