]> git.ipfire.org Git - thirdparty/systemd.git/log
thirdparty/systemd.git
2 years agotpm2: add tpm2_index_to_handle() and tpm2_index_from_handle() 29234/head
Dan Streetman [Fri, 30 Jun 2023 16:52:10 +0000 (12:52 -0400)] 
tpm2: add tpm2_index_to_handle() and tpm2_index_from_handle()

Adjust the tpm2_esys_handle_from_tpm_handle() function into better-named
tpm2_index_to_handle(), which operates like tpm2_get_srk() but allows using any
handle index. Also add matching tpm2_index_from_handle().

Also change the references to 'location' in tpm2_persist_handle() to more
appropriate 'handle index'.

2 years agotpm2: add tpm2_serialize() and tpm2_deserialize()
Dan Streetman [Fri, 30 Jun 2023 17:42:25 +0000 (13:42 -0400)] 
tpm2: add tpm2_serialize() and tpm2_deserialize()

Add functions to perform serialization and deserialization of ESYS_TR objects.

2 years agotpm2: add tpm2_marshal_blob() and tpm2_unmarshal_blob()
Dan Streetman [Wed, 28 Jun 2023 15:46:31 +0000 (11:46 -0400)] 
tpm2: add tpm2_marshal_blob() and tpm2_unmarshal_blob()

Add functions to marshal and unmarshal our 'blob' object.

2 years agotpm2: move measurement log to /run/log/ (from /var/log/)
Lennart Poettering [Fri, 29 Sep 2023 09:59:40 +0000 (11:59 +0200)] 
tpm2: move measurement log to /run/log/ (from /var/log/)

I have no idea what went on in my mind when I used a path in /var/ for
the tpm2 event log we now keep for userspace measurements. The
measurements are only valid for the current boot, hence should not be
persisted (in particular as they cannot be rotated, hence should not
grow without bounds).

Fix that, simply move from /var/log/ to /run/log/.

2 years agoMerge pull request #29193 from keszybz/path-util-adjustment
Daan De Meyer [Fri, 29 Sep 2023 09:33:12 +0000 (11:33 +0200)] 
Merge pull request #29193 from keszybz/path-util-adjustment

Make unit mangling follow paths

2 years agoMerge pull request #29241 from poettering/pidref-watch
Daan De Meyer [Fri, 29 Sep 2023 09:32:19 +0000 (11:32 +0200)] 
Merge pull request #29241 from poettering/pidref-watch

pid1: move unit_watch_pid()/unit_unwatch_pid() logic over to PidRef

2 years agoukify: explicitly import attribute
Valentin Lefebvre [Wed, 27 Sep 2023 14:41:58 +0000 (16:41 +0200)] 
ukify: explicitly import attribute

* Explicitly import attributes rsa and serialization from cryptography.hazmat

Signed-off-by: Valentin Lefebvre <valentin.lefebvre@suse.com>
2 years agoMerge pull request #29183 from ddstreet/tpm2_openssl_functions
Daan De Meyer [Fri, 29 Sep 2023 08:25:39 +0000 (10:25 +0200)] 
Merge pull request #29183 from ddstreet/tpm2_openssl_functions

Add openssl functions for use by TPM2 sealing calculations

2 years agoautomount: fix unused value coverity warnings
Luca Boccassi [Fri, 29 Sep 2023 01:00:34 +0000 (02:00 +0100)] 
automount: fix unused value coverity warnings

'r' is no longer used, so no point in setting it before returning

CID#1522310
CID#1522312

Follow-up for bfeb10911e58bf8890eb7687cf12eddb09ab0c5d

2 years agofix: do not check/verify slice units if recursive errors are to be ignored
Maanya Goenka [Wed, 27 Sep 2023 15:44:04 +0000 (15:44 +0000)] 
fix: do not check/verify slice units if recursive errors are to be ignored

Before this fix, when recursive-errors was set to 'no' during a systemd-analyze
verification, the parent slice was checked regardless. The 'no' setting means that,
only the specified unit should be looked at and verified and errors in the slices should be
ignored. This commit fixes that issue.

Example:

Say we have a sample.service file:

[Unit]
Description=Sample Service

[Service]
ExecStart=/bin/echo "a"
Slice=support.slice

Before Change:

systemd-analyze verify --recursive-errors=no maanya/sample.service
Assertion 'u' failed at src/core/unit.c:153, function unit_has_name(). Aborting.
Aborted (core dumped)

After Change:
systemd-analyze verify --recursive-errors=no maanya/sample.service
{No errors}

2 years agoupdate TODO 29241/head
Lennart Poettering [Tue, 19 Sep 2023 14:22:05 +0000 (16:22 +0200)] 
update TODO

2 years agocore: move pid watch/unwatch logic of the service manager to pidfd
Lennart Poettering [Tue, 19 Sep 2023 19:58:55 +0000 (21:58 +0200)] 
core: move pid watch/unwatch logic of the service manager to pidfd

This makes sure unit_watch_pid() and unit_unwatch_pid() will track
processes by pidfd if supported. Also ports over some related code.
Should not really change behaviour.

