In order to mount procfs and sysfs in an unprivileged container the
kernel requires that a fully visible instance is already present in the
target mount namespace. Mount one here so the inner child can mount its
own instances. Later we umount the temporary instances created here
before we actually exec the payload. Since the rootfs is shared the
umount will propagate into the container. Note, the inner child wouldn't
be able to unmount the instances on its own since it doesn't own the
originating mount namespace. IOW, the outer child needs to do this.
So far nspawn didn't run into this issue because it used MS_MOVE which
meant that the shadow mount tree pinned a procfs and sysfs instance
which the kernel would find. The shadow mount tree is gone with proper
pivot_root() semantics.
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>
In order to support pivot_root() we need to move mount propagation
changes after the pivot_root(). While MS_MOVE requires the source mount
to not be a shared mount pivot_root() also requires the target mount to
not be a shared mount. This guarantees that pivot_root() doesn't leak
any mounts.
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>
Currently, services use mount_move_root() in order to setup the root
directory of services using a mount namespace. This relies on MS_MOVE
and chroot(). However, this has serious drawbacks even for relatively
simple mount propagation scenarios.
What systemd currently does is roughly equivalent to the following shell
code:
unshare --mount --propagation=shared
cd /
mount --make-rslave /
mkdir /new-root
mount --rbind / /new-root
cd /new-root
mount --move /new-root /
chroot .
This looks simple enough but has the consequence that two separate mount
trees exist for the lifetime of the service. The first one was created
when the mount namespace was created, and the second one when a new
mount for the rootfs was created. The first mount tree sticks around as
a shadow mount tree. Both mount trees are dependent mounts with the host
rootfs as their dominating mount.
Now, when mount propagation is triggered by the host by e.g.,
mount --bind /opt /mnt
it means that two propagation events are generated. I'm skipping over
the exact kernel details as they aren't that important. The gist is that
for every propagation event that is generated a second one is generated
for the shadow mount tree. In other words, the kernel creates two copies
for each mount that is propagated instead of one.
This isn't necessary. We can simply change the sequence above to:
unshare --mount --propagation=shared
cd /
mount --make-rslave /
mkdir /new-root
# stash fd to old rootfs
# stash fd to new rootfs
mount --rbind / /new-root
mkdir /new-root
cd /new-root
pivot_root . .
# new root is tucked under old root
# chdir into old rootfs via stashed fd
umount -l /old-root
The pivot_root allows us to get rid of the old mount tree that was
created when the mount namespace was created. So after this sequence
only one mount tree is alive. Plus, it's safer and nicer. Moving mounts
isn't pleasnt.
This patch doesn't convert nspawn yet as the requirements are more
tricky given that it wants to preserve the rootfs as a shared mount
which goes against pivot_root() requirements.
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>
When attaching and /etc/systemd/system.attached can't be created or used
(eg: dead symlink) the logs are pretty much useless as even at debug
level there's no indication of what is going wrong.
Add some debug logs, and return a more specific error string over D-Bus.
I was looking at a bug in bugzilla about some boot loader issue, and it was
hard to say if the boot entry files were generated by our plugin or something
else. Add a header to make this clear.
kernel-install invokes the plugins via absolute path always, so $0 gives as
the full path the location where the plugin is installed. This is what we want:
title Fedora Linux 37 (Workstation Edition)
# Boot Loader Specification type#1 entry
# File created by /usr/lib/kernel/install.d/90-loaderentry.install (systemd 252-409-g5028904^)
When relaxed checks are requested, let's not require the efi/xbootldr
directory to be the root of the filesystem. When building images, image
builders might install all efi/xbootldr files to a regular directory
first before packing them up into a partition. To allow bootctl to be
used in such scenarios to install systemd-boot, we need to relax the
fsroot check.
It's only used to avoid BLKDISCARD on individual partitions at the moment.
It can take a lot of time to run on very slow devices, so avoid it for
them too.
sd-stub has an opportunity to handle the seed the same way sd-boot does,
which would have benefits for UKIs when sd-boot is not in use. This
commit wires that up.
It refactors the XBOOTLDR partition discovery to also find the ESP
partition, so that it access the random seed there.
This reenables epoll_pwait2() use, i.e. undoes the effect of
39f756d3ae.
