925119 - Gecko can call socketpair after sandbox initialization?

Status: RESOLVED DUPLICATE of bug 936252

(Core :: Security, defect)

Description

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

This happened once and I haven't been able to reproduce it.

E/Sandbox ( 4082): PID 4082 is missing syscall 288, arg1 1

This was around the same time that the parent process died for reasons I couldn't find in the log.  I had started CubeVid and switched to the window manager, I think, but I don't remember exactly what I was doing.

This was on mozilla-aurora with my patch for bug 921817 applied.

Comment 1

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Ran into this again while running mochitests, with a gecko I'd hacked to dump a stack.

E/Sandbox (  323): PID 323 is missing syscall 288, arg1 1
E/Sandbox (  323): frame 0: sp=0xbeed686c pc=0x405de206
E/Sandbox (  323): frame 1: sp=0xbeed68c4 pc=0xbeed6af0

$ addr2line -Cfie objdir-gecko/dist/bin/libxul.so $(printf '%x' $[ 0x405de206 - 0x40300002])
sys_socketpair
/home/jld/src/B2G.goldfish/gecko/toolkit/crashreporter/google-breakpad/src/third_party/lss/linux_syscall_support.h:3281
google_breakpad::CrashGenerationClient::RequestDump(void const*, unsigned int)
/home/jld/src/B2G.goldfish/gecko/toolkit/crashreporter/google-breakpad/src/client/linux/crash_generation/crash_generation_client.cc:48

So... we were in the process of crashing?

Comment 2

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

Not strictly speaking a "duplicate", but this was fixed as part of bug 936252, and  I verified that the crash reporter can get a minidump out of a sandboxed child process with that change.