Note that this does *not* add support waiting for POLLIN on the pidfds
as additional exit notification. This is left for a later commit (this
commit is already large enough), in particular as that would add new
logic and not just convert existing logic.

2 years agotest-watch-pid: use a real PID, not a made up one
Lennart Poettering [Wed, 20 Sep 2023 09:34:57 +0000 (11:34 +0200)] 
test-watch-pid: use a real PID, not a made up one

This matters once we track processes with pidfds rather than just pid_t,
because made up PIDs likely won't exist.

The essence of the test remains unmodified, we just use a real, existing
PID instead of 4711.

2 years agopidref: add pidref_verify() helper
Lennart Poettering [Tue, 19 Sep 2023 14:22:29 +0000 (16:22 +0200)] 
pidref: add pidref_verify() helper

This new helper can be used after reading process info from procfs, to
verify that the data that was just read actually matches the pidfd, and
does not belong to some new process that just reused the numeric PID of
the process we originally pinned.

2 years agopidref: add pidref_hash_ops
Lennart Poettering [Tue, 19 Sep 2023 14:11:34 +0000 (16:11 +0200)] 
pidref: add pidref_hash_ops

This adds a "hash_ops" structure, which allows using PidRef structures
as keys in Hashmap and Set objects.

2 years agopidref: add helpers for managing PidRef on the heap
Lennart Poettering [Tue, 19 Sep 2023 14:15:00 +0000 (16:15 +0200)] 
pidref: add helpers for managing PidRef on the heap

Usually we want to embed PidRef in other structures, but sometimes it
makes sense to allocate it on the heap in case it should be used
standalone. Add helpers for that.

Primary usecase: use as key in Hashmap objects, that for example map
process to unit objects in PID 1.

This adds pidref_free()/pidref_freep() for freeing such an allocated
struct, as well as pidref_dup() (for duplicating an existing PidRef
on the heap 1:1), and pidref_new_pid() (for allocating a new PidRef from a
PID).

2 years agopidref: add PIDREF_MAKE_FROM_PID()
Lennart Poettering [Tue, 19 Sep 2023 14:05:42 +0000 (16:05 +0200)] 
pidref: add PIDREF_MAKE_FROM_PID()

This helper truns a pid_t into a PidRef. It's different from
pidref_set_pid() in being "passive", i.e. it does not attempt to acquire
a pidfd for the pid.

This is useful when using the PidRef as a lookup key that shall also
work after a process is already dead, and hence no conversion to a pidfd
is possible anymore.

2 years agocgroup-util: add cg_read_pidref() helper
Lennart Poettering [Tue, 19 Sep 2023 14:05:01 +0000 (16:05 +0200)] 
cgroup-util: add cg_read_pidref() helper

Just like cg_read_pid() but returns a PidRef

2 years agoMerge pull request #29249 from poettering/pid1-error-message
Luca Boccassi [Thu, 28 Sep 2023 21:18:15 +0000 (22:18 +0100)] 
Merge pull request #29249 from poettering/pid1-error-message

pid1: refactoring of unit state machine logging and unit timer refactoring

2 years agoresolved: register ipv4only.arpa are private domain
Bertrand Jacquin [Wed, 27 Sep 2023 18:39:52 +0000 (19:39 +0100)] 
resolved: register ipv4only.arpa are private domain

From RFC 8880:

Because the 'ipv4only.arpa' zone has to be an insecure delegation,
DNSSEC cannot be used to protect these answers from tampering by
malicious devices on the path.

Consequently, the 'ipv4only.arpa' zone MUST be an insecure delegation to
give DNS64/NAT64 gateways the freedom to synthesize answers to those
queries at will, without the answers being rejected by DNSSEC-capable
resolvers. DNSSEC-capable resolvers that follow this specification MUST
NOT attempt to validate answers received in response to queries for the
IPv6 AAAA address records for 'ipv4only.arpa'. Note that the name
'ipv4only.arpa' has no use outside of being used for this special DNS
pseudo-query used to learn the DNS64/NAT64 address synthesis prefix, so
the lack of DNSSEC security for that name is not a problem.

See: https://datatracker.ietf.org/doc/html/rfc8880#name-security-considerations

2 years agoopenssl: add kdf_ss_derive() 29183/head
Dan Streetman [Fri, 7 Jul 2023 14:13:27 +0000 (10:13 -0400)] 
openssl: add kdf_ss_derive()

Add function to perform KDF-SS ("concat" KDF).

