]> git.ipfire.org Git - thirdparty/systemd.git/log
thirdparty/systemd.git
12 months agofs-util: move attempts counter in openat_report_new() into loop 34850/head
Lennart Poettering [Tue, 22 Oct 2024 15:49:40 +0000 (17:49 +0200)] 
fs-util: move attempts counter in openat_report_new() into loop

12 months agolabel: move label_ops_reset() up a bit
Lennart Poettering [Tue, 22 Oct 2024 15:47:22 +0000 (17:47 +0200)] 
label: move label_ops_reset() up a bit

Let#s move it close to label_ops_set(), since it is somewhat symmetric
to it.

12 months agolabel: add missing assert() to label_ops_set()
Lennart Poettering [Tue, 22 Oct 2024 15:47:16 +0000 (17:47 +0200)] 
label: add missing assert() to label_ops_set()

12 months agoshared: get rid of fileio-label.[ch]
Lennart Poettering [Tue, 22 Oct 2024 08:25:26 +0000 (10:25 +0200)] 
shared: get rid of fileio-label.[ch]

Move the renaming function to reboot-util.h (since it writes out
/run/nologin at shutdown), and let's get rid of fileio-label.[ch] now
that it serves no purpose anymore.

12 months agofileio: port write_string_file_full() to openat_report_new()
Lennart Poettering [Tue, 22 Oct 2024 08:19:03 +0000 (10:19 +0200)] 
fileio: port write_string_file_full() to openat_report_new()

This brings two benefits: we will label the created file only if it is
actually created, and we can correctly delete any file we create again
on failure.

12 months agofileio: port write_string_file() to LabelOps, and thus add WRITE_STRING_FILE_LABEL...
Lennart Poettering [Tue, 22 Oct 2024 08:09:20 +0000 (10:09 +0200)] 
fileio: port write_string_file() to LabelOps, and thus add WRITE_STRING_FILE_LABEL flag

Given that we have the LabelOps abstraction these days, we can teach
write_string_file() to use it, which means we can get rid of
fileio-label.[ch] as a separate concept.

(The only reason that fileio-label.[ch] exists independently of
fileio.[ch] was that the former linekd to libselinux potentially, and
thus had to be in src/shared/ while the other always was in src/basic/.
But the LabelOps vtable provides us with a nice work-around)

12 months agofs-util: tweak how openat_report_new() operates when O_CREAT is used on a dangling...
Lennart Poettering [Mon, 21 Oct 2024 20:58:25 +0000 (22:58 +0200)] 
fs-util: tweak how openat_report_new() operates when O_CREAT is used on a dangling symlink

One of the big mistakes of Linux is that when you create a file with
open() and O_CREAT and the file already exists as dangling symlink that
the symlink will be followed and the file created that it points to.
This has resulted in many vulnerabilities, and triggered the creation of
the O_MOFOLLOW flag, addressing the problem.

O_NOFOLLOW is less than ideal in many ways, but in particular one: when
actually creating a file it makes sense to set, because it is a problem
to follow final symlinks in that case. But if the file is already
existing, it actually does make sense to follow the symlinks. With
openat_report_new() we distinguish these two cases anyway (the whole
function exists only to distinguish the create and the exists-already
case after all), hence let's do something about this: let's simply never
create files "through symlinks".

This can be implemented very easily: just pass O_NOFOLLOW to the 2nd
openat() call, where we actually create files.

And then basically remove 0dd82dab91eaac5e7b17bd5e9a1e07c6d2b78dca
again, because we don't need to care anymore, we already will see ELOOP
when we touch a symlink.

Note that this change means that openat_report_new() will thus start to
deviate from plain openat() behaviour in this one small detail: when
actually creating files we will *never* follow the symlink. That should
be a systematic improvement of security.

Fixes: #34088
12 months agofs-util: always call label post ops in xopenat_full(), in both success and error...
Lennart Poettering [Mon, 21 Oct 2024 20:43:18 +0000 (22:43 +0200)] 
fs-util: always call label post ops in xopenat_full(), in both success and error path

For SELinux it is essential that we reset the file creation label both
in the success and in the error path, hence do so.

Moreover, when calling the label post ops do it if possible with the
opened fd of the inode itself, rather than always going via its path,
simply to reduce the attack surface.

12 months agofs-util: don't second guess openat_report_new() return values
Lennart Poettering [Mon, 21 Oct 2024 20:40:52 +0000 (22:40 +0200)] 
fs-util: don't second guess openat_report_new() return values

If openat_report_new() fails, then 'made_file' will be false, as no file
was created, hence there's no need to skip the unlinkat() explicitly
early, given that we check for 'made_file' anyway in the error path. The
extra error code checks are hence entirely redundant.

