glibc (and also musl, though we do not officially support it yet)
silently ignores colon prefix in $TZ. Let's always not prefix the
timezone.
tzset(3) states:
> A nonempty value of TZ can be one of two formats, either of which can
> be preceded by a colon which is ignored.
Addresses https://github.com/systemd/systemd/pull/38876#discussion_r2384347594.
The next version of swtpm will support RSA4096, but it needs to be called
with a new parameter in order to do so. Try with it first, and if
execution fails, fallback to running without it.
This is especially needed for OBS builds, as the signing key is RSA4096
and cannot be changed by users, so the generated UKIs have RSA4096 signatures
for the pcrsig sections, and swtpm refuses them without the new support.
It helps nobody to break compatibility for a missing definition
for printing an error.
Just add the missing definition if not present, as it is already
done for thousands of others from the kernel, glibc, etc.
This partially reverts commit d8b60944f5.
Previously, an input string ends with short timezone spec e.g. WET,
was parsed by setting $TZ environment variable to the timezone.
But the timezone might be different from the original local timezone,
thus the result might not follow the timezone change in the original
local timezone.
This makes the check of the short timezone spec with tzname[] earlier,
then it is not necessary to load another timezone file for e.g. WET,
and provides expected time.
This also make it use SAVE_TIMEZONE macro and drop use of forking
process. This makes greatly improve performance when parsing string
that contains timezone different from the current local timezone.
Unfortunately, there is still one corner case that our test fails.
When tzdata is built with rearguard enabled, then at least
Africa/Windhoek timezone does not provide correct time, but time shifted
1 hour from the original.
Add proper description section for udev_enumerate_new(), udev_enumerate_ref(),
and udev_enumerate_unref() functions. The description matches the source code
comments for consistency.
Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
When -f/--force is used, wipefs does not call ioctl(BLKRRPART), hence
the partitions will remain in sysfs. Let's not use the option.
This also makes it use 'udevadm lock' for safety.
linux https://www.kernel.org/doc/html/latest/admin-guide/nfs/nfsroot.html
states `ntp0-ip` as parameter to `ip=` command line.
Until now systemd-network-generator would fail if ntp was provided with
-EINVAL due to trailing arguments.
Stay in line with kernel doc and handle ntp value properly.
Previously, we shift arrays on read and then shift back on write.
It is inefficient and not necessary.
Let's directly use the buffer that kernel provides as is.
The very initial implementation sets cmd with zero when the bitmap
length does not match. But, it is fixed by
793cf87de9
Fortunately, our code does not follow the outdated comment, but checks
cmd field correctly.
The methods are deprecated since kernel v4.20,
9b30049535
and they are trivial wrapper of ETHTOOL_GLINKSETTINGS and ETHTOOL_SLINKSETTINGS,
respectively. Hence, the fallback logic is nowadays completely
meaningless. Let's drop them.
Let's check both the per user machined and the system machined instead
of only the system machined. We give preference to the per user machined
and fall back to the system machined.
This is also preparation for the installer later, split out of #38764.
It makes the experience a lot nicer if our nice little tools aren't
constantly interrupted by log spew from the kernel.
Fixes: #34448
This allows Type 1 entries to take advantage of multiprofile efi files
by allowing the configuration of the profile associated with that entry.
It will now parse the profile key in a Type 1 config to select that
profile when booting that efi file instead of relaying in the cmdline
Signed-off-by: Itxaka <itxaka@kairos.io>
For "wizard" style interactive tools it's very annoying if they are
interrupted by kernel log output or PID1's status output. let's add some
infra to disable this temporarily. I decided to implement this as an IPC
service so that we can make this robust: if the client request the
muting dies we can automatically unmute again.
This is hence a tiny varlink service, but it can also be started
directly from the cmdline.
It seems
- the address sanitizer on fedora 42 reports false-positive, or
- probing partitions in libblkid 2.40.4 has a bug.
Not sure which causes the issue, but anyway the address sanitizer
kills udev-worker when sym_blkid_partition_get_name() is called
in udev-builtin-blkid.c.
```
systemd-udevd[488]: ==488==ERROR: AddressSanitizer: stack-buffer-underflow on address 0x7ffdd716e020 at pc 0x563e3ca66fcb bp 0x7ffdd716d970 sp 0x7ffdd716d968
systemd-udevd[488]: READ of size 8 at 0x7ffdd716e020 thread T0 ((udev-worker))
(snip)
systemd-udevd[488]: HINT: this may be a false positive if your program uses some custom stack unwind mechanism, swapcontext or vfork
systemd-udevd[488]: (longjmp and C++ exceptions *are* supported)
systemd-udevd[488]: SUMMARY: AddressSanitizer: stack-buffer-underflow (/usr/bin/udevadm+0x187fca) (BuildId: 1fb56dbdf0447aba1185d6e34560b782b76098be)
(snip)
systemd-udevd[488]: Command: (udev-worker)
systemd-udevd[488]: ==488==ABORTING
```
Let's bump fedora from 42 to 43.
Fixes mkosi fedora 42 job failures after c349edfe49.
In integration tests (for example TEST-85-NETWORK-NetworkctlTests)
LINK_OPERSTATE_RANGE_INVALID and required_for_online == -1 are serialized as
```
"RequiredForOnline": "true",
"RequiredOperationalStateForOnline": [null, null]
```
Such link should be reported as required_for_online=False and not
serialize nulls.
