Requiring /sys/fs/bpf path to be a mount point at the moment of cgroup
controllers realization does more harm than good, because:
* Realization happens early on boot, the mount point may not be ready at
the time. That happens if mounts are made by a .mount unit (the issue we
encountered).
* BPF filesystem may be mounted on another point.
Remove the check. Instead verify that path provided by BPFProgram= is
within BPF fs when unit properties are parsed.
Split in two commits for simple backport.
Add ip protocol token to SocketBind{Allow|Deny}= property parser.
Use parse_socket_bind_item helper.
Replace int32_t with int in cgroup item for socket-bind as it was
requested in [0].
Update tests.
[0] https://github.com/systemd/systemd/pull/19942#discussion_r652150024
Parse address family, ip protocol and ports, any of them can be
optional. If neither is specified, a special value 'any' is expected.
Helper is placed in shared to be reused in both fragment and dbus.
Add unit tests with valid and invalid examples.
Thin wrappers of ip_protocol_{from|to}_name targeting IPPROTO_TCP and
IPPROTO_UDP only.
Used to parse IP protocol configuration restricted only to TCP and UDP,
e.g. in SocketBind{Allow|Deny}= unit property.
These helpers are inspired by af_{from|to}_ipv4_ipv6 and potentially
extendable with other IP protocols if there is a use-case to expose
them.
Lookup ip protocol in a socket address to allow or deny binding a socket
to the address.
Matching rule is extended with 'protocol' field. If its value is 0
(IPPROTO_IP) ip protocol comparison is omitted and matching is passed to
the next token which is ip ports.
Documentation is updated.
Support filtering by ip protocol (L4) in SocketBind{Allow|Deny}=
properties.
The signature of dbus methods must be finalized before new release is
cut, hence reserve a parameter for ip protocol.
Implementation will follow.
Closes https://github.com/systemd/systemd/issues/19891
Verify that service exited correctly if valid ports are passed to
SocketBind{Allow|Deny}=
Use `ncat` program starting a listening service binding to a specified
port, e.g.
"timeout --preserve-status -sSIGTERM 1s /bin/nc -l -p ${port} -vv"
Add supported and install unit interface for socket-bind feature.
supported verifies that
- unified cgroup hierarchy (cgroup v2) is used
- BPF_FRAMEWORK (libbpf + clang + llvm + bpftool) was available in
compile time
- kernel supports BPF_PROG_TYPE_CGROUP_SOCK_ADDR
- bpf programs can be loaded into kernel
- bpf link can be used
install:
- load bpf_object from bpf skeleton
- resize rules map to fit socket_bind_allow and socket_bind deny rules
from cgroup context
- populate cgroup-bpf maps with rules
- get bpf programs from bpf skeleton
- attach programs to unit cgroup using bpf link
- save bpf link in the unit
* Add `bpf-framework` feature gate with 'auto', 'true' and 'false' choices
* Add libbpf [0] dependency
* Search for clang llvm-strip and bpftool binaries in compile time to
generate bpf skeleton.
For libbpf [0], make 0.2.0 [1] the minimum required version.
If libbpf is satisfied, set HAVE_LIBBPF config option to 1.
If `bpf-framework` feature gate is set to 'auto', means that whether
bpf feature is enabled or now is defined by the presence of all of
libbpf, clang, llvm and bpftool in build
environment.
With 'auto' all dependencies are optional.
If the gate is set to `true`, make all of the libbpf, clang and llvm
dependencies mandatory.
If it's set to `false`, set `BPF_FRAMEWORK` to false and make libbpf
dependency optional.
libbpf dependency is dynamic followed by the common pattern in systemd.
meson, bpf: add build rule for socket_bind program
Add a build script to compile bpf source code. A program in restricted
C is compiled into an object file. Object file is converted to BPF
skeleton [0] header file.