12 months agolabel: tweak LabelOps post() hook to take "created" boolean
Lennart Poettering [Mon, 21 Oct 2024 20:07:56 +0000 (22:07 +0200)] 
label: tweak LabelOps post() hook to take "created" boolean

We have two distinct implementations of the post hook.

1. For SELinux we just reset the selinux label we told the kernel
   earlier to use for new inodes.

2. For SMACK we might apply an xattr to the specified file.

The two calls are quite different: the first call we want to call in all
cases (failure or success), the latter only if we actually managed to
create an inode, in which case it is called on the inode.

12 months agofs-util: always go through the unlink cleanup paths in xopenat_full()
Lennart Poettering [Mon, 21 Oct 2024 19:56:05 +0000 (21:56 +0200)] 
fs-util: always go through the unlink cleanup paths in xopenat_full()

We didn't go through it at all if label_ops_post() failed.

12 months agofs-util: remove misplaced RET_NERRNO()
Lennart Poettering [Mon, 21 Oct 2024 19:54:37 +0000 (21:54 +0200)] 
fs-util: remove misplaced RET_NERRNO()

12 months agoMerge pull request #34851 from DaanDeMeyer/medium
Daan De Meyer [Tue, 22 Oct 2024 11:37:59 +0000 (13:37 +0200)] 
Merge pull request #34851 from DaanDeMeyer/medium

bus-util: Return ENOMEDIUM if XDG_RUNTIME_DIR is unset

12 months agobus-util: Fix bus_log_connect_error() 34851/head
Daan De Meyer [Tue, 22 Oct 2024 10:32:02 +0000 (12:32 +0200)] 
bus-util: Fix bus_log_connect_error()

12 months agobus-util: Special case when DBUS_SESSION_BUS_ADDRESS is set and XDG_RUNTIME_DIR isn't
Daan De Meyer [Tue, 22 Oct 2024 09:12:17 +0000 (11:12 +0200)] 
bus-util: Special case when DBUS_SESSION_BUS_ADDRESS is set and XDG_RUNTIME_DIR isn't

We noticed some failures because we have code that connects to user
managers by setting DBUS_SESSION_BUS_ADDRESS without setting XDG_RUNTIME_DIR.
If that's the case, connect to the user session bus instead of the
private manager bus as we can't connect to the latter if XDG_RUNTIME_DIR
is not set.

12 months agobus-util: Return ENOMEDIUM if XDG_RUNTIME_DIR is unset
Daan De Meyer [Tue, 22 Oct 2024 08:59:27 +0000 (10:59 +0200)] 
bus-util: Return ENOMEDIUM if XDG_RUNTIME_DIR is unset

bus_log_connect_error() checks for ENOMEDIUM, not ENXIO.

12 months agoMerge pull request #34761 from ikruglov/ikruglov/io-systemd-Machine-GetAddresses
Lennart Poettering [Tue, 22 Oct 2024 07:06:56 +0000 (09:06 +0200)] 
Merge pull request #34761 from ikruglov/ikruglov/io-systemd-Machine-GetAddresses

machine: add Addresses, OSRelease, and UIDShift fields in varlink io.systemd.Machine.List output

12 months agoMerge pull request #30952 from rpigott/resolved-dnr
Lennart Poettering [Tue, 22 Oct 2024 07:05:36 +0000 (09:05 +0200)] 
Merge pull request #30952 from rpigott/resolved-dnr

RFC9463: Discovery of Network-designated Resolvers

12 months agotest: CET/EET are deprecated, use Europe/Berlin and Kyiv
Luca Boccassi [Mon, 21 Oct 2024 16:31:09 +0000 (17:31 +0100)] 
test: CET/EET are deprecated, use Europe/Berlin and Kyiv

The links moved to the legacy dataset so they won't be available by
default, so stop using them and just use the city ones instead

12 months agoMerge pull request #34628 from DaanDeMeyer/measure
Daan De Meyer [Mon, 21 Oct 2024 16:55:33 +0000 (18:55 +0200)] 
Merge pull request #34628 from DaanDeMeyer/measure

Rework TEST-86-MULTI-PROFILE-UKI + associated bugfixes

12 months agoresolve: move sd-* api into libsystemd-network 30952/head
Ronan Pigott [Wed, 26 Jun 2024 00:01:59 +0000 (17:01 -0700)] 
resolve: move sd-* api into libsystemd-network

This duplicates the svc param constants for the benefit of the
resolved-core library.

12 months agondisc: implement ndisc_option_build_encrypted_dns
Ronan Pigott [Thu, 4 Apr 2024 02:04:33 +0000 (19:04 -0700)] 
ndisc: implement ndisc_option_build_encrypted_dns

This is only used by the fuzzer so far.

