Since c6552ad we now try to collect coverage even in situations where
it's basically impossible (like in test-mount-util where the whole / is
mounted as read-only). As dealing with this is not worth the trouble,
let's ignore the missing coverage errors thrown by gcov in such cases.
Mariusz Tkaczyk [Tue, 29 Mar 2022 10:49:54 +0000 (12:49 +0200)]
shutdown: get only active md arrays.
Current md_list_get() implementation filters all block devices, started from
"md*". This is ambiguous because list could contain:
- partitions created upon md device (mdXpY)
- external metadata container- specific type of md array.
For partitions there is no issue, because they aren't handle STOP_ARRAY
ioctl sent later. It generates misleading errors only.
Second case is more problematic because containers are not locked in kernel.
They are stopped even if container member array is active. For that reason
reboot or shutdown flow could be blocked because metadata manager cannot be
restarted after switch root on shutdown.
Add filters to remove partitions and containers from md_list. Partitions
can be excluded by DEVTYPE. Containers are determined by MD_LEVEL
property, we are excluding all with "container" value.
9afd5e7b975e8051c011ff9c07c95e80bd954469 introduced a build-time
taint, introduce a runtime one as well, in preparation for
removing support for unmerged-usr in a future release
hwdb: drop boilerplate about match patterns being unstable
We've had this text since the beginning, but in fact the patterns must be
stable in order for people to create local hwdb entries. And we support that
and can't change the match patterns without being very careful. So let's just
drop the text.
Previously, systemd-analyze verify would return 0 even if warnings
were raised during analysis of the specified units or their
dependencies. With 3cc3dc7, verify was changed to return 1 when
warnings were raised.
This commit changes the default mode to _RECURSIVE_ERRORS_INVALID
so that verify returns zero again by default when warnings are
raised.
Michal Sekletar [Wed, 23 Mar 2022 16:34:12 +0000 (17:34 +0100)]
udev/net_id: avoid slot based names only for single function devices
If we have two or more devices that share the same slot but they are
also multifunction then it is OK to use the slot information even if it
is the same for all of them. Name conflict will be avoided because we
will append function number and form names like, ens1f1, ens1f2...
meson: also allow setting GIT_VERSION via templates
GIT_VERSION is not available as a config.h variable, because it's rendered
into version.h during builds. Let's rework jinja2 rendering to also
parse version.h. No functional change, the new variable is so far unused.
I guess this will make partial rebuilds a bit slower, but it's useful
to be able to use the full version string.
A long name of one parameter was making the whole thing very wide.
I think that it's obvious from the context what the argument is,
so a shorter name should be just as good.
tests: add a smoke test for --version option in binaries
This is very similar to (and directly based on) the test for --help. I think
it's nice to do this: the test is very quick, but it'll catch cases where we
forgot to hook up the option, or forgot to exit after printing --version, and
it'll also increase our test coverage a bit.
README: say kernel 4.15 is the minimum recommended
After various long discussions
(https://lists.freedesktop.org/archives/systemd-devel/2022-March/047587.html,
https://lwn.net/Articles/889610/), there is no clear answer what the minimum
version should be. Bumping the version above 3.15 doesn't allow us to make any
significant simplifications (unless we went *much* higher). In particular, even
renameat2() is not fully supported with latest kernel versions, e.g. nfs still
doesn't have it. And the bpf stuff is optional anyway. So let's just say that
4.15 is what we recommend, because it provides fairly complete cgroups-v2, but
without any removals of compat in the code.
Nick Rosbrook [Mon, 4 Apr 2022 19:06:07 +0000 (15:06 -0400)]
oomd: calculate 'used' memory with MemAvailable instead of MemFree
The calculation for used memory in oomd_system_context_acquire is given
by MemTotal - MemFree from /proc/meminfo. This is too strict of a
calculation because it does not consider memory that is still available
for starting new applictions without swapping (MemAvailable). As a
result, systemd-oomd can start to kill processes before it is necessary.
This is more apparent on systems with low swap space.