If build with custom meson build rule, the target header will reside in
build/ directory (not in source tree), e.g the path for socket_bind:
`build/src/core/bpf/socket_bind/socket-bind.skel.h`
Script runs the phases:
* clang to generate *.o from restricted C
* llvm-strip to remove useless DWARF info
* bpf skeleton generation with bpftool
These phases are logged to stderr for debug purposes.
To include BTF debug information, -g option is passed to clang.
[0] https://lwn.net/Articles/806911/
Introduce BPF program compiled from BPF source code in
restricted C - socket-bind.
It addresses feature request [0].
The goal is to allow systemd services to bind(2) only to a predefined set
of ports. This prevents assigning socket address with unallowed port
to a socket and creating servers listening on that port.
This compliments firewalling feature presenting in systemd:
whereas cgroup/{egress|ingress} hooks act on packets, this doesn't
protect from untrusted service or payload hijacking an important port.
While ports in 0-1023 range are restricted to root only, 1024-65535
range is not protected by any mean.
Performance is another aspect of socket_bind feature since per-packet
cost can be eliminated for some port-based filtering policies.
The feature is implemented with cgroup/bind{4|6} hooks [1].
In contrast to the present systemd approach using raw bpf instructions,
this program is compiled from sources. Stretch goal is to
make bpf ecosystem in systemd more friendly for developer and to clear
path for more BPF programs.
[0] https://github.com/systemd/systemd/pull/13496#issuecomment-570573085
[1] https://www.spinics.net/lists/netdev/msg489054.html
- Handle BPFProgram= property in string format
"<bpf_attach_type>:<bpffs_path>", e.g. egress:/sys/fs/bpf/egress-hook.
- Add dbus getter to list foreign bpf programs attached to a cgroup.
- Pin trivial bpf programs to bpf filesystem, compose BPFProgram= option
string and pass it to a unit. Programs store `0` in r0 BPF register for
denying action, e.g. drop a packet.
- Load trivial BPF programs
- Test is skipped if not run under root or if can not lock enough
memory.
- For egress and ingress hooks, test BPFProgram= option along with
with IP{Egress|Ingress}FilterPath=, expected result should not depend on
which rule is executed first.
Expected results for BPF_CGROUP_INET_INGRESS:
5 packets transmitted, 0 received, 100% packet loss, time 89ms
For BPF_CGROUP_INET_SOCK_CREATE:
ping: socket: Operation not permitted
- Introduce support of cgroup-bpf programs managed (i.e. compiled,
loaded to and unloaded from kernel) externally. Systemd is only
responsible for attaching programs to unit cgroup hence the name
'foreign'.
Foreign BPF programs are identified by bpf program ID and attach type.
systemd:
- Gets kernel FD of BPF program;
- Makes a unique identifier of BPF program from BPF attach type and
program ID. Same program IDs mean the same program, i.e the same
chunk of kernel memory. Even if the same program is passed multiple
times, identical (program_id, attach_type) instances are collapsed
into one;
- Attaches programs to unit cgroup.
- Store foreign bpf programs in cgroup context. A program is considered
foreign if it was loaded to a kernel by an entity external to systemd,
so systemd is responsible only for attach and detach paths.
- Support the case of pinned bpf programs: pinning to bpffs so a program
is kept loaded to the kernel even when program fd is closed by a user
application is a common way to extend program's lifetime.
- Aadd linked list node struct with attach type and bpffs path
fields.
Introduce bpf_cgroup_attach_type_table with accustomed attached type
names also used in bpftool.
Add bpf_cgroup_attach_type_{from|to}_string helpers to convert from|to
string representation of pinned bpf program, e.g.
"egress:/sys/fs/bpf/egress-hook" for
/sys/fs/bpf/egress-hook path and BPF_CGROUP_INET_EGRESS attach type.
Add helpers to:
- Create new BPFProgram instance from a path in bpf
filesystem and bpf attach type;
- Pin a program to bpf fs;
- Get BPF program ID by BPF program FD.
Julia Kartseva