12 months agonetwork: add dnr resolvers to networkctl status json output
Ronan Pigott [Wed, 3 Apr 2024 20:16:05 +0000 (13:16 -0700)] 
network: add dnr resolvers to networkctl status json output

12 months agotest/fuzz: add dnr packets
Ronan Pigott [Thu, 25 Jan 2024 06:49:03 +0000 (23:49 -0700)] 
test/fuzz: add dnr packets

The structure of DNR options is considerably more complicated than most
DHCP options, and as a result the fuzzer has poor coverage of these code
paths.

This adds some DNR packets to the fuzzing corpus, not with the intent of
capturing some specific edge case, but with the intent to rapidly
improve the fuzzers' coverage of these codepaths by giving it a valid
example to begin with.

Also include an ndisc router advert with a few Encrypted DNS options,
for the same purpose.

12 months agonetwork: Serialize ipv6ra DNR
Ronan Pigott [Mon, 22 Jan 2024 21:37:42 +0000 (14:37 -0700)] 
network: Serialize ipv6ra DNR

Serialize DNR servers acquired by ipv6ra option, same as the V4/V6 DNR
DHCP options.

12 months agonetwork: Introduce IPv6RA UseDNR= option
Ronan Pigott [Mon, 22 Jan 2024 21:37:42 +0000 (14:37 -0700)] 
network: Introduce IPv6RA UseDNR= option

Same as the DHCP v4/v6 options, this controls the use of DNR received
from ipv6ra.

12 months agondisc: Parse RFC9463 encrypted DNS (DNR) option
Ronan Pigott [Sat, 20 Jan 2024 01:26:26 +0000 (18:26 -0700)] 
ndisc: Parse RFC9463 encrypted DNS (DNR) option

This option is equivalent to the V4/V6 DNR options for DHCP.

12 months agotest-network: add DHCPv6 DNR test
Ronan Pigott [Thu, 18 Jan 2024 02:35:20 +0000 (19:35 -0700)] 
test-network: add DHCPv6 DNR test

Same as the DHCPv4 test.

12 months agonetwork: Serialize DHCPv6 DNR servers
Ronan Pigott [Thu, 18 Jan 2024 02:27:21 +0000 (19:27 -0700)] 
network: Serialize DHCPv6 DNR servers

This serializes DNR servers acquired by V6_DNR option, equivalent to the
V4_DNR option.

12 months agonetwork: Introduce UseDNR DHCPv6 option
Ronan Pigott [Thu, 18 Jan 2024 01:16:47 +0000 (18:16 -0700)] 
network: Introduce UseDNR DHCPv6 option

This is equivalent to the DHCPv4 option introduced earlier.

12 months agonetwork: Parse RFC9463 DHCPv6 DNR option
Ronan Pigott [Wed, 17 Jan 2024 23:49:02 +0000 (16:49 -0700)] 
network: Parse RFC9463 DHCPv6 DNR option

Implement the parsing for V6_DNR DHCPv6 option. This does the same as
the DHCP V4_DNR option.

12 months agodhcp6: use dns_name_from_wire_format
Ronan Pigott [Tue, 23 Jan 2024 20:12:30 +0000 (13:12 -0700)] 
dhcp6: use dns_name_from_wire_format

Convert some of the option parsing to use dns_name_from_wire_format,
introduced earlier. No change in behavior intended.

12 months agotest-dhcp6: terminate fqdn option
Ronan Pigott [Thu, 25 Jan 2024 00:53:35 +0000 (17:53 -0700)] 
test-dhcp6: terminate fqdn option

The encoded fqdn in this option must be properly terminated. We will
soon validate that this field is correctly encoded, so correct it in the
test.

12 months agotest-network: add test for DHCPv4 DNR
Ronan Pigott [Tue, 16 Jan 2024 23:07:54 +0000 (16:07 -0700)] 
test-network: add test for DHCPv4 DNR

This will test that networkd/resolved can understand the V4_DNR DHCP
option.

12 months agonetwork: Serialize DNR servers
Ronan Pigott [Tue, 16 Jan 2024 07:04:41 +0000 (00:04 -0700)] 
network: Serialize DNR servers

Implement serialization/deserialization for DNR servers. This re-uses
the string format in place for user configuration of DoT servers, and as
a consequence non-DoT servers are discarded when recording the link
configuration, for correctness.

This also enables sd-resolved to use these servers as it would other DNS
servers.

12 months agonetwork: Add serialization for DoT resolvers
Ronan Pigott [Sat, 24 Feb 2024 00:24:40 +0000 (17:24 -0700)] 
network: Add serialization for DoT resolvers

For now only DoT is supported, so DoT resolvers are represented using
the existing configuration format.

