test: check if we correctly propagate /run mounts during switch root
Since 7c764d4 we bind mount certain directories during switch root
instead of moving the mount directly, and for /run we do this without
MS_REC. This, unfortunately, leaves all mounts under /run behind
in the old root, which breaks certain use cases.
The commit intended to fix a race reported at #28009. However,
unfortunately, it does not fix the root of the race, and reveals
the race in more simple setups. See reports in #28358.
Dan Streetman [Fri, 21 Jul 2023 15:23:22 +0000 (11:23 -0400)]
tpm2: instead of adjusting authValue trailing 0(s), trim them as required by tpm spec
To keep compatibility with any existing object authValues with trailing 0's,
change tpm2_get_pin_auth() to trim trailing 0's, which is what the TPM
implementation will do. This should retain compatibility with any existing
authValues that contain trailing 0's.
Note that any existing authValues with trailing 0's are unlikely to have worked
in the way that systemd uses them in object sealing, which is as a bind key for
the encryption (and policy) session. However, it is better to be compatible
with the TPM spec (and implementations) even if previously created objects that
are affected may not have worked.
fd-util: do not call statx() twice when it does not provide mount ID
Previously, in path_is_root_at(), if statx() does not provide mount ID,
path_get_mnt_id_at() was called, but it also calls statx(). Let's avoid
the second trial.
When newer glibc is used, but kernel does not support statx(), then
glibc try to fallback with fstatat(). That's quite similar to our
implementation, but the supported flags are different, and if
unsupported flags are specified, it returns EINVAL.
* We don't support filesystems smaller than 300MB anymore. Tiny
* filesystems have never been XFS' design target. This limit has been
* carefully calculated to prevent formatting with a log smaller than
* the "realistic" size.
*
* If the realistic log size is 64MB, there are four AGs, and the log
* AG should be at least 1/8 free after formatting, this gives us:
*
* 64MB * (8 / 7) * 4 = 293MB
So let's accommodate and bump the minimal XFS filesystem size to 300M.
Previously, if the priority is same, devlinks are always replaced by
newer events. The commit 331aa7aa15ee5dd12b369b276f575d521435eb52 changes
that to keep the existing devlink. That should not change any behavior
when the devices that request the same symlink do not have any
dependency, e.g. when /dev/sda1 and /dev/adb1 request the same
/dev/disk/by-label symlink, as there are no guarantee that which device
is processed first.
However, when devices has dependency, e.g. /dev/sda and /dev/sda1
request the same /dev/disk/by-label symlink, previously the symlink
always pointed to the partition, as the partition is always processed
later. But, 331aa7aa15ee5dd12b369b276f575d521435eb52 makes the symlink
point to the whole disk.
The change by 331aa7aa15ee5dd12b369b276f575d521435eb52 is crucial to
improve performance of devlink handling, especially when a system has
large number of disks with same label or so. Hence, cannot and should
not be reverted.
So, let's workaround the case, as such situation should happen only when
the disk is a hybrind ISO image, I guess.
When an explicit sector size is set by the user it is also necessary to set the
sector size of any loopback devices. If the sector size is not set for loopback
devices then it can cause file system creation to fail or run into odd issues.
run: disable --expand-environment by default for --scope
The intention was to have this option enabled by default everywhere,
but unfortunately at least one case was found where it breaks
compatibility of a program using systemd-run --scopes and expecting
variables not to be expanded:
Restore backward compatibility and make the option disabled by default
when --scope is used, and enabled by default for other types.
In case --expand-environment is not specified and a '$' character is
detected, print a warning to nudge users toward specifying the
parameter as needed. In the future we can then flip the default.
This fixes the test failure when invoked by a user.
===
Running ./systemd-tmpfiles --user on 'f /tmp/test-systemd-tmpfiles.1foag_ur/test-content.n_9r_xhm/arg - - - - %S'
expect: '/home/watanabe/.config'
actual: '/home/watanabe/.local/state'
Traceback (most recent call last):
File "/home/watanabe/git/systemd/test/test-systemd-tmpfiles.py", line 233, in <module>
test_valid_specifiers(user=True)
File "/home/watanabe/git/systemd/test/test-systemd-tmpfiles.py", line 135, in test_valid_specifiers
test_content('f {} - - - - %S',
File "/home/watanabe/git/systemd/test/test-systemd-tmpfiles.py", line 88, in test_content
assert content == expected
^^^^^^^^^^^^^^^^^^^
AssertionError
===
Template names can be learned from the filesystem, so there isn't a need
to parse the output of systemctl list-unit-files in this case. This
should accelerate the completion of some verbs like enable.
