Make IPV6_[RECV]PKTINFO work for IPv4-mapped addresses

Bug Description

Currently, the Linux kernel doesn't provide IPV6_RECVPKTINFO ancillary data on datagrams coming in from IPv4-mapped clients (e.g., ::ffff:192.0.2.1) on INET6 sockets in the default dual personality mode, nor does it honour IPV6_PKTINFO when sending datagrams on such a socket to an IPv4-mapped destination.

This means that an UDP application that requires a server to respond with the same address as it was contacted on cannot reliably work for IPv4 clients, because 1) the server has no way of knowing which address it was contacted on, and even if it did, 2) the kernel would ignore requests to use a specific source address.

For a real-life manifestation of this problem, see this OpenVPN bug report: https://community.openvpn.net/openvpn/ticket/306

This has recently been fixed in the net-next upstream tree with the following commits (the first one of which made the 3.14 merge window, the second one will be in 3.15):

https://git.kernel.org/cgit/linux/kernel/git/davem/net-next.git/commit/?id=4b261c75a99f29c93a0b6babfc180cdf566bd654
https://git.kernel.org/cgit/linux/kernel/git/davem/net-next.git/commit/?id=c8e6ad0829a723a74cd2fea9996a3392d2579a18

Enabling IPv6 is getting increasingly important in these days, but at the same time maintaining backwards compatibility with IPv4-only clients is also essential. I'm therefore requesting a backport of the above two commits to the Ubuntu LTS kernel images so that it becomes possible to use OpenVPN (and any other software packages) in dual-stack mode.

Tore

This bug is missing log files that will aid in diagnosing the problem. From a terminal window please run:

apport-collect 1284535

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

This is really more a RFE/missing functionality than a bug per se, so confirming.

This bug was fixed in the package linux - 3.13.0-14.34

---------------
linux (3.13.0-14.34) trusty; urgency=low

[ Tim Gardner ]

* Release Tracking Bug
- LP: #1285851

[ Andy Whitcroft ]

* [Config] d-i -- add hyperv_keyboard to serial-modules udeb
- LP: #1285434

[ Hannes Frederic Sowa ]

* SAUCE: ipv6: honor IPV6_PKTINFO with v4 mapped addresses on sendmsg
- LP: #1284535

[ Jason Wang ]

* SAUCE: x86, hyperv: bypass the timer_irq_works() check
- LP: #1282693

[ Paolo Pisati ]

* [Config] disable FB_OMAP2, DRM_OMAP=m

[ Upstream Kernel Changes ]

* ipv6: make IPV6_RECVPKTINFO work for ipv4 datagrams
- LP: #1284535
* mei: remove flash_work_queue
* mei: drop redundant list_del_init
* mei: cleanup mei_irq_read_handler
* mei: enable marking internal commands
* mei: me: set dma mask using DMA mapping API
* Documentation/misc-devices/mei/mei-amt-version.c: remove unneeded call
of mei_deinit()
* mei: do not run reset flow from the interrupt thread
* mei: nfc: mei_nfc_free has to be called under lock
* mei: fix syntax in comments and debug output
* mei: revamp mei reset state machine
* mei: limit the number of consecutive resets
* mei: set client's read_cb to NULL when flow control fails
-- Tim Gardner <email address hidden> 2014年2月26日 08:43:20 -0500

I'm using kernel 3.2. This issue was fixed in 3.14. But I still have this issue with 3.2.