12 months agonetwork: Introduce UseDNR DHCPv4 option
Ronan Pigott [Tue, 16 Jan 2024 07:04:07 +0000 (00:04 -0700)] 
network: Introduce UseDNR DHCPv4 option

This option will control the use of DNR for choosing DNS servers on the
link. Defaults to the value of UseDNS so that in most cases they will be
toggled together.

12 months agomachine: add tests for Addresses/OSRelease/UIDShift fields in io.systemd.Machine... 34761/head
Ivan Kruglov [Thu, 17 Oct 2024 11:56:43 +0000 (13:56 +0200)] 
machine: add tests for Addresses/OSRelease/UIDShift fields in io.systemd.Machine.List output

12 months agomachine: use AcquireMetadata in io.systemd.MachineImage.List method
Ivan Kruglov [Thu, 17 Oct 2024 11:56:25 +0000 (13:56 +0200)] 
machine: use AcquireMetadata in io.systemd.MachineImage.List method

12 months agomachine: add Addresses field in varlink io.systemd.Machine.List output
Ivan Kruglov [Thu, 17 Oct 2024 11:54:07 +0000 (13:54 +0200)] 
machine: add Addresses field in varlink io.systemd.Machine.List output

This is equivalent to DBus implementation of GetMachineAddresses.

12 months agomachine: add OSRelease and UIDShift fields in varlink io.systemd.Machine.List output
Ivan Kruglov [Thu, 17 Oct 2024 11:51:34 +0000 (13:51 +0200)] 
machine: add OSRelease and UIDShift fields in varlink io.systemd.Machine.List output

This commit adds support of the above mentioned fields. This is equivalent to DBus implementation of:
- GetMachineOSRelease
- GetMachineUIDShift

12 months agomachine: enum AcquireMetadata
Ivan Kruglov [Thu, 17 Oct 2024 11:46:40 +0000 (13:46 +0200)] 
machine: enum AcquireMetadata

12 months agoRework TEST-86-MULTI-PROFILE-UKI 34628/head
Daan De Meyer [Fri, 4 Oct 2024 14:46:16 +0000 (16:46 +0200)] 
Rework TEST-86-MULTI-PROFILE-UKI

Now that mkosi supports generating UKI profiles, let's make use of
that to generate the UKI profiles required for the test instead of
doing it within the test itself.

12 months agoTEST-70-TPM2: Disable public key enrollment explicitly
Daan De Meyer [Mon, 21 Oct 2024 13:01:59 +0000 (15:01 +0200)] 
TEST-70-TPM2: Disable public key enrollment explicitly

Otherwise, when the test is executed on a system with signed PCRs,
cryptenroll will automatically pick up the public key from the UKI
which results in a volume that can't be unlocked because the pcrextend
tests appends extra things to pcr 11.

12 months agocryptenroll: Remove faulty assert()
Daan De Meyer [Sun, 20 Oct 2024 14:31:23 +0000 (16:31 +0200)] 
cryptenroll: Remove faulty assert()

We can break out of the preceeding for loop in certain scenarios
which would trigger the assert so let's drop it.

12 months agopcrlock: Take VirtualSize > SizeOfRawData into account
Daan De Meyer [Mon, 7 Oct 2024 15:39:27 +0000 (17:39 +0200)] 
pcrlock: Take VirtualSize > SizeOfRawData into account

If VirtualSize > SizeOfRawData, measure extra zeros to take into
account the extra zeros also measured by the stub.

12 months agomachine: rework Operation logic to reuse in varlink interface
Ivan Kruglov [Fri, 11 Oct 2024 14:26:38 +0000 (16:26 +0200)] 
machine: rework Operation logic to reuse in varlink interface

12 months agoMerge pull request #34403 from poettering/askpw-per-user
Lennart Poettering [Mon, 21 Oct 2024 14:37:28 +0000 (16:37 +0200)] 
Merge pull request #34403 from poettering/askpw-per-user

modernize the ask-password logic, and add unpriv askpw agents to the concept

12 months agoMerge pull request #34787 from yuwata/core-ip-address-allow-deny
Lennart Poettering [Mon, 21 Oct 2024 14:35:49 +0000 (16:35 +0200)] 
Merge pull request #34787 from yuwata/core-ip-address-allow-deny

core/cgroup: fix IPAddressAllow=/IPAddressDeny= set through DBus

12 months agomkosi: update opensuse commit reference
Luca Boccassi [Mon, 21 Oct 2024 13:26:05 +0000 (14:26 +0100)] 
mkosi: update opensuse commit reference

It was force-pushed again

12 months agovarlinkctl: respect $COLUMNS when rebreaking lines and we are not connected to a TTY
Lennart Poettering [Mon, 21 Oct 2024 09:10:35 +0000 (11:10 +0200)] 
varlinkctl: respect $COLUMNS when rebreaking lines and we are not connected to a TTY

Let's provide a mechanism to select the number of screen columns for
rebreaking comments in Varlink IDL connected to a TTY, by honouring the
$COLUMNS env var then too. Previously we'd only honour when connected to
a TTY, but it's also useful otherwise for rebreaking ridiculously long
comments, hence honour it in this case too.