While Openssl allows a digest, HMAC, or KMAC for the auxiliary function H, this
currently only allows using a digest for H.

2 years agoopenssl: add ecc_edch()
Dan Streetman [Fri, 7 Jul 2023 14:11:07 +0000 (10:11 -0400)] 
openssl: add ecc_edch()

Add function to perform ECC EDCH.

2 years agoopenssl: add openssl_cipher_many()
Dan Streetman [Tue, 27 Jun 2023 19:04:59 +0000 (15:04 -0400)] 
openssl: add openssl_cipher_many()

Add function to perform openssl cipher operations.

2 years agoopenssl: add kdf_kb_hmac_derive()
Dan Streetman [Tue, 27 Jun 2023 19:04:59 +0000 (15:04 -0400)] 
openssl: add kdf_kb_hmac_derive()

Add function to perform key-based (KB) key derivation function (KDF) using
hash-based message authentication code (HMAC).

Also alphabetize openssl-util.c header list, and include string-util.h.

2 years agoopenssl: add rsa_oaep_encrypt_bytes()
Dan Streetman [Thu, 15 Jun 2023 12:19:51 +0000 (08:19 -0400)] 
openssl: add rsa_oaep_encrypt_bytes()

Add function to encrypt bytes, similar to rsa_encrypt_bytes() but using OAEP
(Optimal Asymmetric Encryption Padding).

2 years agoopenssl: add openssl_hmac_many()
Dan Streetman [Mon, 26 Jun 2023 21:40:18 +0000 (17:40 -0400)] 
openssl: add openssl_hmac_many()

Add function to perform HMAC on multiple buffers.

Also update test-openssl with associated testing, and replace some memcmp()
with memcmp_nn().

2 years agoopenssl: replace openssl_hash() with openssl_digest()
Dan Streetman [Wed, 5 Jul 2023 20:53:01 +0000 (16:53 -0400)] 
openssl: replace openssl_hash() with openssl_digest()

The openssl_hash() function was used only by string_hashnum(); change it to use
openssl_digest() instead.

2 years agoopenssl: add openssl_digest_many()
Dan Streetman [Wed, 5 Jul 2023 16:28:39 +0000 (12:28 -0400)] 
openssl: add openssl_digest_many()

Add function to perform openssl digest calculation on multiple buffers.

2 years agoopenssl: add openssl_digest_size()
Dan Streetman [Wed, 5 Jul 2023 16:59:47 +0000 (12:59 -0400)] 
openssl: add openssl_digest_size()

Add function to get digest hash size for provided digest name.

2 years agoMerge pull request #29361 from keszybz/kernel-install-work
Luca Boccassi [Thu, 28 Sep 2023 16:16:15 +0000 (17:16 +0100)] 
Merge pull request #29361 from keszybz/kernel-install-work

Advertise installkernel ↔ kernel-install duality

2 years agoRemove json_variant_merge_pair() in favor of json_variant_set_field_non_null()
Daan De Meyer [Thu, 28 Sep 2023 09:51:25 +0000 (11:51 +0200)] 
Remove json_variant_merge_pair() in favor of json_variant_set_field_non_null()

2 years agoMerge pull request #28545 from bluca/softreboot_survive
Luca Boccassi [Thu, 28 Sep 2023 16:12:03 +0000 (17:12 +0100)] 
Merge pull request #28545 from bluca/softreboot_survive

pid1: add SurviveFinalKillSignal= to skip units on final sigterm/sigkill spree

2 years agokmod-setup: Load virtiofs and virtio_pci early
Daan De Meyer [Thu, 28 Sep 2023 07:50:36 +0000 (09:50 +0200)] 
kmod-setup: Load virtiofs and virtio_pci early

There's no way for us to wait for specific virtiofs tags to appear,
so we have to try and make sure that the tags are all available by
the time we try to mount any virtiofs tag. Let's try to do that by
loading the necessary modules as early as we can.

2 years agoshow-logs: add assert and fix local variable type
наб [Thu, 28 Sep 2023 14:21:15 +0000 (16:21 +0200)] 
show-logs: add assert and fix local variable type

Follows-up for: 0693e6b246053d31c0eb405c6abe9db8a4d00aaf

#29355

2 years agocore: mark units as need daemon-reload if unit file operations are
Mike Yuan [Wed, 27 Sep 2023 10:38:10 +0000 (18:38 +0800)] 
core: mark units as need daemon-reload if unit file operations are
performed

systemctl would issue daemon-reload after unit file operations
(enable/disable/preset/...) succeed. However, such operations
are not atomic, meaning that the unit file state could still change
even if the operation generally fails, and the unit_file_state
cached by manager becomes outdated.

