We can simplify our code quite a bit if we explicitly check for the
ifindex being 1 on Linux as a loopback check. Apparently, this is
hardcoded on Linux on the kernel, and effectively exported to userspace
via rtnl and such, hence we should be able to rely on it.
kdbus_msg.timeout_ns now takes an absolute value, based on CLOCK_MONOTONIC,
in order to eventually support automatically restarted syscalls.
Signed-off-by: Daniel Mack <daniel@zonque.org>
In kdbus.h, the following details changed:
* All commands gained a 'kernel_flags' field to report the flags supported
by the driver. Before, this was done in the 'flags' field in a
bidirectional way, which turned out to be a problem for the code in
sd-bus, as many parts of it reuse the same ioctl struct more than once
and consider them to be owned by userspace.
* Name listings are now returned by a new struct instead of reusing struct
kdbus_cmd_name for that matter. This way, we don't add more unneeded
fields to it and make the API cleaner.
* 'conn_flags' was renamed to 'flags' in struct kdbus_cmd_hello to make
the API a bit more unified.
Quoting from Jon Corbet's report of Stephen Hemminger's talk at Linux
Plumbers Conference 2014 (https://lwn.net/Articles/616241/):
[...] So Stephen encouraged everybody to run a command like:
sysctl -w net.core.default_qdisc=fq_codel
That will cause fq_codel to be used for all future connections
[Qdiscs apply to interfaces, not connections. Pointed out by TomH
in the article comments. -- mschmidt] (up to the next reboot).
Unfortunately, the default queuing discipline cannot be changed,
since it will certainly disturb some user's workload somewhere.
Let's have the recommended default in systemd.
Thanks to Dave Täht for advice and the summary at
https://lists.bufferbloat.net/pipermail/cerowrt-devel/2014-October/003701.html
kdbus.h now has KDBUS_ATTACH_COMM split into KDBUS_ATTACH_TID_COMM and
KDBUS_ATTACH_PID_COMM. The items were split already, so the change in
systemd is easy.
In kdbus.h, the 'features' field has been dropped again. Instead of
negotiating features that way, we decided to make the kernel return the
set of supported flags in each ioctl struct's .flags field, in both the
success and error cases.
Without the socket open we are going to crash and burn. If for
whatever reason we fail during deserialization we will fail when
trying to open the socket. In this case it is better to unlink the old
socket and maybe lose some messages, than to continue without the
notification socket.
Of course this situation should not happen, but we should handle
it as gracefully as possible anyway.
https://bugzilla.redhat.com/show_bug.cgi?id=1099299
Commit 864e17068 ("nspawn: actually allow access to /dev/net/tun in the
container") added "/dev/net/tun" to the list of allowed devices but forgot
to tweak the array length, which caused "/dev/kdbus/*" to be missed.
If we don't have privileges to setup the namespaces then we are most likely
running inside some sort of unprivileged container, hence not being able to
create namespace is not a problem because spawned service can't access host
system anyway.