12 months agotty-askpw-agent: modernize wall_tty_match() a bit 34403/head
Lennart Poettering [Mon, 21 Oct 2024 12:10:52 +0000 (14:10 +0200)] 
tty-askpw-agent: modernize wall_tty_match() a bit

12 months agoask-password-api: don't accidentally create a dir, when we don't want one
Lennart Poettering [Fri, 18 Oct 2024 19:36:34 +0000 (21:36 +0200)] 
ask-password-api: don't accidentally create a dir, when we don't want one

Previously, we were using touch(), which usually works fine, because the
path should always refer to an existing directory, in which case it just
updates the timestamp. However, if the dir does not exist yet (which
shouldn't happen), it would be created as regular file, which is just
wrong.

Hence, let's instead create the dir as dir if it is missing, and then
update its timestamp.

12 months agoman: update PASSWORD_AGENTS spec, and introduce unpriv pw queries
Lennart Poettering [Fri, 13 Sep 2024 12:55:49 +0000 (14:55 +0200)] 
man: update PASSWORD_AGENTS spec, and introduce unpriv pw queries

Fixes: #1232 #2217
12 months agoask-password-tool: add --user/--system flag to systemd-ask-password tool
Lennart Poettering [Fri, 13 Sep 2024 11:45:08 +0000 (13:45 +0200)] 
ask-password-tool: add --user/--system flag to systemd-ask-password tool

This allows selecting which agents to ask about this: system-level
agents, or per-user agents.

Fixes: #1232 #2217
12 months agoask-password-api: add support for querying pws from unpriv agents
Lennart Poettering [Fri, 13 Sep 2024 11:53:35 +0000 (13:53 +0200)] 
ask-password-api: add support for querying pws from unpriv agents

12 months agoask-password-api: minor modernizations
Lennart Poettering [Fri, 13 Sep 2024 11:49:25 +0000 (13:49 +0200)] 
ask-password-api: minor modernizations

12 months agotty-ask-password-agent: support for watching both system-wide and per-user askpw dir
Lennart Poettering [Fri, 13 Sep 2024 11:54:36 +0000 (13:54 +0200)] 
tty-ask-password-agent: support for watching both system-wide and per-user askpw dir

Fixes: #1232 #2217
12 months agotty-ask-password-agent: minor modernizations
Lennart Poettering [Fri, 13 Sep 2024 09:06:28 +0000 (11:06 +0200)] 
tty-ask-password-agent: minor modernizations

12 months agocore: modernize askpw handling a bit
Lennart Poettering [Fri, 13 Sep 2024 09:40:11 +0000 (11:40 +0200)] 
core: modernize askpw handling a bit

12 months agoMerge pull request #33398 from AdrianVovk/sysupdate-optional
Lennart Poettering [Mon, 21 Oct 2024 10:36:43 +0000 (12:36 +0200)] 
Merge pull request #33398 from AdrianVovk/sysupdate-optional

sysupdate: Add support for optional features

12 months agoMerge pull request #34667 from rpigott/resolved-bypass
Lennart Poettering [Mon, 21 Oct 2024 10:34:24 +0000 (12:34 +0200)] 
Merge pull request #34667 from rpigott/resolved-bypass

resolve: fixes for sd-resolved bypass

12 months agosysupdate: Use camelCase for JSON field names
Adrian Vovk [Fri, 18 Oct 2024 22:11:47 +0000 (18:11 -0400)] 
sysupdate: Use camelCase for JSON field names

Seems like we missed some snake_case field names in previous reviews of
systemd-sysupdate

12 months agomeasure: Take SizeOfImage into account as well for .linux section
Daan De Meyer [Fri, 4 Oct 2024 08:22:37 +0000 (10:22 +0200)] 
measure: Take SizeOfImage into account as well for .linux section

Same change as https://github.com/systemd/systemd/pull/34583 but for
systemd-measure. Otherwise we end up with PCR policy digest mismatches
as systemd-stub will measure the full virtual size of the kernel image
after it has been loaded while systemd-measure will disregard the extra
size introduced by SizeOfImage.