The existing caching policy isn't very sensible for this cache. We could
write a different policy, but I don't think there is much value in
caching these values, as in my experience the command used to generate
them is quick.
In the first stanza, despite the comment, the pattern given would
consider the cache invalid if it was more than 1 hour old.
The second stanza was also incorrect, since the output of `systemctl
--all` is not unit file paths, but unit names. When they were being
tested against the cachefile mtime, the test would always fail becuase
of the nonexistant file (hopefully).
In fact it's not very useful to test if the unit files have newer mtime
in this case anyway, since we are only caching their names. Also,
`systemctl --all` is an unfortunately slow operation to be used in
testing for the cache validity — we want this operation to at least be
faster than rebuilding the cache.
I've rewritten this stanza with my best guess at its original intent. It
now checks against the mtime of the parent directories in the search
path, which should be updated and cause the cache to rebuild when we
add, remove, or rename any unit files.
The previous definition was not quite appropriate for the library code
because it relied on the message domain set by textdomain() invocation
which is not necessarily the same message domain defined in
GETTEXT_PACKAGE macro.
The only code that uses _() so far is located in pam_systemd_home.c.
kernel-install: silently ignore unexpected arguments for 'remove' command
The shell script version of kernel-install silently ignored unexpected
arguments, but C version refused that. Unfortunately, Fedora's kernel
script specifies kernel file even for 'remove' command. Let's accept
extra arguments and silently ignore them to keep backward compatibility.
This fixes regression introduced in 5a0c810462 with which all requests
for link properties ended up with EINVAL as we kept hitting
the signature_is_single() assert in sd_bus_get_property().
core: copy the host's os-release for /run/host/os-release
Currently for portable services we automatically add a bind mount
os-release -> /run/host/os-release. This becomes problematic for the
soft-reboot case, as it's likely that portable services will be configured
to survive it, and thus would forever keep a reference to the old host's
os-release, which would be a problem because it becomes outdated, and also
it stops the old rootfs from being garbage collected.
Create a copy when the manager starts under /run/systemd/propagate instead,
and bind mount that for all services using RootDirectory=/RootImage=, so
that on soft-reboot the content gets updated (without creating a new file,
so the existing bind mounts will see the new content too).
This expands the /run/host/os-release protocol to more services, but I
think that's a nice thing to have too.
As mentioned in the NEWS entry, it seems to see very little use, but adds
complexity in our code. It was added mainly with the goal of making it easier
for people using grub2 to modify their boot configuration, but grub2 is gaining
support for BLS snippets. On the systemd side, we now have credentials. So
let's deprecate this, and if there's no outcry, remove it in a few releases.
Petr Menšík [Sat, 15 Jul 2023 02:11:25 +0000 (04:11 +0200)]
Include in manual what DNSSEC=no means in detail
https://www.rfc-editor.org/rfc/rfc4035.html#section-3.2.1 says
security-aware recursive name server MUST set DO bit when sending
requests. systemd-resolved does not do that by design. State it more
clearly in manual page. Unlike other implementations it disables not
only validation as it stated, but complete DNSSEC awareness.
analyze: don't warn about version spec compliant versions
This commits adds version_is_valid_versionspec and uses it in
analyze-compare-version.c.
version_is_valid_versionspec differs from version_is_valid in that it acepts
empty strings and since valid characters in a version spec version are all
ASCII letters and digits as well as "-.~^", but ",_+" allowed by
version_is_valid are not.
Also give a more specific warning message on invalid characters.
test-network: check for captive portals received via NDISC
This requires fairly recent radvd that supports sending RAs with captive
portals [0].
Also, this should hopefully provide coverage for issues like:
- https://github.com/systemd/systemd/issues/28229
- https://github.com/systemd/systemd/issues/28231
- https://github.com/systemd/systemd/issues/28277
Rebuilding whenever the cached parameter is not set forces each new
shell to rebuild the cache, which often defeates the purpose of caching
in the first place.
This used to work correctly, before the change was reverted in e09d0d46c297. In fact it is important to specify the manager explicity
in the completion because the argument is reused in the caching
policies. An empty argument here caused the completion to create
separate caches with and without the --system parameter. We can simplify
the given pattern a little here too.
