Security/Sandbox/2017-10-26

From MozillaWiki
Jump to: navigation, search

« previous week | index | next week »

Alex_Gaynor

  • bug 1409768 - Significant performance regression for printing
    • Patch landed. Flushing your buffers is not the same thing as fsync! (Gecko Profiler is super great for anyone who hasn't used it)
    • (Hopefully) small bit of follow up work to use buffered IO - bug 1411984
  • bug 1407693 - Don't create files from content process on process crash
    • Wrote up an alternate approach, slightly more invasive, waiting for feedback

haik

  • bug 1403260 - Remove access to print server from content process sandbox
    • Landed, backed out last night due to test failure
  • bug 1393259 - Tighten font rules in the Mac content sandbox
    • Working on remoting the async font loading code path (AsyncFontInfoLoader)
    • There are some other code paths on main thread that will need sync remote loading

gcp

  • bug 1386404 Stop allowing Linux content processes to access /tmp
  • Testing path replacement at runtime
  • Adapted the tests, failing due to no(?) access to chrome dir

bobowen

  • bug 1400637 - Crash in mozilla::layers::ImageBridgeChild::InitForContent
    • Symantec still causing lots of crashes in Nightly, going to block just the later loading DLL, which hopefully won't cause same issues as last time
  • Chromium update
    • Lots more painful to take the latest changes particularly with C++14 things for Linux.
    • Have this building now, with a small try push running.
    • I've also decided to move to having patches in tree instead of references to the changesets.


jld

  • Regressions
    • bug 1410280 - prctl PR_GET_NAME, PulseAudio
    • bug 1411115 - fcntl F_SETLK, Nvidia GL and fontconfig
      • F_SETLKW already allowed for PulseAudio; also fontconfig apparently
  • "Fix it later" and now it's later:
    • Syscalls with filesystem paths -> problems for chrooting
    • bug 1408497 - inotify; exthandler -> gio, can just deny
    • bug 1409895 - getcwd; have a polyfill but it's just for this one test
      • If anyone knows a better way for mochitests to find their files....
    • bug 1409900 - quotactl can be blocked; statfs = open+fstatfs
  • Minor cleanups
    • bug 1410191 - all errors are EPERM
      • Fixed so the statfs handler can use it
    • bug 1410241 - possible use-after-destroy in SIGSYS handler
      • Trying to eliminate reasons for mysterious failures when testing new things
  • IPC stuff: landed Mac things; next is LaunchOptions (bug 1401786)

handyman

  • bug 1382251 - Brokering https in NPAPI process
    • fixing leakcheck
  • bug 1411379 - Flash updates need reg keys
    • Jimm asking Adobe

Round table

Win32K lockdown write up and research