While ideally the stub would only measure the data that's actually on
disk and not the uninitialized data introduced by VirtualSize > SizeOfRawData,
we want newer systemd-measure to work with older stubs so we have to fix
systemd-measure and can't fix this in the stub.

12 months agotest: exercise bypass mode on the sd-resolved stub 34667/head
Ronan Pigott [Fri, 11 Oct 2024 05:40:03 +0000 (22:40 -0700)] 
test: exercise bypass mode on the sd-resolved stub

A basic test will verify that we provide the right flags.

12 months agoresolved: update condition for caching full packets
Ronan Pigott [Sat, 12 Oct 2024 17:00:56 +0000 (10:00 -0700)] 
resolved: update condition for caching full packets

Previously a full packet was cached only if the CD bit was set, but this
no longer corresponds to the cases where bypass is enabled.

Update the cache to retain a full packet in the cases where it might
actually be useful.

12 months agoresolved: enable CD bit without DO set
Ronan Pigott [Sat, 19 Oct 2024 04:10:57 +0000 (21:10 -0700)] 
resolved: enable CD bit without DO set

This is useful for a validating resolver to indicate to a non-validating
resolver when checking was disabled for the query. This matches the
behavior of the major public resovlers in response to queries with CD bu
tnot DO set.

12 months agoresolved: authenticate bypass queries
Ronan Pigott [Mon, 7 Oct 2024 18:05:18 +0000 (11:05 -0700)] 
resolved: authenticate bypass queries

Following 13e15dae9f0b, resolved does not forward the AD bit for bypass
queries, but resolved also didn't do it's own validation, making these
replies appear to never be authentic. We should enable validation for
bypass queries.

Let's disable our own validation when processing a +cd query, and also
ensure that it skips the cache so that we don't accidentally fail to
return inauthentic replies from upstream.

Previously, when we had a bypass transaction without cd, a cached,
authenticated, reply with cd could be served, leaving the cd bit
erroneously set in the reply. Only reply with a CD bit if the client
requested it.

Fixes: 13e15dae9f0b (resolved: clear the AD bit for bypass packets)
12 months agoTEST-55-OOMD: workaround for kernel regression in 6.12-rcX
Yu Watanabe [Fri, 11 Oct 2024 17:44:22 +0000 (02:44 +0900)] 
TEST-55-OOMD: workaround for kernel regression in 6.12-rcX

This ignore failures when running on kernel-6.12-rcX, which has a
regression in the kernel scheduler that breaks PSI.

From https://github.com/systemd/systemd/issues/32730#issuecomment-2415312260
> There is a known scheduler bug in 6.12 that breaks psi. It leaks
> "running tasks" counts, which matches your symptoms of seeing partial
> pressure only.
>
> Do you see "inconsistent task state" warnings in dmesg | grep psi?
>
> A fix is queued in the scheduler tree, should be sent to Linus shortly:
> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?id=c6508124193d42bbc3224571eb75bfa4c1821fbb

Workaround for #32730.

12 months agoupdatectl: Introduce optional feature verbs 33398/head
Adrian Vovk [Fri, 5 Jul 2024 00:03:09 +0000 (20:03 -0400)] 
updatectl: Introduce optional feature verbs

This introduces a nice UX for listing, inspecting, enabling, and
disabling optional features from the command line.

12 months agosysupdated: Plumb through optional features
Adrian Vovk [Tue, 2 Jul 2024 18:41:31 +0000 (14:41 -0400)] 
sysupdated: Plumb through optional features

This adds APIs to enumerate/inspect/enable/disable optional features.

12 months agosysupdate: Add verb to inspect features
Adrian Vovk [Wed, 3 Jul 2024 21:49:36 +0000 (17:49 -0400)] 
sysupdate: Add verb to inspect features

12 months agosysupdate: Add tests for optional features
Adrian Vovk [Tue, 2 Jul 2024 18:56:27 +0000 (14:56 -0400)] 
sysupdate: Add tests for optional features

Makes sure we don't regress on #33343 and #33344

12 months agosysupdate: Introduce optional features
Adrian Vovk [Sat, 22 Jun 2024 18:48:50 +0000 (14:48 -0400)] 
sysupdate: Introduce optional features

Optional features allow distros to define sets of transfers that can
be enabled or disabled by the system administrator. This is useful for
situations where a distro may want to ship some resources version-locked
to the core OS, but many people have no need for the resource, such as:
development tools/compilers, drivers for specialized hardware, language
packs, etc

We also rename sysupdate.d/*.conf -> sysupdate.d/*.transfer, because
now there are more than one type of definition in sysupdate.d/. For
backwards compat, we still load *.conf files as long as no *.transfer
files are found and the *.conf files don't try to declare themselves
as part of any features

