No functional change, just moving a bunch of things around. Before
we needed a rather complicated setup to test hostname_setup(), because
the code was in src/core/. When things are moved to src/shared/
we can just test it as any function.
The test is still "unsafe" because hostname_setup() may modify the
hostname.
When someone runs 'nft flush ruleset' in the same net namespace
this will also tear down the systemd nat table.
Unlike iptables -t nat -F, which will remove all rules added by
the systemd iptables backend, iptables has builtin chains that cannot
be deleted. IOW, the next add operation will 'just work'.
In the nftables case however, the entire table gets removed.
When the systemd nat table is removed by an external entity next
attempt to add a set element will yield -ENOENT.
If this happens, recreate the table, and, if successful, re-do
the add operation.
Note that this doesn't protect against external sabotage such as
a running 'while true; nft flush ruleset;done'. However, there is
nothing that could be done short of extending the kernel to allow
tables to be "frozen" or otherwise tied to a process such as
systemd-networkd.
Idea is to use a static ruleset, added when the first attempt to
add a masquerade or dnat rule is made.
The alternative would be to add the ruleset when the init function is called.
The disadvantage is that this enables connection tracking and NAT in the kernel
(as the ruleset needs this to work), which comes with some overhead that might
not be needed (no nspawn usage and no IPMasquerade option set).
There is no additional dependency on the 'nft' userspace binary or other libraries.
sd-netlinks nfnetlink backend is used to modify the nftables ruleset.
The commit message/comments still use nft syntax since that is what
users will see when they use the nft tool to list the ruleset.
The added initial skeleton (added on first fw_add_masquerade/local_dnat
call) looks like this:
table ip io.systemd.nat {
set masq_saddr {
type ipv4_addr
flags interval
elements = { 192.168.59.160/28 }
}
map map_port_ipport {
type inet_proto . inet_service : ipv4_addr . inet_service
elements = { tcp . 2222 : 192.168.59.169 . 22 }
}
chain prerouting {
type nat hook prerouting priority dstnat + 1; policy accept;
fib daddr type local dnat ip addr . port to meta l4proto . th dport map @map_port_ipport
}
chain output {
type nat hook output priority -99; policy accept;
ip daddr != 127.0.0.0/8 oif "lo" dnat ip addr . port to meta l4proto . th dport map @map_port_ipport
}
chain postrouting {
type nat hook postrouting priority srcnat + 1; policy accept;
ip saddr @masq_saddr masquerade
}
}
Next calls to fw_add_masquerade/add_local_dnat will then only add/delete the
element/mapping to masq_saddr and map_port_ipport, i.e. the ruleset doesn't
change -- only the set/map content does.
Running test-firewall-util with this backend gives following output
on a parallel 'nft monitor':
$ nft monitor
add table ip io.systemd.nat
add chain ip io.systemd.nat prerouting { type nat hook prerouting priority dstnat + 1; policy accept; }
add chain ip io.systemd.nat output { type nat hook output priority -99; policy accept; }
add chain ip io.systemd.nat postrouting { type nat hook postrouting priority srcnat + 1; policy accept; }
add set ip io.systemd.nat masq_saddr { type ipv4_addr; flags interval; }
add map ip io.systemd.nat map_port_ipport { type inet_proto . inet_service : ipv4_addr . inet_service; }
add rule ip io.systemd.nat prerouting fib daddr type local dnat ip addr . port to meta l4proto . th dport map @map_port_ipport
add rule ip io.systemd.nat output ip daddr != 127.0.0.0/8 fib daddr type local dnat ip addr . port to meta l4proto . th dport map @map_port_ipport
add rule ip io.systemd.nat postrouting ip saddr @masq_saddr masquerade
add element ip io.systemd.nat masq_saddr { 10.1.2.3 }
add element ip io.systemd.nat masq_saddr { 10.0.2.0/28 }
delete element ip io.systemd.nat masq_saddr { 10.0.2.0/28 }
delete element ip io.systemd.nat masq_saddr { 10.1.2.3 }
add element ip io.systemd.nat map_port_ipport { tcp . 4711 : 1.2.3.4 . 815 }
delete element ip io.systemd.nat map_port_ipport { tcp . 4711 : 1.2.3.4 . 815 }
add element ip io.systemd.nat map_port_ipport { tcp . 4711 : 1.2.3.5 . 815 }
delete element ip io.systemd.nat map_port_ipport { tcp . 4711 : 1.2.3.5 . 815 }
CTRL-C
Things not implemented/supported:
1. Change monitoring. The kernel allows userspace to learn about changes
made by other clients (using nfnetlink notifications). It would be
possible to detect when e.g. someone removes the systemd nat table.
