This part of the copyright blurb stems from the GPL use recommendations:
https://www.gnu.org/licenses/gpl-howto.en.html
The concept appears to originate in times where version control was per
file, instead of per tree, and was a way to glue the files together.
Ultimately, we nowadays don't live in that world anymore, and this
information is entirely useless anyway, as people are very welcome to
copy these files into any projects they like, and they shouldn't have to
change bits that are part of our copyright header for that.
hence, let's just get rid of this old cruft, and shorten our codebase a
bit.
Files which are installed as-is (any .service and other unit files, .conf
files, .policy files, etc), are left as is. My assumption is that SPDX
identifiers are not yet that well known, so it's better to retain the
extended header to avoid any doubt.
I also kept any copyright lines. We can probably remove them, but it'd nice to
obtain explicit acks from all involved authors before doing that.
We maintain an "extra" set of IP accounting counters that are used when
we systemd is reloaded to carry over the counters from the previous run.
Let's reset these to zero whenever IP accounting is turned off. If we
don't do this then turning off IP accounting and back on later wouldn't
reset the counters, which is quite surprising and different from how our
CPU time counting works.
So, the kernel's management of cgroup/BPF programs is a bit misdesigned:
if you attach a BPF program to a cgroup and close the fd for it it will
stay pinned to the cgroup with no chance of ever removing it again (or
otherwise getting ahold of it again), because the fd is used for
selecting which BPF program to detach. The only way to get rid of the
program again is to destroy the cgroup itself.
This is particularly bad for root the cgroup (and in fact any other
cgroup that we cannot realistically remove during runtime, such as
/system.slice, /init.scope or /system.slice/dbus.service) as getting rid
of the program only works by rebooting the system.
To counter this let's closely keep track to which cgroup a BPF program
is attached and let's implicitly detach the BPF program when we are
about to close the BPF fd.
This hence changes the bpf_program_cgroup_attach() function to track
where we attached the program and changes bpf_program_cgroup_detach() to
use this information. Moreover bpf_program_unref() will now implicitly
call bpf_program_cgroup_detach().
In order to simplify things, bpf_program_cgroup_attach() will now
implicitly invoke bpf_program_load_kernel() when necessary, simplifying
the caller's side.
Finally, this adds proper reference counting to BPF programs. This
is useful for working with two BPF programs in parallel: the BPF program
we are preparing for installation and the BPF program we so far
installed, shortening the window when we detach the old one and reattach
the new one.
This new kernel 4.15 flag permits that multiple BPF programs can be
executed for each packet processed: multiple per cgroup plus all
programs defined up the tree on all parent cgroups.
We can use this for two features:
1. Finally provide per-slice IP accounting (which was previously
unavailable)
2. Permit delegation of BPF programs to services (i.e. leaf nodes).
This patch beefs up PID1's handling of BPF to enable both.
Note two special items to keep in mind:
a. Our inner-node BPF programs (i.e. the ones we attach to slices) do
not enforce IP access lists, that's done exclsuively in the leaf-node
BPF programs. That's a good thing, since that way rules in leaf nodes
can cancel out rules further up (i.e. for example to implement a
logic of "disallow everything except httpd.service"). Inner node BPF
programs to accounting however if that's requested. This is
beneficial for performance reasons: it means in order to provide
per-slice IP accounting we don't have to add up all child unit's
data.
b. When this code is run on pre-4.15 kernel (i.e. where
BPF_F_ALLOW_MULTI is not available) we'll make IP acocunting on slice
units unavailable (i.e. revert to behaviour from before this commit).
For leaf nodes we'll fallback to non-ALLOW_MULTI mode however, which
means that BPF delegation is not available there at all, if IP
fw/acct is turned on for the unit. This is a change from earlier
behaviour, where we use the BPF_F_ALLOW_OVERRIDE flag, so that our
fw/acct would lose its effect as soon as delegation was turned on and
some client made use of that. I think the new behaviour is the safer
choice in this case, as silent bypassing of our fw rules is not
possible anymore. And if people want proper delegation then the way
out is a more modern kernel or turning off IP firewalling/acct for
the unit algother.
This improves the BPF/cgroup detection logic, and looks whether
BPF_ALLOW_MULTI is supported. This flag allows execution of multiple
BPF filters in a recursive fashion for a whole cgroup tree. It enables
us to properly report IP accounting for slice units, as well as
delegation of BPF support to units without breaking our own IP
accounting.
I figure saying "systemd" here was a typo, and it should have been
"system". (Yes, it becomes very hard after a while typing "system"
correctly if you type "systemd" so often.) That said, "systemd" in some
ways is actually more correct, since BPF might be available for the
system instance but not in the user instance.
Either way, talking of "this systemd" is weird, let's reword this to be
"this manager", to emphasize that it's the local instance of systemd
where BPF is not available, but that it might be available otherwise.
Currently we allowed delegation for alluntis with cgroup backing
except for slices. Let's make this a bit more strict for now, and only
allow this in service and scope units.
Let's also add a generic accessor unit_cgroup_delegate() for checking
whether a unit has delegation turned on that checks the new bool first.
Also, when doing transient units, let's explcitly refuse turning on
delegation for unit types that don#t support it. This is mostly
cosmetical as we wouldn't act on the delegation request anyway, but
certainly helpful for debugging.
So far, we assumed that kernels where TRIE was on also supported
BPF/cgroup stuff. That's not a correct assumption to make, hence check
for both features separately.
Fixes: #7054
Let's permit installing BPF programs in cgroup subtrees if
Delegeate=yes. Let's not document this precise behaviour for now though,
as most likely the logic here should become recursive, but that's only
going to happen if the kernel starts supporting that. Until then,
support this in a non-recursive fashion.
Zbigniew Jędrzejewski-Szmek