Security/Sandbox/2017-04-13

« previous week | index | next week »

haik

• bug 1334550 - Proxy moz-extension protocol requests to the parent process
  • https://treeherder.mozilla.org/#/jobs?repo=try&revision=f6e5730da552aacad5ed43aa82dafdfd7825baf3
  • tests looks pretty good
  • founds an instance of loading an extension script without moz-extension via process scripts
• bug 1354674 - Introduce a (not-yet-enabled) level 3 Mac content sandbox with home directory read access disabled
• bug 1356167 - [Mac] Remove $PROFILE/extensions read access from level 3 content sandbox

bobowen

• Fair bit of time getting bug 1329294 and bug 1347646 landed and into Beta.
• bug 1344465 - Can't submit form using post method form WebExtensions or file:// page (related to bug 1347921?)
  • Working through scenarios for this, need to add fair bit of testing.

Alex_Gaynor

• bug 1355083 Made the macOS sandbox policies full constant strings - landed.
• bug 1294641 Make it possible to deny read access to $HOME by whitelisting the source location for non-packaged builds
  • Works locally, still figuring out what the design-correct way to detect non-packaged and the source location.
• bug 1354678 - Moved macOS sandbox policies from c-strings to files - not really started

gcp

• PTO
• bug 1308400 Debugging read policy with symlink handling

jld

• Nobody had triaged the SIGSYS crashes in a while; I did.
  • Filed a bunch of bugs; have patches for some
  • bug 1354731 - GPU driver mknod; landed
  • bug 1355274 - Pulseaudio libasyncns socketpair; have patch, tested
  • bug 1325242 - Did needinfo (sandbox problems via GConf -> ORBit -> utime)
  • bug 1294528 - The JAR bug is back
  • bug 1355270 - fstatat(2) <- ftw(3) <- ?????
  • Possibly more I'm forgetting, but they're all linked to crash signatures.
• Picked up some minor bugs that had been annoying me; did patches; sent to try
  • bug 1320834 - prctl
  • bug 1302711 - ioctl
  • bug 1328896 - fcntl

roundtable

• Telemetry for Linux?
  • It's working, but need to select “Don't Sanitize” to see any data
  • Some of it is also on crash-stats, but not all of it
  • msgsnd: We allowed msgget (bug 1285902) but nothing else, so the process can create message queues but not use them.
    • And that bug is old enough that the crash reports have expired, so we don't know *what* was using message queues