Fixes #29341

2 years agocore: improve error message when setting up service mounts
Luca Boccassi [Wed, 27 Sep 2023 17:35:26 +0000 (18:35 +0100)] 
core: improve error message when setting up service mounts

Right now we include the private working directory when we say some files
where not found, which is confusing. Strip it from the error string.

For example, with a BindPaths=/var/bar that does not exist on the host:

Before:

  foo.service: Failed to set up mount namespacing: /run/systemd/unit-root/var/bar: No such file or directory

After:

  foo.service: Failed to set up mount namespacing: /var/bar: No such file or directory

2 years agoMerge pull request #29295 from valentindavid/valentindavid/sysupdate-patterns-in...
Luca Boccassi [Thu, 28 Sep 2023 14:18:45 +0000 (15:18 +0100)] 
Merge pull request #29295 from valentindavid/valentindavid/sysupdate-patterns-in-directory

sysupdate: Allow patterns to match path with directories

2 years agoMerge pull request #29359 from poettering/bootctl-uki-measured
Luca Boccassi [Thu, 28 Sep 2023 14:18:12 +0000 (15:18 +0100)] 
Merge pull request #29359 from poettering/bootctl-uki-measured

bootctl: show whether we booted in a measured UKI in status output (plus some minor other stuff)

2 years agodocs: note root storage daemons can now also use SurviveFinalKillSignal=yes 28545/head
Luca Boccassi [Mon, 25 Sep 2023 23:18:00 +0000 (00:18 +0100)] 
docs: note root storage daemons can now also use SurviveFinalKillSignal=yes

2 years agotest: check soft-reboot behavior wrt argv[0][0] == '@'
Frantisek Sumsal [Sat, 23 Sep 2023 16:52:03 +0000 (18:52 +0200)] 
test: check soft-reboot behavior wrt argv[0][0] == '@'

2 years agopid1: add SurviveFinalKillSignal= to skip units on final sigterm/sigkill spree
Luca Boccassi [Thu, 27 Jul 2023 22:14:31 +0000 (23:14 +0100)] 
pid1: add SurviveFinalKillSignal= to skip units on final sigterm/sigkill spree

Add a new boolean for units, SurviveFinalKillSignal=yes/no. Units that
set it will not have their process receive the final sigterm/sigkill in
the shutdown phase.

This is implemented by checking if a process is part of a cgroup marked
with a user.survive_final_kill_signal xattr (or a trusted xattr if we
can't set a user one, which were added only in kernel v5.7 and are not
supported in CentOS 8).

2 years agoupdate TODO
Lennart Poettering [Thu, 28 Sep 2023 11:22:39 +0000 (13:22 +0200)] 
update TODO

2 years agoRework unit_name_mangle_with_suffix() to (very slightly) simplify the path 29193/head
Zbigniew Jędrzejewski-Szmek [Sat, 16 Sep 2023 18:22:29 +0000 (20:22 +0200)] 
Rework unit_name_mangle_with_suffix() to (very slightly) simplify the path

'systemctl status /../dev' now looks for 'dev.mount', not '-..-dev.service',
and 'systemctl status /../foo' looks for 'foo.mount', not '-..-foo.service'. I
think this much more useful. I think the escaping is not very useful, so I plan
to submit a later series which changes that behaviour. But I think this first
step here is already useful on its own.

Note that the patch is smaller than it seems: before, is_device_path() would
return true only for absolute paths, so moving of is_device_path() under the
path_is_absolute() conditional doesn't influence the logic.

2 years agoexec-util: print executed commands in do_execute() 29361/head
Zbigniew Jędrzejewski-Szmek [Wed, 27 Sep 2023 07:55:57 +0000 (09:55 +0200)] 
exec-util: print executed commands in do_execute()

kernel-install uses do_execute(). We would log whenever a spawned child
finished, but we would not log anything when the child is launched. When the
children log output without a prefix (as the kernel-install plugins do), it
is hard to see where that output is coming from.

2 years agokernel-install: describe usage as installkernel
Zbigniew Jędrzejewski-Szmek [Tue, 26 Sep 2023 19:54:18 +0000 (21:54 +0200)] 
kernel-install: describe usage as installkernel

For us, this is a compatibility mode, but most likely it is there to stay: the
kernel Makefile's install target expects to be able to call /bin/installkernel.
We want people who build their own kernels to use this, so that they use
kernel-install and get support for all the functionality provided by it,
including building of UKIs and other new features. So let's actually advertise
that this exists and works.

