In a nutshell, my problem is this: how can I hack, or communicate with, the netd daemon to force it to disable this feature?
(Android offers no configuration for this. I bug-reported the issue to Google ad android bug #31102 aka http : / / code.google.com/p/android/issues/detail?id=31102 (sorry I'm not allowed to post links) but they, of course, ignored it. Please note that there are lots of pages dealing with the question of how to enable IPv6 privacy extensions, because old versions of Android did not enable them: my question is how to disable them, permanently.)
What the netd daemon actually does is that when a network interface $IFACE is brought up, it opens the file /proc/sys/net/ipv6/conf/$IFACE/use_tempaddr and writes "2" there (this asks the Linux kernel to enable the feature). Up to Android 4.2, what I did was binary patch netd to replace the string "/proc/sys/net/ipv6/conf/%s/use_tempaddr" by "/dev/null\000" so the daemon would simply write that "2" to /dev/null and nothing would happen. But in Android 4.3 that part of the code has been slightly refactored, see InterfaceController.cpp from the netd source code, around line 134 (https : / / android.googlesource.com/platform/system/netd/+/android-4.3.1_r1/InterfaceController.cpp — again I'm not allowed to put links, what a pain), so a binary patch is not so trivial. (I could probably replace "use_tempaddr" by "hfr_grzcnqqe", but it would cause an error message in the logs and I'd like to avoid that.)
(Changing netd's source would be absolutely trivial. But I want to avoid recompiling it, because I'd probably spend many sleepless nights getting the correct native toolchain and convincing the Android makefiles to recompile just this bit: I don't have the resources to do a full Android build. Maybe I'm being pessimistic.)
In principle, it seems that netd reacts to commands that are sent to it (see https : / / android.googlesource.com/platform/system/netd/+/android-4.3.1_r1/CommandListener.cpp starting from around line 434). What I don't know is how to communicate with it to send it such commands, let alone do it precisely when a new interface is brought up. I know that at the other end of the line there is, for example, android.net.wifi.WifiStateMachine (see https : / / android.googlesource.com/platform/frameworks/base/+/android-4.3.1_r1/wifi/java/android/net/wifi/WifiStateMachine.java around line 2104). I'm a Linux dev, not so familiar with the Android IPC mechanisms or daemons, so I was hoping someone more knowledgeable could think of a way to pass a command at the right time.
PS: I'm aware that there's an app called to.doc.android.ipv6config which claims to solve the problem I'm talking about. But, looking at the code (https : / / gitorious.org/android-ipv6config/android-ipv6config/source/58e2060162485b54d4f8c147a558aeed708fa4b4:src/to/doc/android/ipv6config/LinuxIPCommandHelper.java around line 103), it's obvious that it does so in a completely wrong way, namely by talking to the kernel, bypassing netd's role as the network gatekeeper altogether.
PPS: I'm using CyanogenMod (currently 10.1.3, and this issue is preventing me from upgrading to 10.2), but I don't think this is relevant at all (I must admit I didn't check to see if CyanogenMod patched netd and/or android.net.wifi.WifiStateMachine in any way).
Meta-question: Is this the right place to ask? Apparently I have to ask 10 stupid questions before I'm allowed to post anywhere else.