Daan De Meyer [Mon, 27 Mar 2023 11:23:40 +0000 (13:23 +0200)]
mkosi: Narrow glob used to install python packages
Let's make sure we only install the python3 and python39 python
packages, instead of all the packages of all versions that are
packaged.
This also fixes the CentOS 8 CI because python3.11-pytest was failing
to install.
We have to ship our own powertools repo definition because we need to
enable module_hotfixes for powertools to coerce dnf into installing
some of the python packages.
Frantisek Sumsal [Sun, 26 Mar 2023 15:04:57 +0000 (17:04 +0200)]
test: verify the journal with and without a sealing key
The bit flips during journal verification cause various types of journal
corruptions, so it's useful to go through it even without a sealing key
to see how we handle corrupted stuff.
Also, provide a sealing key if running in "CI mode" (i.e. arguments),
to check the FSS-related codepaths in CIs as well.
Yu Watanabe [Sun, 26 Mar 2023 07:33:43 +0000 (16:33 +0900)]
bootctl: split-out entry token related definitions into boot-entry.[ch]
No functional change, just preparation for later commits.
These can be used in kernel-install later.
Note, unlike the our usual coding style, the arguments for
boot_entry_token_ensure() and parse_boot_entry_token_type() are
referenced, updated, and may freed, hence, always pass initialized
values. That's why they are not named as 'ret_xyz'.
Gertalitec [Wed, 22 Mar 2023 21:09:10 +0000 (01:09 +0400)]
Define $KERNEL_INSTALL_UKI_GENERATOR
Define $KERNEL_INSTALL_UKI_GENERATOR in case one wants it to be different from $KERNEL_INSTALL_INITRD_GENERATOR. This can be useful if one wants to use mkinitcpio / Dracut to generate the initrd, but without creating the UKI so this can be left for e.g. ukify or something else. Right now these initrd generators will read /etc/kernel/install.conf and generate the UKI
core/main: restore the correct assert about array position
'pos' is incremented after each assignment. If we use the maximum number
of arguments, we end up with pos==9 after all the assignments, and it
points to where the next value would be assigned. This position must remain
NULL.
The assert I "fixed" was intentionally introduced in 26abdc73a212b90f7c4b71808a1028d2e87ab09f as a bugfix. So my "fix" repeated
the same error that was fixed back then.
Frantisek Sumsal [Fri, 24 Mar 2023 20:15:28 +0000 (21:15 +0100)]
test: do an initial coverage capture
I noticed that our coverage reports miss some files completely - this
happens when the test doesn't touch the code in them at all, so the
generated coverage data (and resulting reports) have no information
about them. Let's fix this by doing an initial zero coverage capture
that contains a zeroed counter for every instrumented line in every
file, so when we later merge it with a capture from the test, it shows up
with a missing coverage instead of not showing at all.
Fixup for d2ebd50d7f9740dcf30e84efc75610af173967d2
and 6920049fad4fa39db5fec712f82f7f75b98fd4b9:
- add a comment that the last arg must be NULL and adjust the assert.
- move initialization around so that fields are declared,
initialized, and consumed in the same order.
- move declaration of pos adjacent do declaration of command_line.
This makes it easy to see that it was not initialized correctly.
- initialize buffers before writing the pointer into the args array.
This makes no difference for the compiler, but it just feels "wrong"
to do it in opposite order.
Because pos was off, we would ignore args after the timeout, and also
overwrite the buffer if enough args were used.
I think this is case shows clearly that declaring all variables at the
top of the function, with some initialized and other not, is very
error-prone. The compiler has no issue with declaring variables whereever,
and we should take advantage of this to make it keep declaration,
initialization, and use close. (Within reason of course.)
This is a little helper I used when preparing the tests for auxv
parsing. Just looking at hexdump output is pretty hard. We could
enhance it to display some specific data types better.
Franck Bui [Fri, 17 Mar 2023 13:18:50 +0000 (14:18 +0100)]
meson: make sure the unit test scripts find testdata/ even if they are not installed in the same directory
testdata/ is shared by both the unit tests and the integration tests. Hence it
makes sense to place them right under /usr/lib/systemd/tests/ while the tests
themself are located under dedicated sub-directories:
However the unit tests implemented as shell scripts (such as
test-fstab-generator.sh) expect to find testdata/ where the scripts are
installed, ie they do something like:
Franck Bui [Thu, 26 Jan 2023 15:07:17 +0000 (16:07 +0100)]
test: install unit tests in a dedicated subdirectory below '$testsdir'
/usr/lib/systemd/tests may contain more than the unit tests. For example on
SUSE we also install the integration tests there.
