udev bridge fails to get events from kernel (missing FHANDLE)

There is indeed a race in the upstart job, which makes it unable to receive the kernel events. It works every time now after changing the upstart job to start on 'started' instead of 'starting' udev.

Looking at the libudev code, this is what is currently happening:
"""
struct udev_monitor *udev_monitor_new_from_netlink_fd(struct udev *udev, const char *name, int fd)
{
        struct udev_monitor *udev_monitor;
        unsigned int group;

        if (udev == NULL)
                return NULL;

        if (name == NULL)
                group = UDEV_MONITOR_NONE;
        else if (streq(name, "udev")) {
                /*
                 * We do not support subscribing to uevents if no instance of
                 * udev is running. Uevents would otherwise broadcast the
                 * processing data of the host into containers, which is not
                 * desired.
                 *
                 * Containers will currently not get any udev uevents, until
                 * a supporting infrastructure is available.
                 *
                 * We do not set a netlink multicast group here, so the socket
                 * will not receive any messages.
                 */
                if (access("/run/udev/control", F_OK) < 0 && !udev_has_devtmpfs(udev)) {
                        log_debug("the udev service seems not to be active, disable the monitor");
                        group = UDEV_MONITOR_NONE;
                } else
                        group = UDEV_MONITOR_UDEV;
        } else if (streq(name, "kernel"))
                group = UDEV_MONITOR_KERNEL;
        else
                return NULL;
....
"""

Added some debug lines in the upstart-udev-bridge job and /run/udev/control is not necessarily available when this job starts (I'd guess it only happens on ubuntu-touch, since we start udev a bit later than the desktop).

After spending a bit more time looking at why this happens with the touch devices, I found out that the root cause of this issue is the lack of CONFIG_FHANDLE in the kernel, which is needed systemd >= 209 (http://lists.freedesktop.org/archives/systemd-devel/2014-February/017362.html), and we're not enabling that option for any of our kernel packages.

We should probably enforce this option at our kernel scripts, since we're officially moving to systemd quite soon.

Removing the RTM bug tasks as we're still using systemd 208 there, so this bug doesn't happen on RTM.

Adding linux task to add this to the enforcement rules.

This bug was fixed in the package linux-mako - 3.4.0-6.35

---------------
linux-mako (3.4.0-6.35) vivid; urgency=low

  [ Andy Whitcroft ]

  * [Config] CONFIG_FHANDLE=y
    - LP: #1412543

  [ Tim Gardner ]

  * [Debian] getabis: Commit new ABI directory, remove the old
 -- Andy Whitcroft <email address hidden> Tue, 20 Jan 2015 10:17:37 +0000

This bug was fixed in the package linux-flo - 3.4.0-4.16

---------------
linux-flo (3.4.0-4.16) vivid; urgency=low

  [ Andy Whitcroft ]

  * [Config] CONFIG_FHANDLE=y
    - LP: #1412543

  [ Tim Gardner ]

  * [Debian] getabis: Commit new ABI directory, remove the old
 -- Andy Whitcroft <email address hidden> Tue, 20 Jan 2015 12:15:35 +0000

This bug was fixed in the package linux-manta - 3.4.0-7.30

---------------
linux-manta (3.4.0-7.30) vivid; urgency=low

  [ Andy Whitcroft ]

  * [Config] CONFIG_FHANDLE=y
    - LP: #1412543

  [ Tim Gardner ]

  * [Debian] getabis: Commit new ABI directory, remove the old
 -- Andy Whitcroft <email address hidden> Tue, 20 Jan 2015 12:07:58 +0000