This would need more work. Its also not clear on how to react to
external changes -- it doesn't seem like a good idea to just auto-undo
everthing.
2. 'set masq_saddr' doesn't handle overlaps.
Example:
fw_add_masquerade(true, AF_INET, "10.0.0.0" , 16);
fw_add_masquerade(true, AF_INET, "10.0.0.0" , 8); /* fails */
With the iptables backend the second call works, as it adds an
independent iptables rule.
With the nftables backend, the range 10.0.0.0-10.255.255.255 clashes with
the existing range of 10.0.0.0-10.0.255.255 so 2nd add gets rejected by the
kernel.
This will generate an error message from networkd ("Could not enable IP
masquerading: File exists").
To resolve this it would be needed to either keep track of the added elements
and perform range merging when overlaps are detected.
However, the add erquests are done using the configured network on a
device, so no overlaps should occur in normal setups.
IPv6 support is added in a extra changeset.
Fixes: #13307
Will be used by nftables nfnetlink backend.
It sends a series of netlink messages that form a nftables
update transaction.
The transaction will then generate a series of ack messages
(or an error).
This function will be used to read these acks.
nftables uses a transaction-based netlink model: one netlink write
comes with multiple messages.
A 'BEGIN' message to tell nf_tables/kernel that a new transaction starts.
Then, one more messages to add/delete tables/chains/rules etc.
Lastly, an END message that commits all changes.
This function will be used to send all the individual messages that should
make up a single transaction as a single write.
add nfnetlink_nftables helper functions to:
* open a new nfnetlink socket to kernel
* add tables, chains, rules, sets and maps
* delete/flush table
* add and delete elements from sets/maps
for planned nft backend we have three choices:
- open/close a new nfnetlink socket for every operation
- keep a nfnetlink socket open internally
- expose a opaque fw_ctx and stash all internal data here.
Originally I opted for the 2nd option, but during review it was
suggested to avoid static storage duration because of perceived
problems with threaded applications.
This adds fw_ctx and new/free functions, then converts the existing api
and nspawn and networkd to use it.
Next patch will need to pass two pointers to the callback instead
of just the addr mask. Caller will pass a compound structure, so
make this 'void *userdata' to de-clutter the next patch.
In a nutshell:
1. git mv firewall-util.c firewall-util-iptables.c
2. existing external functions gain _iptables_ in their names
3. firewall-util.c provides old function names
4. build system always compiles firewall-util.c,
firewall-util-iptables.c is conditional instead (libiptc).
5. On first call to any of the 'old' API functions performs
a probe that should return the preferred backend.
In a future step, can add firewall-util-FOOTYPE.c, add its
probe function to firewall-util.c and then have calls to
fw_add_masq/local_dnat handed to the detected backend.
For now, only iptables backend exists, and no special probing
takes place for it, i.e. when systemd was built with iptables,
that will be used. If not, requets to add masquerade/dnat will
fail with same error (-EOPNOTSUPP) as before this change.