2 years agoresolved: never respond to .alt pseudo-TLD.
Bertrand Jacquin [Wed, 27 Sep 2023 18:39:52 +0000 (19:39 +0100)] 
resolved: never respond to .alt pseudo-TLD.

From RFC 9476:

Because names beneath .alt are in an alternative namespace, they have no
significance in the regular DNS context. DNS stub and recursive
resolvers do not need to look them up in the DNS context.

See: https://datatracker.ietf.org/doc/html/rfc9476#name-the-alt-namespace

2 years agobootctl: highlight SecureBoot enabled state in green 29359/head
Lennart Poettering [Thu, 28 Sep 2023 08:45:49 +0000 (10:45 +0200)] 
bootctl: highlight SecureBoot enabled state in green

2 years agobootctl: if we can't access the ESP, show this in regular status output
Lennart Poettering [Thu, 28 Sep 2023 08:45:39 +0000 (10:45 +0200)] 
bootctl: if we can't access the ESP, show this in regular status output

2 years agoMerge pull request #29333 from YHNdnzj/systemctl-warn-half-masked
Mike Yuan [Thu, 28 Sep 2023 09:58:21 +0000 (17:58 +0800)] 
Merge pull request #29333 from YHNdnzj/systemctl-warn-half-masked

systemctl-enable: warn if disabled/masked unit has active triggering units

2 years agosysupdate: Add documentation for new MatchPattern behavior 29295/head
Valentin David [Sun, 24 Sep 2023 15:30:41 +0000 (17:30 +0200)] 
sysupdate: Add documentation for new MatchPattern behavior

2 years agosysupdate: Allow patterns to match path with directories
Valentin David [Sun, 24 Sep 2023 12:35:59 +0000 (14:35 +0200)] 
sysupdate: Allow patterns to match path with directories

`MatchPattern` for regular-file and directory as target can now match
subdirectories This is useful to install files for examples in `.extra.d`
directories:

```
[Target]
Type=regular-file
Path=/EFI/Linux
PathRelativeTo=boot
MatchPattern=gnomeos_@v.efi.extra.d/apparmor.addon.efi
```

The if the directories in the path do not exist, they will be created.  Whereas
the part in `Path` is not created.

2 years agobootctl: report if have been booted with a measured UKI
Lennart Poettering [Thu, 28 Sep 2023 08:33:00 +0000 (10:33 +0200)] 
bootctl: report if have been booted with a measured UKI

Just expose the result of efi_measured_uki() to the user.

2 years agosystemctl-enable: warn if disabled/masked units has active triggering units 29333/head
Mike Yuan [Tue, 26 Sep 2023 15:21:23 +0000 (23:21 +0800)] 
systemctl-enable: warn if disabled/masked units has active triggering units

Closes #311

2 years agosystemctl-start: suppress the triggering unit warning when --no-warn
Mike Yuan [Tue, 26 Sep 2023 15:01:52 +0000 (23:01 +0800)] 
systemctl-start: suppress the triggering unit warning when --no-warn

2 years agosystemctl: clean up check_triggering_units
Mike Yuan [Tue, 26 Sep 2023 14:26:07 +0000 (22:26 +0800)] 
systemctl: clean up check_triggering_units

Preparation for #311

2 years agosystemctl: make unit_is_masked always query manager
Mike Yuan [Wed, 27 Sep 2023 15:43:47 +0000 (23:43 +0800)] 
systemctl: make unit_is_masked always query manager

2 years agosystemctl: don't duplicate string needlessly
Mike Yuan [Tue, 26 Sep 2023 13:46:15 +0000 (21:46 +0800)] 
systemctl: don't duplicate string needlessly

2 years agosystemctl: reflect that statically enabled units can be in .upholds/
Mike Yuan [Tue, 26 Sep 2023 13:41:29 +0000 (21:41 +0800)] 
systemctl: reflect that statically enabled units can be in .upholds/

Follow-up for 38f901791f3c4b1cbd04b71323bbef2fdab65f83

2 years agoMerge pull request #29353 from YHNdnzj/nft-followup
Luca Boccassi [Wed, 27 Sep 2023 20:02:43 +0000 (21:02 +0100)] 
Merge pull request #29353 from YHNdnzj/nft-followup

man/org.freedesktop.systemd1: add version info for NFTSet

2 years agoMerge pull request #29265 from YHNdnzj/sleep-util-refactor
Mike Yuan [Wed, 27 Sep 2023 19:06:48 +0000 (03:06 +0800)] 
Merge pull request #29265 from YHNdnzj/sleep-util-refactor

sleep-util: split into three and first round of cleanups

2 years agoman/org.freedesktop.systemd1: add version info for NFTSet 29353/head
Mike Yuan [Wed, 27 Sep 2023 19:04:28 +0000 (03:04 +0800)] 
man/org.freedesktop.systemd1: add version info for NFTSet

