Daan De Meyer [Sat, 12 Aug 2023 11:30:46 +0000 (13:30 +0200)]
repart: Use 4096 as the fallback sector size for verity/luks/filesystems
When we don't know the sector size of the actual block device, because
we're building an image in a loopback file and no sector size was specified
explicitly, let's use 4096 as the sector size for filesystems, verity and
LUKS. This should be the most compatible option, since 4096 will also work
on devices with sector size 512 or 2048.
For the actual GPT partition table size, we stick with 512 as the default
value since UEFI firmware and the kernel will only try to read the GPT
partition table from the first LBA on the device and the sector size for
most devices is still 512. It can also be trivially modified when copying
the image to another device using --copy-from + --sector-size.
(I guess in the original patch authors usecase the root fs actually
*does* remain in memory, but that's a special case and does not belong
in the man pages this way).
This change makes sure a data copy using copy_bytes() does not exceed the
max_bytes value when using COPY_HOLES and max_bytes stops before the next
data section.
Yu Watanabe [Thu, 10 Aug 2023 19:48:01 +0000 (04:48 +0900)]
coredump: fix various invalid memory access
Previously, we did not check error from iovw_put(). If it fails, the
target iovw may have no iov or partial iovs from the journal importar.
So, the finalization may cause underflow and may access and free invalid
memory.
vconsole-setup: use "@kernel" rather than "kernel" as special string to leave keymap untouched
This is a magic string, and we should avoid stepping into the territory
of normal keymap names with that, given that users can pick names
otherwise freely.
Hence, prefix the name with a special char to avoid any namespace
issues.
file-io: let's use offsetof() rather than sizeof() for determining EFI_FILE_INFO prefix size
The gnu-efi definition of the struct uses [1], our local one [0] to size
the filename array. Let's avoid an ambiguity and use offsetof() so that
this difference doesn't matter. Also, doing it this way makes very clear
to the read what happens here: it's a structure with a variable size
suffix.
Daan De Meyer [Thu, 10 Aug 2023 15:05:55 +0000 (17:05 +0200)]
ukify: Use length= instead of ignore_padding= in inspect
ignore_padding= was only added in a recent version of pefile. Let's
set length= to the virtual size instead which is what ignore_padding
does behind the scenes so we're compatible with older versions of
pefile.
Michal Koutný [Wed, 9 Aug 2023 19:31:58 +0000 (21:31 +0200)]
mkosi: Copy sources under /usr in the image
Originally, the source code was copied under /root/src.
This home directory is part of root FS and the new mkosi building
paradigm has only ephemeral root FS that is generated lazily.
Any files placed on the root FS in the build environment are that
excluded from the final image.
It is useful to have source codes available in the image's runtime (not
build time) environment for debugging.
ExtraTrees= as used currently are ineffective, so change the destination
to copy files under /usr to achieve the intention.
gdb sees source files as:
> 1354 ../src/src/systemctl/systemctl.c: No such file or directory.
Modify gdb configration in the built image accordingly (that file cannot
be in /root neither) to resolve to the moved sources.
(Commit fdecbf7 ("Enable unprivileged image builds") envisions bind
mounting or virtiofsd for nspawn or qemu containers respectively.)
journalcl: simplify handling of stdout being a regular file and epoll()
Let's not check the fd type beforehand, let's instead gracefully handle
if we get EPERM back from epoll_ctl() because the fd doesn't do epoll.
THis should be safer and more generic.
The epoll_ctl(2) man page clearly documents EPERM is being returned in
this case, hence it's safe to check for exactly that case.
man: let's tone down the recommendation to use Type=exec a bit
This is a follow-up for #28596.
I think the suggestion to use Type=exec uses too strong wording:
Type=exec has non-trivial drawbacks over Type=simple, and they deserve
to be mentioned.
Hence drop the <emphasis> and turn this around so that Type=exec is
*recommended*, but Type=simple is not expressly discouraged, because
there are plenty reasons to use it.