Instead, compute 'used' memory as MemTotal - MemAvailable in
oomd_system_context_acquire and procfs_memory_get (which is used by
oomd_cgroup_context_acquire). And, rename oomd_mem_free_below to
oomd_mem_available_below for clarity.
man: add annotated example for bootctl status & list
This is based on the output on my laptop, with various manual adjustments.
If people have other types of entries, it'd be useful to add them here. In
particular, some dual-boot entries would be nice.
Strangely enough, having <varlistenetry>s outside of <variablelist> wasn't
causing visual problems. But having two <listitem>s in one <varlistentry>
resulted in the paragraphs running together in the rendered man page.
bootctl: unify boot entry loading for "status" and "list"
We must be consistent in the two listings, so let's split out the loading code
and call it from both verb_status() and verb_list(). This effectively makes
verb_status() also call efi_loader_get_entries().
There is still some code duplicated, but that's hard to avoid. Error messages
are made identical for the same errors in various places.
Let's make it easier to discern the "header" of device records from the
"body", i.e. non-property data from property data, by using some
conservative coloring.
udevadm: show more fields of sd_device objects in "udevadm info"
Let's make things easier to debug, and show a more comprehensive set of
fields, extending on the existing output syntax that starts with one
marker character followed by a colon and a space.
man: properly conditionalize kernel-install man page
Change f887eab1da85d0053321d43228042d90720eb77e conditionalized the
building of ther kenel-install man page in the generated meson output,
instead of the source in the XML markup. Thus, whenever the rules file
is updated the conditionalization is lost. Correct that.
Franck Bui [Fri, 18 Mar 2022 07:12:06 +0000 (08:12 +0100)]
journald: make sure journal_file_open() doesn't leave a corrupted file around after failing
This can be problematic especially when there's no more free disk
space. Consider the following:.
When disk space becomes sparse, writting to the system journal can lead to
error. In this case journald attempts to make room by rotating the journals,
which consists in archiving online journals and opening new ones.
However opening new files is likely to fail too and in this case
journal_file_open() leaves half initialized file around but in online
state. Then the error is propagated and journald switches into volatile mode.
Next time a new message is received by journald, it tries to open the
persistent system journal file to switch automatically back to persistent
mode.
When opening the system journal, journal_file_open(), called by
managed_journal_file_open_reliably(), finds the persistent system journal left
previously and assumes that it was uncleanly closed and considers it as
corrupted. The error is reported to managed_journal_file_open_reliably(), which
backs the file up and attempts to create a new system file, which fails and
leaves a corrupted system file again.
Since this is done for each message received by journald, /var/log/message can
be filled with backup files pretty quickly.
To prevent this, the patch makes sure to delete the newly created file in case
of error.
Let's order dev_t's by their major first, minor secondary. The binary
encoding of the two fields is weirdly interleaved and different in
kernel and glibc, hence let's focus on the generic part that works like
users would expect it.
So far the function is only used to compare for equality, not for
sorting, hence this has no immediate effect.
test: use the new `udevadm wait` verb to wait for the loop device
The original workaround didn't work, as `systemd-repart` kept failing
even when the `/dev/loopX` device was present:
```
[ 13.959419] H testsuite-58.sh[280]: + LOOP=/dev/loop1
[ 13.959636] H testsuite-58.sh[280]: + :
[ 13.959764] H testsuite-58.sh[280]: + test -e /dev/loop1
[ 13.959895] H testsuite-58.sh[280]: + break
[ 13.960023] H testsuite-58.sh[280]: + systemd-repart --pretty=yes --definitions=/tmp/testsuite-58-sector/ --seed=750b6cd5c4ae4012a15e7be3c29e6a47 --empty=require --dry-run=no /dev/loop1
[ 13.970538] H testsuite-58.sh[363]: Device '/dev/loop1' has no dm-crypt/dm-verity device, no need to look for underlying block device.
[ 13.970538] H testsuite-58.sh[363]: Failed to determine canonical path for '/dev/loop1': No such file or directory
[ 13.970538] H testsuite-58.sh[363]: Failed to open file or determine backing device of /dev/loop1: No such file or directory
```