Follow-up for dc7d69b3c1eb4aa78a5ba2791c6e146a365c4092

2 years agocore/unit: use RET_GATHER in one more function
Mike Yuan [Wed, 27 Sep 2023 19:00:13 +0000 (03:00 +0800)] 
core/unit: use RET_GATHER in one more function

2 years agotest: testing for core NFTSet= feature
Topi Miettinen [Wed, 6 Sep 2023 18:48:02 +0000 (21:48 +0300)] 
test: testing for core NFTSet= feature

2 years agocore: add user and group to NFTSet=
Topi Miettinen [Sun, 22 May 2022 12:17:24 +0000 (15:17 +0300)] 
core: add user and group to NFTSet=

The benefit of using this setting is that user and group IDs, especially dynamic and random
IDs used by DynamicUser=, can be used in firewall configuration easily.

Example:

```
[Service]
NFTSet=user:inet:filter:serviceuser
```

Corresponding NFT rules:

```
table inet filter {
        set serviceuser {
                typeof meta skuid
        }
        chain service_output {
                meta skuid @serviceuser accept
                drop
        }
}
```

```
$ cat /etc/systemd/system/dunft.service
[Service]
DynamicUser=yes
NFTSet=user:inet:filter:serviceuser
ExecStart=/bin/sleep 1000

[Install]
WantedBy=multi-user.target
$ sudo nft list set inet filter serviceuser
table inet filter {
        set serviceuser {
                typeof meta skuid
                elements = { 64864 }
        }
}
$ ps -n --format user,group,pid,command -p `systemctl show dunft.service -P MainPID`
    USER    GROUP     PID COMMAND
   64864    64864   55158 /bin/sleep 1000
```

2 years agocore: firewall integration of cgroups with NFTSet=
Topi Miettinen [Sat, 2 Sep 2023 18:55:36 +0000 (21:55 +0300)] 
core: firewall integration of cgroups with NFTSet=

New directive `NFTSet=` provides a method for integrating dynamic cgroup IDs
into firewall rules with NFT sets. The benefit of using this setting is to be
able to use control group as a selector in firewall rules easily and this in
turn allows more fine grained filtering. Also, NFT rules for cgroup matching
use numeric cgroup IDs, which change every time a service is restarted, making
them hard to use in systemd environment.

This option expects a whitespace separated list of NFT set definitions. Each
definition consists of a colon-separated tuple of source type (only "cgroup"),
NFT address family (one of "arp", "bridge", "inet", "ip", "ip6", or "netdev"),
table name and set name. The names of tables and sets must conform to lexical
restrictions of NFT table names. The type of the element used in the NFT filter
must be "cgroupsv2". When a control group for a unit is realized, the cgroup ID
will be appended to the NFT sets and it will be be removed when the control
group is removed.  systemd only inserts elements to (or removes from) the sets,
so the related NFT rules, tables and sets must be prepared elsewhere in
advance.  Failures to manage the sets will be ignored.

If the firewall rules are reinstalled so that the contents of NFT sets are
destroyed, command systemctl daemon-reload can be used to refill the sets.

Example:

```
table inet filter {
...
        set timesyncd {
                type cgroupsv2
        }

        chain ntp_output {
                socket cgroupv2 != @timesyncd counter drop
                accept
        }
...
}
```

/etc/systemd/system/systemd-timesyncd.service.d/override.conf
```
[Service]
NFTSet=cgroup:inet:filter:timesyncd
```

```
$ sudo nft list set inet filter timesyncd
table inet filter {
        set timesyncd {
                type cgroupsv2
                elements = { "system.slice/systemd-timesyncd.service" }
        }
}
```

2 years agoupdate TODO
Lennart Poettering [Wed, 27 Sep 2023 17:08:56 +0000 (19:08 +0200)] 
update TODO

2 years agocore: generalize service_arm_timer() for all unit types 29249/head
Lennart Poettering [Wed, 20 Sep 2023 15:38:23 +0000 (17:38 +0200)] 
core: generalize service_arm_timer() for all unit types

2 years agoscope: also modernize state machine logging
Lennart Poettering [Wed, 20 Sep 2023 15:38:46 +0000 (17:38 +0200)] 
scope: also modernize state machine logging

2 years agopath: also modernize path state machine logging
Lennart Poettering [Wed, 20 Sep 2023 15:07:32 +0000 (17:07 +0200)] 
path: also modernize path state machine logging

2 years agotimer: also modernize timer state machine error logging
Lennart Poettering [Wed, 20 Sep 2023 15:04:41 +0000 (17:04 +0200)] 
timer: also modernize timer state machine error logging