For reference, the rules added by the libiptc/iptables backend look like this:
for service export (via systemd-nspawn):
[0:0] -A PREROUTING -p tcp -m tcp --dport $exportedport -m addrtype --dst-type LOCAL -j DNAT --to-destination $containerip:$port
[0:0] -A OUTPUT ! -d 127.0.0.0/8 -p tcp -m tcp --dport $exportedport -m addrtype --dst-type LOCAL -j DNAT --to-destination $containerip:$port
for ip masquerade:
[0:0] -A POSTROUTING -s network/prefix -j MASQUERADE
Make sure we don't add masquerading rules without a explicitly
specified network range we should be masquerading for.
The only caller aside from test case is
networkd-address.c which never passes a NULL source.
As it also passes the network prefix, that should always be > 0 as well.
This causes expected test failure:
Failed to modify firewall: Invalid argument
Failed to modify firewall: Invalid argument
Failed to modify firewall: Invalid argument
Failed to modify firewall: Protocol not available
Failed to modify firewall: Protocol not available
Failed to modify firewall: Protocol not available
Failed to modify firewall: Protocol not available
The failing test cases are amended to expect failure on
NULL source or prefix instead of success.
This commit extends $SYSTEMD_COLORS to an enum variable (compared to
a simple boolean) which specifies the "colors mode". This means that, in
addition to disabling colors altogether, it's now possible to restrict
the console output to 16 or 256 colors only.
Previously, we'd already have explicit logging for the case where
$XDG_RUNTIME_DIR is not set. Let's also add some explicit logging for
the EPERM/ACCESS case. Let's also in both cases suggest the
--machine=<user>@.host syntax.
And while we are at it, let's remove side-effects from the macro.
By checking for both the EPERM/EACCES case and the $XDG_RUNTIME_DIR case
we will now catch both the cases where people use "su" to issue a
"systemctl --user" operation, and those where they (more correctly, but
still not good enough) call "su -".
Fixes: #17901
So far, the bridge always acted as if "--system" was used, i.e. would
unconditionally connect to the system bus. Let's add "--user" too, to
connect to the users session bus.
This is mostly for completeness' sake.
I wanted to use this when making sd-bus's ability to connect to other
user's D-Bus busses work, but it didn't exist so far. In the interest of
keeping things compatible the implementation in sd-bus will not use the
new "--user" switch, and instead manually construct the right bus path
via "--path=", but we still should add the proper switches, as
preparation for a brighter future, one day.
This is unfortunately harder to implement than it sounds. The user's bus
is bound a to the user's lifecycle after all (i.e. only exists as long
as the user has at least one PAM session), and the path dynamically (at
least theoretically, in practice it's going to be the same always)
generated via $XDG_RUNTIME_DIR in /run/.
To fix this properly, we'll thus go through PAM before connecting to a
user bus. Which is hard since we cannot just link against libpam in the
container, since the container might have been compiled entirely
differently. So our way out is to use systemd-run from outside, which
invokes a transient unit that does PAM from outside, doing so via D-Bus.
Inside the transient unit we then invoke systemd-stdio-bridge which
forwards D-Bus from the user bus to us. The systemd-stdio-bridge makes
up the PAM session and thus we can sure tht the bus exists at least as
long as the bus connection is kept.
Or so say this differently: if you use "systemctl -M lennart@foobar"
now, the bus connection works like this:
1. sd-bus on the host forks off:
systemd-run -M foobar -PGq --wait -pUser=lennart -pPAMName=login systemd-stdio-bridge
2. systemd-run gets a connection to the "foobar" container's
system bus, and invokes the "systemd-stdio-bridge" binary as
transient service inside a PAM session for the user "lennart"
3. The systemd-stdio-bridge then proxies our D-Bus traffic to
the user bus.
sd-bus (on host) → systemd-run (on host) → systemd-stdio-bridge (in container)
Complicated? Well, to some point yes, but otoh it's actually nice in
various other ways, primarily as it makes the -H and -M codepaths more
alike. In the -H case (i.e. connect to remote host via SSH) a very
similar three steps are used. The only difference is that instead of
"systemd-run" the "ssh" binary is used to invoke the stdio bridge in a
PAM session of some other system. Thus we get similar implementation and
isolation for similar operations.
