Security/Sandbox/2017-09-21
From MozillaWiki
« previous week | index | next week »
Contents
gcp
- bug 1382323 Firefox 54 on Fedora 26 doesn't launch custom protocol handler
- bug 1384804 Broken browser when running under a firejail sandbox
haik
- bug 1328975 - Mac e10s printing needs refactoring, causes sandbox violations to be logged
- bug 1401756 - [Mac] Remove unneeded mach-lookups from plugin sandbox rules
- State of Mac Content Sandboxing in Firefox with Chrome Comparison
bobowen
- bug 1372823 - Extend BaseThreadInitThunk gatekeeping to support Windows 64-bit
- Landed.
- bug 1397301 - Crash in sandbox::SharedMemIPCClient::DoCall
- Hopefully fixed by bug 1372823, probably wait to see if it re-occurs in Beta as it's fairly rare on Nightly.
- bug 1385207 - Audio over RDP connections not working in 57
- Decision to live with regression with workaround in 56.
- Have a patch for automatically weakening sandbox, possibly this should be behind a pref.
- bug 1400826 - [META] Tracking bug for Windows sandboxed process failed launches.
- Start filing bugs for these, numbers are much lower now errors are once per session. It will be interesting to see what happens in Beta.
- bug 1230910 - Get sandbox compiled with mingw-w64
- Just a note that Georg Koppen, got a working version of Tor Browser with the sandbox enabled.
jld
- Commented on bug 1386297, about ASan Nightlies and interaction with sandboxing
- Considered trying to get LSan to work, but, just no. (Blocks SIGSYS with inlined sigprocmask, uses ptrace, …)
- Suggestion: ASan Nightly uses ASan + sandboxing (we'd need to un-bit-rot and disable LSan by default), while tests use ASan+LSan
- And minimal tests of ASan + sandboxing — mainly, don't break the crash reporting
- FIled bug 1401062 for The Clone Thing, and a few others
- Commented on bug 1151624 (pid namespaces) to update it and describe current direction
- Filed bug 1401053 for pid namespace isolation for content
- Once again, PulseAudio is a problem
- bug 1401786 for cleaning up launch options
- Still just guessing whether this might be useful on other platforms….
- bug 1401790 to remove ProcessArchitecture (cross-arch NPAPI on OS X)
- bug 1396542 - The mysterious Goobuntu bug is about /var/lib/dbus/machine-id, and our QA independently found it on Ubuntu 14.04 and Arch
- :jesup's fd exhaustion (is this one really our problem, or are we just the one "on top" of the stacks?)
- The fd exhaustion isn't our problem, but we could, at least, print more useful error messages if it hits us.
- (It's vaguely *my* problem, because IPC shmem is part of the problem and I'm an IPC peer.)
- (Also there was a log message about an IPDL message losing attached fds, so maybe I should file a bug about that too….)
- Filed bug 1401774 to suggest having the broker handle this more gracefully
- Filed bug 1401776 against IPC to suggest raising the limit (but currently it's handled in Necko)
- The fd exhaustion isn't our problem, but we could, at least, print more useful error messages if it hits us.
handyman
- bug 1382251 - Brokering https in NPAPI process
- Work on dynamic memory management
- bug 1400169 - Crash in CallHookWithSEH
- Very old crash with an even older top-line call signature. Goes back at least to 2014.
- Suspicious of AV theory. Was unable to repro with Immunet AV.
- Seems to crash handling window messages in user mode.
- https://crash-stats.mozilla.com/signature/?proto_signature=~ShowFilePicker&signature=CallHookWithSEH&date=%3E%3D2017-03-21T15%3A19%3A43.000Z&date=%3C2017-09-21T15%3A19%3A43.000Z#graphs