2 years agoautomount: also modernize log logic
Lennart Poettering [Wed, 20 Sep 2023 15:00:52 +0000 (17:00 +0200)] 
automount: also modernize log logic

2 years agoswap: also modernize state engine log message generation
Lennart Poettering [Wed, 20 Sep 2023 14:46:28 +0000 (16:46 +0200)] 
swap: also modernize state engine log message generation

2 years agomount: also rework log message generation
Lennart Poettering [Wed, 20 Sep 2023 13:59:24 +0000 (15:59 +0200)] 
mount: also rework log message generation

2 years agosocket: clean up error message generation/fail paths also for the socket state engine
Lennart Poettering [Wed, 20 Sep 2023 13:09:00 +0000 (15:09 +0200)] 
socket: clean up error message generation/fail paths also for the socket state engine

2 years agosocket: modernize socket_acquire_peer() a bit
Lennart Poettering [Wed, 20 Sep 2023 13:07:26 +0000 (15:07 +0200)] 
socket: modernize socket_acquire_peer() a bit

2 years agosocket: drop redundant TAKE_FD(cfd) line
Lennart Poettering [Wed, 27 Sep 2023 15:21:51 +0000 (17:21 +0200)] 
socket: drop redundant TAKE_FD(cfd) line

In the only two codepaths we reach this place we know that cfd is
already invalidated. In the Accept=yes case there's already a
TAKE_FD() a few lines further up, and in the Accept=no case there is no
connection fd anyway.

2 years agoservice: add error handling for all service_arm_timer() invocations
Lennart Poettering [Wed, 20 Sep 2023 11:11:39 +0000 (13:11 +0200)] 
service: add error handling for all service_arm_timer() invocations

Let's clean this up a bit, and catch all errors and do something
reasonable in case this happens.

2 years agoservice: clean up logging a bit
Lennart Poettering [Wed, 20 Sep 2023 11:09:42 +0000 (13:09 +0200)] 
service: clean up logging a bit

This rearranges various cases of "goto fail" in service.c: sometimes the
whole "goto fail" logic was redundant, since only jumped to form a
single place. Sometimes the log message was generated in the fail
section, instead of the place jumped to from, which resulted in
duplicate or misleading error messages.

No real codeflow changes, just refactoring primarily around log
messages.

2 years agoMerge pull request #29345 from poettering/measured-uki-condition
Lennart Poettering [Wed, 27 Sep 2023 14:39:46 +0000 (16:39 +0200)] 
Merge pull request #29345 from poettering/measured-uki-condition

pid1: introduce ConditionSecurity=measured-uki

2 years agosleep-config: add explanatory comment on "modes" 29265/head
Mike Yuan [Sun, 24 Sep 2023 10:28:54 +0000 (18:28 +0800)] 
sleep-config: add explanatory comment on "modes"

2 years agosleep-config: several cleanups
Mike Yuan [Thu, 21 Sep 2023 06:59:26 +0000 (14:59 +0800)] 
sleep-config: several cleanups

* Rename free_sleep_config to sleep_config_free
* Rearrange functions
* Make SleepConfig.modes and .states only contain
  operations that needs configuration
* Add missing assert

2 years agosleep/battery-capacity: rearrange functions
Mike Yuan [Thu, 21 Sep 2023 05:45:05 +0000 (13:45 +0800)] 
sleep/battery-capacity: rearrange functions

2 years agosleep/battery-capacity: drop unused error-handling
Mike Yuan [Thu, 21 Sep 2023 05:43:31 +0000 (13:43 +0800)] 
sleep/battery-capacity: drop unused error-handling

2 years agosleep/battery-capacity: don't report we have trip alarm if no battery is found
Mike Yuan [Wed, 20 Sep 2023 13:49:31 +0000 (21:49 +0800)] 
sleep/battery-capacity: don't report we have trip alarm if no battery is found

2 years agosleep-util: split into sleep-config and hibernate-util
Mike Yuan [Wed, 27 Sep 2023 13:35:45 +0000 (21:35 +0800)] 
sleep-util: split into sleep-config and hibernate-util

2 years agosleep-util: split battery-capacity into sleep/
Mike Yuan [Wed, 27 Sep 2023 13:35:19 +0000 (21:35 +0800)] 
sleep-util: split battery-capacity into sleep/

This is only used by sleep.c. Let's start shrinking down
the "mixed" sleep-util.