Fixes: #14580
So far when asked for augmented bus credentials and the process was
already gone we'd fail fatally. Let's make this graceful instead, and
never allow augmenting fail due to PID having vanished — unless the
augmenting is the explicit and only purpose of the requested operation.
This should be safe as clients have to explicitly query the acquired
creds anyway and handle if they couldn't be acquired. Moreover we
already handle permission problems gracefully, thus clients must be
ready to deal with missing creds.
This is useful to make selinux authorization work for short-lived client
proceses. PReviously we'd augment creds to have more info to log about
(the selinux decision would not be based on augmented data however,
because that'd be unsafe), and would fail if we couldn't get it. Now,
we'll try to acquire the data, but if we cannot acquire it, we'll still
do the selinux check, except that logging will be more limited.
Let's clean up hostname_is_valid() a bit: let's turn the second boolean
argument into a more explanatory flags field, and add a flag that
accepts the special name ".host" as valid. This is useful for the
container logic, where the special hostname ".host" refers to the "root
container", i.e. the host system itself, and can be specified at various
places.
let's also get rid of machine_name_is_valid(). It was just an alias,
which is confusing and even more so now that we have the flags param.
The immediately following container_get_leader() call validate the name
anyway, no need to twice exactly the same way twice immediately after
each other.
Use symlink_atomic_label() instead of symlink_atomic() as the symlink
may need a different label than the parent directory.
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
```
❯ ssh sus@xx.xx.xx.xx
Last login: Sat Nov 14 17:32:08 2020 from 10.104.45.138
17:36:19 up 0 min, 0 users, load average: 0.00, 0.00, 0.00
> systemd-analyze blame
Bootup is not yet finished (org.freedesktop.systemd1.Manager.FinishTimestampMonotonic=0).
Please try again later.
Hint: Use 'systemctl list-jobs' to see active jobs
> systemd-analyze blame
43.954s systemd-time-wait-sync.service
1.969s systemd-networkd-wait-online.service
1.559s cloud-init-local.service
1.039s cloud-init.service
414ms cloud-final.service
387ms dracut-initqueue.service
382ms initrd-switch-root.service
380ms cloud-config.service
198ms systemd-journal-flush.service
136ms systemd-udev-trigger.service
115ms initrd-parse-etc.service
97ms systemd-timesyncd.service
84ms systemd-journald.service
```
After made it configurable and set to 5s
```
❯ ssh sus@xx.xx.xx.xx
Last login: Sat Nov 14 18:41:42 2020 from 10.104.45.138
18:42:36 up 0 min, 0 users, load average: 0.16, 0.03, 0.01
> systemd-analyze blame
10.450s systemd-time-wait-sync.service
8.303s systemd-networkd-wait-online.service
1.621s cloud-init-local.service
1.068s cloud-init.service
```
This changes the retransmission timeout algorithm for requests
other than RENEW and REBIND. Previously, the retransmission timeout
started at 2 seconds, then doubling each retransmission up to a max
of 64 seconds. This is changed to match what RFC2131 section 4.1 describes,
which skips the initial 2 second timeout and starts with a 4 second timeout
instead. Note that -1 to +1 seconds of random 'fuzz' is added to each
timeout, in previous and current behavior.
This change is therefore slightly slower than the previous behavior in
attempting retransmissions when no server response is received, since the
first transmission times out in 4 seconds instead of 2.
Since TRANSIENT_FAILURE_ATTEMPTS is set to 3, the previous length of time
before a transient failure was reported back to systemd-networkd was
2 + 4 + 8 = 14 seconds, plus, on average, 3 seconds of random 'fuzz' for
a transient failure timeout between 11 and 17 seconds. Now, since the
first timeout starts at 4, the transient failure will be reported at
4 + 8 + 16 = 28 seconds, again plus 3 random seconds for a transient
failure timeout between 25 and 31 seconds.
Additionally, if MaxAttempts= is set, it will take slightly longer to
reach than with previous behavior.
Use the request timeout algorithm specified in RFC2131 section 4.4.5 for
handling timed out RENEW and REBIND requests.