Add a brief discussion where Type=simple might be preferable.
Also, fix the outright unruth that Type=exec was the "simplest and
fastest", because it certainly is a lot, but not that.
Daan De Meyer [Thu, 10 Aug 2023 09:32:55 +0000 (11:32 +0200)]
execute: Chown credentials files and directories to service group
For a userns root user to be able to access the credentials, both
the uid and gid of the credentials directory have to be mapped into
the userns. Currently, the credentials directory group is root, which
we obviously do not want to map in to a userns, so let's make sure
that the credentials directory and files are owned by the service
group instead, which can generally be safely mapped into the userns.
Since we use permissions mode 0600, this shouldn't cause any change
in who is able to access the credentials.
udev: re-introduce symlinks for loopback block device
But the directories are changed from /dev/loop/by-ref/ -> /dev/disk/by-loop-ref/
and /dev/loop/by-inode/ -> /dev/disk/by-loop-inode/.
As /dev/loop/ is used by losetup command for other purpose.
See issue #28475.
tree-wide: drop _pure_ + _const_ from local, static functions
This is supposed to be a help for compilers to apply optimizations on
functions where they can't determine whether they are const/pure on
their own. For static, local functions the compiler can do this on its
own easily however, hence the decoration with pure/const is just noise.
Let's drop it, and let the compiler to its thing better.
(Use it for exported functions, since compilers can't 'reach-over' into
other modules to determine if they are pure, except if LTO is used)
sysusers: move global variables into a Context object
Our coding style says static variables suck except for very special
cases, i.e. things like the log level or very per-process stuff, such as
parsed version of cmdline args and such. sysusers departed from that as
one of the very few exceptions in our codebases: it keeps its
operational state in global variables.
Address that. Introduce a Context object that carries the fields that so
far have been global, and pass it around as needed.
This has the nice effect that state and configuration is clearly
separated in code, and we can very clearly see which functions mangle
state and which ones do not.
bootctl: don't advertise systemd-efi-options in --help/man anymore
The command is deprecated, as per NEWS of 254. Let's go one step further
and remove it from the help text and man page, so that people are not
inspired to use it at this time anymore.
This is how we usually have done this before: remove it from visibility,
but support it for a while still.
Yu Watanabe [Tue, 8 Aug 2023 17:40:21 +0000 (02:40 +0900)]
fstab-generator: update cache in fstab_enabled_full() based on the parsed result of fstab=
Currently, fstab-generator does not use fstab file parsers in
fstab-util.c. So, this is not necessary. Just for a possible
optimization in the case the parsers used in the future.
Yu Watanabe [Tue, 8 Aug 2023 17:30:33 +0000 (02:30 +0900)]
fstab-util: introduce fstab_enabled() helper function
And refuse to parse fstab when 'fstab=no' is specified in the kernel
command line.
When 'fstab=no' is specified in the kernel command line, fstab-generator
does not parse fstab and will not create e.g. /boot or /efi mount entry
even if fstab contains entries for the mount points. However, gpt-auto
generator may parse fstab file, and adjust or ignore mounts for EFI or
XBOOTLDR partitions based on the fstab file.
This makes gpt-auto also ignore fstab entries if 'fstab=no' is set in
the kernel command line.
The commit b42482af904ae0b94a6e4501ec595448f0ba1c06 dropped
'--exclude-prefix=/dev' from systemd-tmpfiles-setup.service. So, the
possibly later invocation of the service changes the permission set by
udevd.
As commmented in the head of this file, settings should be consistent
with udev rules. Only missing entry here is vfio. Let's re-add the
entry for the device.
The commit makes systemd-tmpfiles-setup.service also updates the
permission or owner of device nodes. However, the service does not have
ordering for systemd-udevd.service. So, the service may set different
permission from the one udevd already set.
projects / thirdparty / systemd.git / log