Instead of just reverting that, this PR will change things so that we
strictly rely on glibc's new epoll_pwait2() wrapper (which was added
earlier this year), and drop our own manual fallback syscall wrapper.
That should nicely side-step any issues with correct syscall wrapping
definitions (which on some arch seem not to be easy, given the sigset_t
size final argument), by making this a glibc problem, not ours.
Given that the only benefit this delivers are time-outs more granular
than msec, it shouldn't really matter that we'll miss out on support
for this on systems with older glibcs.
We said "query the journal". This is true but also very generic. Let's say
"print log entries from the journal" instead, so that users who are looking for
"logging" are more likely to figure out that the journalctl is the tool for
them.
Also, mention systemd-journal-remote.service which can write the journal too.
And give some hints how to figure out how to write *to* the journal.
In sd_bus_wait(), let's convert EINTR to a return code of 0, thus asking
the caller do loop again and enter sd_bus_process() again (which will
not find any queued events). This way we'll not return an error on
something that isn't really an error. This should typically make sure
things are properly handled by the caller, magically, without eating up
the event entirely, and still giving the caller time to run some code if
they want.
Sometimes, RTM_NEWLINK message with carrier is received earlier than
NL80211_CMD_CONNECT. To make SSID= or other WiFi related settings in
[Match] section work, let's try to reconfigure the interface.
Fixes a bug introduced by 96f5f9ef9a.
Fixes#25384.
If no root= switch is specified on the kernel command line we'll use the
root disk on which the partition the LoaderDevicePartUUID efi var is
located – as long as that partition is an ESP. Let's slightly liberalize
that and also allow it if that partition is an XBOOTLDR partition. This
ensures that UKIs spawned directly from XBOOTLDR work the same as those
from the ESP.
(Note that this makes no difference if sd-boot is in the mix, as in that
case LoaderDevicePartUUID is always set to the ESP, as that's where
sd-boot is located, and sd-boot will set the var first, sd-stub will
only set it later if it#s not set yet.)
The compiler should recognize that these are constant expressions, but
let's better make this explicit, so that the linker can safely share the
initializations all over the place.
Now that the random seed is used on virtualized systems, there's no
point in having a random-seed-mode toggle switch. Let's just always
require it now, with the existing logic already being there to allow not
having it if EFI itself has an RNG. In other words, the logic for this
can now be automatic.
The second argument to _printf_() specifies where the arguments start. We need to
use 0 in two cases: when the args in a va_list and can't be checked, and with journald
logging functions which accept multiple format strings with multiple argument sets,
which the _printf_ checker does not understand. But strv_extendf() can be checked.
Make sure that the sym_xyz function pointers have the types that the
functions we'll assign them have.
And of course, this found a number of incompatibilities right-away, in
particular in the bpf hookup.
(Doing this will trigger deprecation warnings from libbpf. I simply
turned them off locally now, since we are well aware of what we are
doing in that regard.)
There's one return type fix (bool → int), that actually matters I think,
as it might have created an incompatibility on some archs.
Removing the virtualization check might not be the worst thing in the
world, and would potentially get many, many more systems properly seeded
rather than not seeded. There are a few reasons to consider this:
- In most QEMU setups and most guides on how to setup QEMU, a separate
pflash file is used for nvram variables, and this generally isn't
copied around.
- We're now hashing in a timestamp, which should provide some level of
differentiation, given that EFI_TIME has a nanoseconds field.
- The kernel itself will additionally hash in: a high resolution time
stamp, a cycle counter, RDRAND output, the VMGENID uniquely
identifying the virtual machine, any other seeds from the hypervisor
(like from FDT or setup_data).
- During early boot, the RNG is reseeded quite frequently to account for
the importance of early differentiation.
So maybe the mitigating factors make the actual feared problem
significantly less likely and therefore the pros of having file-based
seeding might outweigh the cons of weird misconfigured setups having a
hypothetical problem on first boot.
Christian Brauner
Yu Watanabe
Michal Sekletar
Luca Boccassi
Daan De Meyer
Bernd Steinhauser
Lennart Poettering
Zbigniew Jędrzejewski-Szmek
Jason A. Donenfeld
Anita Zhang
Benjamin Fogle
наб