Fixes https://github.com/systemd/systemd/issues/33343
Fixes https://github.com/systemd/systemd/issues/33344

12 months agofs-util: Introduce symlinkat_idempotent
Adrian Vovk [Fri, 18 Oct 2024 21:57:42 +0000 (17:57 -0400)] 
fs-util: Introduce symlinkat_idempotent

12 months agoMerge pull request #34820 from poettering/dissect-image-uclean
Yu Watanabe [Fri, 18 Oct 2024 17:15:01 +0000 (02:15 +0900)] 
Merge pull request #34820 from poettering/dissect-image-uclean

dissect-image: generate better log message for EUCLEAN dissect error

12 months agoMerge pull request #34783 from keszybz/man-nspawn-private-users
Zbigniew Jędrzejewski-Szmek [Fri, 18 Oct 2024 16:44:05 +0000 (18:44 +0200)] 
Merge pull request #34783 from keszybz/man-nspawn-private-users

Change systemd-nspawn man page to strongly recommend private users

12 months agotree-wise: use "lightweight" spelling 34783/head
Zbigniew Jędrzejewski-Szmek [Tue, 15 Oct 2024 16:54:28 +0000 (18:54 +0200)] 
tree-wise: use "lightweight" spelling

Both spellings were used, but the dictionary says that "lightweight"
is the standard spelling.

12 months agoman/systemd-nspawn: emphasise that user namespaces are strongly recommended
Zbigniew Jędrzejewski-Szmek [Tue, 15 Oct 2024 16:53:00 +0000 (18:53 +0200)] 
man/systemd-nspawn: emphasise that user namespaces are strongly recommended

12 months agotest: customize /etc/os-release instead of /usr/lib/os-release
Luca Boccassi [Fri, 18 Oct 2024 14:02:03 +0000 (15:02 +0100)] 
test: customize /etc/os-release instead of /usr/lib/os-release

As per spec image builders can create a local /etc/os-release
with per-image IDs, so modify that one instead of the original
one in /usr/lib. For example we do this when we build debian
unstable images in mkosi.

12 months agodissect-image: generate better log message for EUCLEAN dissect error 34820/head
Lennart Poettering [Fri, 18 Oct 2024 12:16:53 +0000 (14:16 +0200)] 
dissect-image: generate better log message for EUCLEAN dissect error

Fixes: #31799
12 months agodissect-image: uppercase first char of dissect error message systematically
Lennart Poettering [Fri, 18 Oct 2024 12:16:13 +0000 (14:16 +0200)] 
dissect-image: uppercase first char of dissect error message systematically

Some of the log message stricts used proper uppercasing, others didn't.
Fix that to make it uniform.

12 months agoresolved: refresh resolv.conf files when link goes away
Lennart Poettering [Fri, 18 Oct 2024 09:30:12 +0000 (11:30 +0200)] 
resolved: refresh resolv.conf files when link goes away

This might have the effect that some DNS server or search domain
disappears, hence rewrite the relevant files.

See: #27543

12 months agomkosi: update opensuse commit reference and switch branch
Luca Boccassi [Fri, 18 Oct 2024 10:35:56 +0000 (11:35 +0100)] 
mkosi: update opensuse commit reference and switch branch

'factory' was reset to 8 years ago

12 months agoresolved: add some more comments to varlink interface
Lennart Poettering [Fri, 18 Oct 2024 08:20:21 +0000 (10:20 +0200)] 
resolved: add some more comments to varlink interface

This is by no means complete, but gets us a bit closer.

12 months agonetworkd: raise limits on number of address 8x
Lennart Poettering [Fri, 18 Oct 2024 07:19:23 +0000 (09:19 +0200)] 
networkd: raise limits on number of address 8x

Limits should be enforced, but not in a way real setups collide with
them.

There have been multiple reports that current limits are too low, hence
raise them 8x.

Fixes: #24852
12 months agoupdate TODO
Lennart Poettering [Fri, 18 Oct 2024 07:54:32 +0000 (09:54 +0200)] 
update TODO

12 months agoupdate TODO
Lennart Poettering [Fri, 18 Oct 2024 07:23:54 +0000 (09:23 +0200)] 
update TODO

12 months agoGREEDY_REALLOC_APPEND: Make more type safe
Adrian Vovk [Wed, 4 Sep 2024 17:44:26 +0000 (13:44 -0400)] 
GREEDY_REALLOC_APPEND: Make more type safe

Previously, GREEDY_REALLOC_APPEND would compile perfectly fine and cause
subtle memory corruption if the caller messes up the type they're passing
in (i.e. by forgetting to pass-by-reference when appending a Type* to an
array of Type*). Now this will lead to compilation failure