This changes behavior, as previously only 2 RENEW and 2 REBIND requests
were sent, no matter how long the lease lifetime. Now, requests are
send according to the RFC, which results in starting with a timeout
of 1/2 the t1 or t2 period, and halving the timeout for each retry
down to a minimum of 60 seconds.
Fixes: #17909
The parsing of the dhcpv4 lease lifetime, as well as the t1/t2
times, is simplified by this commit.
This differs from previous behavior; previously, the lease lifetime and
t1/t2 values were modified by random 'fuzz' by subtracting 3, then adding
a random number between 0 and (slightly over) 2 seconds. The resulting
values were therefore always between 1-3 seconds shorter than the value
provided by the server (or the default, in case of t1/t2). Now, as
described in RFC2131, the random 'fuzz' is between -1 and +1 seconds,
meaning the actual t1 and t2 value will be up to 1 second earlier or
later than the server-provided (or default) t1/t2 value.
This also differs in handling the lease lifetime, as described above it
previously was adjusted by the random 'fuzz', but the RFC does not state
that the lease expiration time should be adjusted, so now the code uses
exactly the lease lifetime as provided by the server with no adjustment.
RFC2131, providing the details for dhcpv4, has specific retransmission
intervals that it outlines. This adds functions to compute the timeouts
as the RFC describes.
When something fails, we need some logs to figure out what happened.
This is primarily relevant for connection errors, but in general we
want to log about all errors, even if they are relatively unlikely.
We want one log on failure, and generally no logs on success.
The general idea is to not log in static functions, and to log in the
non-static functions. Non-static functions which call other functions
may thus log or not log as appropriate to have just one log entry in the
end.
The commit 6f3ac0d517 drops the prefix and
suffix in TAGS= property. But there exists several rules that have like
`TAGS=="*:tag:*"`. So, the property must be always prefixed and suffixed
with ":".
Fixes#17930.
The ret_size result is a bit of an awkward optimization that in a
sense enables bypassing the mmap-cache API, while encouraging
duplication of logic it already implements.
It's only utilized in one place; journal_file_move_to_object(),
apparently to avoid the overhead of remapping the whole object
again once its header, and thus its actual size, is known.
With mmap-cache's context cache, the overhead of simply
re-getting the object with the now known size should already be
negligible. So it's not clear what benefit this brings, unless
avoiding some function calls that do very little in the hot
context-cache hit case is of such a priority.
There's value in having all object-sized gets pass through
mmap_cache_get(), as it provides a single entrypoint for
instrumentation in profiling/statistics gathering. When
journal_file_move_to_object() bypasses getting the full object
size, you don't capture the full picture on the mmap-cache side
in terms of object sizes explicitly loaded from a journal file.
I'd like to see additional accounting in mmap_cache_get() in a
future commit, taking advantage of this change.
Quoting Andy Lutomirski:
> The upcoming Linux SGX driver has a device node /dev/sgx. User code opens
> it, does various setup things, mmaps it, and needs to be able to create
> PROT_EXEC mappings. This gets quite awkward if /dev is mounted noexec.
We already didn't use noexec in spawn, and this extends this behaviour to other
systems.
Afaik, the kernel would refuse execve() on a character or block device
anyway. Thus noexec on /dev matters only for actual binaries copied to /dev,
which requires root privileges in the first place.
We don't do noexec on either /tmp or /dev/shm (because that causes immediate
problems with stuff like Java and cffi). And if you have those two at your
disposal anyway, having noexec on /dev doesn't seem important. So the 'noexec'
attribute on /dev doesn't really mean much, since there are multiple other
similar directories which don't require root privileges to write to.
C.f. 33c10ef43b.
When an interface gains carrier but udev have not initialized the
interface or link_initialized_handler() has not been called yet,
then link_configure will be called twice. Thus LLDP client will be
configured twice, and triggers assertion.
Fixes#17929.
Zbigniew Jędrzejewski-Szmek