Putting the unit tests in a dedicated directory named 'unit-tests' makes the
layout cleaner.
Note that `run-unit-tests.py` has not been moved so we don't need to adjust
(Fedora) packaging and users also don't need to descend into the subdirectory.
So far we asked via a comment that @default should stay the first group
and @known the last group in the list. Let's enforce that statically, in
code, too.
seccomp-util: add new @sandbox syscall group with landlock/seccomp
Let's group these 4 syscalls, as they offer similar things and I guess
might be used in conjunction quite often, as they offer unprivileged
sandboxing.
No functional change. (We already checked for ELFCLASS32 or ELFCLASS64 before,
so even though there's a new check for other architectures, the only caller
only passes ELFCLASS32 or ELFCLASS64.)
Daan De Meyer [Thu, 23 Mar 2023 13:42:35 +0000 (14:42 +0100)]
loop-util: Add loop_device_make_by_path_at()
On top of taking a directory file descriptor, we use xopenat() so
that the function can also be used to work on existing file
descriptors to image files including all the logic to use O_DIRECT
and fallback to O_RDONLY if needed.
наб [Fri, 3 Mar 2023 14:57:42 +0000 (15:57 +0100)]
find-esp: don't silently error bootctl install if presumed XBOOTLDR part is stx_dev_major=0 but not btrfs
btrfs_get_block_device_fd() returns -ENOTTY if fstatfs().f_type !=
BTRFS_SUPER_MAGIC
btrfs_get_block_device_fd() is run by verify_fsroot_dir() by
verify_xbootldr() by find_xbootldr_and_warn() if
statx($presumed-XBOOTLDR).stx_dev_major == 0 ("maybe a btrfs device")
Every bootctl verb_install() runs find_xbootldr_and_warn(), by default
with /boot
If your /boot .stx_dev_major=0 but /not/ btrfs, bootctl install/update
quietly exits 1 with no note so as to what exactly failed (debug also
empty, and the strace isn't exactly clear since no syscall actually
failed)
This is the case on ZFS and the Debian filesystem layout: /boot/efi is
the ESP, and everything else under / is ZFS:
$ sudo env SYSTEMD_LOG_LEVEL=debug bootctl update
Found cgroup2 on /sys/fs/cgroup/, full unified hierarchy
Found container virtualization none.
File system "/boot" is not a FAT EFI System Partition (ESP) file system.
Using EFI System Partition at /boot/efi.
Checking whether /boot/efi/EFI/systemd/ contains any files…
$ echo $?
1
and funnier still:
$ sudo bootctl update --graceful
$ echo $?
1
Which is great, and also breaks postinst, which runs precisely the
latter, with no feedback at all
By checking for -ENOTTY we accept that the path being investigated
"is not it" if it's on ZFS (and any other filesystem that returns
.stx_dev_major == 0 but isn't btrfs)
The auxv metadata is unaligned, as the length of the prefix
"COREDUMP_PROC_AUXV=" is 19. Hence, parse_auxv{32,64}() may triger
an undefined behavior (or at least cause slow down), which can be
detected when running on an undefined behavior sanitizer.
This also introduces a macro to define `parse_auxv{32,64}()`.
ukify: allow uncompressed kernel images for UNAME detection on aarch64 and riscv64 (#26929)
Uncompressed aarch64 and riscv64 kernels have a different startpoint than x86.
Example output from ukify:
aarch64: NotImplementedError: unknown file format (starts with b'MZ@\xfa')
riscv64: NotImplementedError: unknown file format (starts with b'MZo\x10')
Add check for (b'MZ') to catch both in one call.
Daan De Meyer [Mon, 20 Mar 2023 12:26:57 +0000 (13:26 +0100)]
core: Settle log target if we're going to be closing all fds
Whenever we're going to close all file descriptors, we tend to close
the log and set it into open when needed mode. When this is done with
the logging target set to LOG_TARGET_AUTO, we run into issues because
for every logging call, we'll check if stderr is connected to the
journal to determine where to send the logging message. This check
obviously stops working when we close stderr, so we settle the log
target before we do that so that we keep using the same logging
target even after stderr is closed.