2 years agosleep: rebreak lines in check_wakeup_type
Mike Yuan [Fri, 22 Sep 2023 14:52:56 +0000 (22:52 +0800)] 
sleep: rebreak lines in check_wakeup_type

2 years agosleep-util: move check_wakeup_type to sleep/sleep
Mike Yuan [Wed, 20 Sep 2023 13:19:07 +0000 (21:19 +0800)] 
sleep-util: move check_wakeup_type to sleep/sleep

2 years agoMerge pull request #29134 from nabijaczleweli/short-iso-timestamp
Luca Boccassi [Wed, 27 Sep 2023 13:42:27 +0000 (14:42 +0100)] 
Merge pull request #29134 from nabijaczleweli/short-iso-timestamp

journalctl -o short-iso[-precise]: timezone as +02:00 instead of +0200

2 years agoMerge pull request #29296 from yuwata/sd-journal-several-cleanups-for-boot-id
Zbigniew Jędrzejewski-Szmek [Wed, 27 Sep 2023 12:56:48 +0000 (14:56 +0200)] 
Merge pull request #29296 from yuwata/sd-journal-several-cleanups-for-boot-id

2 years agotest: introduce TEST-09-REBOOT
Frantisek Sumsal [Sat, 23 Sep 2023 16:17:04 +0000 (18:17 +0200)] 
test: introduce TEST-09-REBOOT

To test stuff involving state preserved across (multiple) reboots, like
journal boot IDs.

2 years agoMerge pull request #29296 from keszybz/make-cryptsetup-offical-and-add-docs
Luca Boccassi [Wed, 27 Sep 2023 12:31:11 +0000 (13:31 +0100)] 
Merge pull request #29296 from keszybz/make-cryptsetup-offical-and-add-docs

Make cryptsetup offical and add docs

2 years agounits/blockdev@.target: conflict with umount.target
Mike Yuan [Tue, 26 Sep 2023 22:26:38 +0000 (06:26 +0800)] 
units/blockdev@.target: conflict with umount.target

Follow-up for d120ce478dc0043c89899799b5c1aaf62901bea9

blockdev@.target is used as a synchronization point between
the mount unit and corresponding systemd-cryptsetup@.service.
After the mentioned commit, it doesn't get a stop job enqueued
during shutdown, and thus the stop job for systemd-cryptsetup@.service
could be run before the mount unit is stopped.

Therefore, let's make blockdev@.target conflict with umount.target,
which is also what systemd-cryptsetup@.service does.

Fixes #29336

2 years agoupdate TODO 29345/head
Lennart Poettering [Wed, 27 Sep 2023 10:17:18 +0000 (12:17 +0200)] 
update TODO

2 years agoefi-loader: add caching to efi_measured_uki()
Lennart Poettering [Wed, 27 Sep 2023 10:14:56 +0000 (12:14 +0200)] 
efi-loader: add caching to efi_measured_uki()

EFI variable access is slow, hence let's avoid it if there's no need.
Let's cache the result of efi_measured_uki() so that we don't have to go
to the EFI variables each time.

This only caches in the yes/no case. If we encounter an error we don't
cache, so that we go to disk again.

This should optimize things a bit given we now have a bunch of services
which are conditioned with this at boot.

2 years agounits: move units over to ConditionSecurity=measured-uki
Lennart Poettering [Wed, 27 Sep 2023 10:13:26 +0000 (12:13 +0200)] 
units: move units over to ConditionSecurity=measured-uki

2 years agocondition: add ConditionSecurity=measured-uki
Lennart Poettering [Wed, 27 Sep 2023 10:08:41 +0000 (12:08 +0200)] 
condition: add ConditionSecurity=measured-uki

We have various services that should only run in a measured UKI
environment. Let's add an explicit high-level check for that.

2 years agoefi-loader: rename efi_stub_measured() → efi_measured_uki()
Lennart Poettering [Wed, 27 Sep 2023 09:50:14 +0000 (11:50 +0200)] 
efi-loader: rename efi_stub_measured() → efi_measured_uki()

Let's say "uki" rather than "stub", since that is just too generic, and
we shouldn't limit us to our own stub anyway, but generally define a
concept of a "measured UKI", which is a UKI that measures its part to
PCR 11.

This is mostly preparation for exposing this check to the user via
ConditionSecurity=.

2 years agondisc: Also set link hoplimit
Susant Sahani [Fri, 8 Sep 2023 09:00:38 +0000 (14:30 +0530)] 
ndisc: Also set link hoplimit

The per route hop limit does not place the hop limit in the IPv6 header.
https://github.com/systemd/systemd/issues/28437#issuecomment-1711055181

2 years agoRevert "userdbd: Order systemd-userdbd.service after systemd-remount-fs.service"
Lennart Poettering [Tue, 5 Sep 2023 08:17:41 +0000 (10:17 +0200)] 
Revert "userdbd: Order systemd-userdbd.service after systemd-remount-fs.service"

This reverts commit 9dd88582813b6dbeea6ce336f70cae681e6cbfc6.