It's been a while since we introduced Differential ShellCheck and it
proved to be quite useful (and in some ways even better than the shellcheck
run by super-linter). So, to have only one linter scream at us for not
knowing how to write bash properly, let's drop the super-linter's one in
favor of Differential ShellCheck.
Follow-up for https://github.com/systemd/systemd/pull/24328#pullrequestreview-1074127504
Some of the new units using systemd-pcrphase are missing the --graceful
flag which causes them to error if the tpm libraries are not installed.
Add --graceful just like in the other pcrphase units to make systemd-pcrphase
exit gracefully if the tpm libraries are missing.
test-network: correctly support running systemd-udevd from the build dir
We create the udevadm -> systemd-udevd symlink during the install phase,
so it doesn't exist in the just compiled tree. This worked in CI since
the symlink is manually created there post-build.
Given that ERRNO_IS_PRIVILEGE() also matches positive values,
make sure this macro is not called with arguments that do not have
errno semantics.
In this case the arguments passed to ERRNO_IS_PRIVILEGE() are the values
returned by read_one_line_file() which can legitimately return positive
values without errno semantics, so fix this by moving ERRNO_IS_PRIVILEGE()
invocations to the branches where the return values are known to be negative.
Dan Streetman [Sun, 16 Jul 2023 01:33:50 +0000 (21:33 -0400)]
test: avoid TEST-70 passphrase and password file mode complaints
Minor change, to adjust mode of /tmp/passphrase and /tmp/password test files to
avoid repeated warning logs that each file "...has 0644 mode that is too
permissive, please adjust the ownership and access mode."
Given that ERRNO_IS_DISCONNECT() also matches positive values,
make sure this macro is not called with arguments that do not have
errno semantics.
In this case the argument passed to ERRNO_IS_DISCONNECT() is the value
returned by manager_recv() which can legitimately return 1 without errno
semantics, so fix this by moving ERRNO_IS_DISCONNECT() invocation to the
branch where the return value is known to be negative.
Given that ERRNO_IS_DISCONNECT() also matches positive values,
make sure this macro is not called with arguments that do not have
errno semantics.
In this case the argument passed to ERRNO_IS_DISCONNECT() is the value
returned by bus_socket_process_watch_bind(), bus_socket_process_opening(),
and bus_socket_process_authenticating() which can legitimately return
positive values without errno semantics, so fix this by moving the
ERRNO_IS_DISCONNECT() invocation to the branch where the return value
is known to be negative.
Given that ERRNO_IS_DISCONNECT() also matches positive values,
make sure this macro is not called with arguments that do not have
errno semantics.
In this case the argument passed to ERRNO_IS_DISCONNECT() is the value
returned by socket_acquire_peer() which can legitimately return 1
without errno semantics, so fix this by moving ERRNO_IS_DISCONNECT()
invocation to the branch where the return value is known to be negative.
Given that ERRNO_IS_RESOURCE() also matches positive values,
make sure this macro is not called with arguments that do not have
errno semantics.
In this case the argument passed to ERRNO_IS_RESOURCE() is the value
returned by recurse_dir_at() which can legitimately return positive
values without errno semantics, so fix this by moving the ERRNO_IS_RESOURCE()
invocation to the branch where the return value is known to be negative.
Dan Streetman [Sat, 15 Jul 2023 12:30:40 +0000 (08:30 -0400)]
tpm2: add tpm2_get_pin_auth()
Add function to calculate the hash digest for a provided pin, and also verify
that the final byte in the digest is not 0. This is required because the TPM
will always remove all trailing 0's from an auth value before using it.
Dan Streetman [Fri, 14 Jul 2023 17:15:48 +0000 (13:15 -0400)]
tpm2: handle older tpm enrollments without a saved pcr bank
Older code did not save the pcr bank (i.e. pcr hash algorithm), and instead let
tpm2_unseal() find the best pcr bank to use. In commit 2cd9d57548b0dadd52523df486d33aa4cf7c3b84 we changed tpm2_unseal() to no longer
handle an unset pcr bank. This adds back in the handling of an unset pcr_bank
so older sealed data should continue to work.
Dan Streetman [Fri, 14 Jul 2023 15:21:43 +0000 (11:21 -0400)]
test: reduce the number of loops in tpm2 test_tpms_pcr_selection_mask_and_hash()
This test loops through masks, but is a relatively long test due to the
increment size between loops; this slightly increases the increment size (from
3->5) which greatly speeds up the test.