12 months agoMerge pull request #34801 from poettering/async-sd-notify-close
Lennart Poettering [Thu, 17 Oct 2024 12:43:19 +0000 (14:43 +0200)] 
Merge pull request #34801 from poettering/async-sd-notify-close

pid1: close unexpected fds received via sd_notify() asyncronously

12 months agoMerge pull request #34795 from YHNdnzj/bump-kernel-baseline-5.4
Lennart Poettering [Thu, 17 Oct 2024 12:42:25 +0000 (14:42 +0200)] 
Merge pull request #34795 from YHNdnzj/bump-kernel-baseline-5.4

Bump kernel baseline to 5.4

12 months agoudev: consider serial ports as unconfigured only if both port and iomem_base sysattr...
Lennart Poettering [Thu, 17 Oct 2024 07:57:41 +0000 (09:57 +0200)] 
udev: consider serial ports as unconfigured only if both port and iomem_base sysattr is zero

Various archs do not know the "io port" concept, hence check for the
"iomem" configuration, too.

Fixes: #34800
12 months agopid1: close fds we receive via sd_notify() and cannot make use of asynchronously 34801/head
Lennart Poettering [Thu, 17 Oct 2024 07:46:16 +0000 (09:46 +0200)] 
pid1: close fds we receive via sd_notify() and cannot make use of asynchronously

This addresses #11112 fully. It mostly was addressed by
99620f457ed0886852ba18c9093b59767299121c already, but for fds not
even passed to the fdstore, this adds the missing asynchronous close
codepath.

Fixes: #11112
12 months agofdset: optionally, close remaining fds asynchronously
Lennart Poettering [Thu, 17 Oct 2024 07:43:37 +0000 (09:43 +0200)] 
fdset: optionally, close remaining fds asynchronously

12 months agoMerge pull request #34597 from ryantimwilson/oomd-pressure-duration
Yu Watanabe [Thu, 17 Oct 2024 05:30:13 +0000 (14:30 +0900)] 
Merge pull request #34597 from ryantimwilson/oomd-pressure-duration

Add ManagedOOMMemoryPressureDurationSec override setting for units

12 months agoReformat load-fragment-gperf.gperf.in 34597/head
Ryan Wilson [Tue, 15 Oct 2024 04:13:36 +0000 (21:13 -0700)] 
Reformat load-fragment-gperf.gperf.in

This commit reformats load-fragment-gperf.gperf.in after changes
made for adding the ManagedOOMMemoryPressureDurationSec= property.

12 months agocgroup: Add ManagedOOMMemoryPressureDurationSec= override setting for units
Ryan Wilson [Tue, 15 Oct 2024 03:49:54 +0000 (20:49 -0700)] 
cgroup: Add ManagedOOMMemoryPressureDurationSec= override setting for units

This will allow units (scopes/slices/services) to override the default
systemd-oomd setting DefaultMemoryPressureDurationSec=.

The semantics of ManagedOOMMemoryPressureDurationSec= are:
- If >= 1 second, overrides DefaultMemoryPressureDurationSec= from oomd.conf
- If is empty, uses DefaultMemoryPressureDurationSec= from oomd.conf
- Ignored if ManagedOOMMemoryPressure= is not "kill"
- Disallowed if < 1 second

Note the corresponding dbus property is DefaultMemoryPressureDurationUSec
which is in microseconds. This is consistent with other time-based
dbus properties.

12 months agooomd: Refactor DefaultMemoryPressureDurationSec= to use conf parser
Ryan Wilson [Wed, 16 Oct 2024 17:40:30 +0000 (10:40 -0700)] 
oomd: Refactor DefaultMemoryPressureDurationSec= to use conf parser

Parsing DefaultMemoryPressureDurationSec= is currently split between
conf parser, main() and manager_start() methods. This commit centralizes
parsing and bounds checking logic within a single custom conf parser
function.

12 months agoMerge pull request #34797 from yuwata/test-mount
Yu Watanabe [Wed, 16 Oct 2024 22:23:36 +0000 (07:23 +0900)] 
Merge pull request #34797 from yuwata/test-mount

TEST-60: several cleanups and potential performance improvement

12 months agoTEST-60-MOUNT-RATELIMIT: disable journal ratelimiting 34797/head
Yu Watanabe [Wed, 16 Oct 2024 13:34:48 +0000 (22:34 +0900)] 
TEST-60-MOUNT-RATELIMIT: disable journal ratelimiting

Otherwise, journal check in testcase_mount_ratelimit() may fail and we
need to wait 2 minutes.