Harald Hoyer [Fri, 12 Feb 2021 12:53:34 +0000 (13:53 +0100)]
ci: more disk space for test 30 iscsi
```
mke2fs 1.45.6 (20-Mar-2020)
Filesystem too small for a journal
Discarding device blocks: done
Creating filesystem with 1024 1k blocks and 128 inodes
Allocating group tables: 0/1 done
Writing inode tables: 0/1 done
Writing superblocks and filesystem accounting information: 0/1 done
cp: error writing '/sysroot/usr/bin/bash': No space left on device
cp: error writing '/sysroot/usr/bin/grep': No space left on device
cp: error writing '/sysroot/usr/bin/ping': No space left on device
[…]
```
This rephrases a comment in the check() as requested on another PR
and generalizes the install() section comments on par with other recent
module changes.
Introducing systemd-ask-password module which is used to query a
system password or passphrase from the user.
This module can be used standalone ( systemd-ask-password )
( which is without plymouth & wall support )
With plymouth module ( systemd-ask-password + plymouth )
With wall module ( systemd-ask-password + wall ), which is currently
commented out since there is no forceable usecase for it or even exiting
wall module.
This is a standalone systemd module which means it depends on no other module than systemd.
systemd-run runs nbd-client in parallel. This introduces a possible
race condition when /dev/nbd0 is not directly used by /sysroot mount
but used by another dracut module as in:
netroot=nbd:ip:port:fstype root=live:/dev/nbd0
Before nbd0 usages are synchronized, it is better to keep it serial.
Introducing systemd-modules-load which is an early boot service
that loads kernel modules from a static list, which is required for
kernel modules that do for example not support automatic module loading
( like key type parsers ).
Petr Pavlu [Tue, 9 Feb 2021 12:13:25 +0000 (13:13 +0100)]
fix(shutdown): add timeout to umount calls
When terminating a system, the shutdown module attempts to unmount all
file systems from under /oldroot. This reaps remaining file systems that
systemd cannot unmount and detaches /oldroot itself.
In case that running umount for some file system repeatedly fails, the
module reports this error and continues the processing in order to
shutdown the system. This handles a condition when the umount command
actually terminates but it can happen in some cases that it waits
indefinitely.
An example with NFS mounts:
# mount -t nfs 192.168.0.1:/srv/nfs/dir /mnt/nfs
# mkdir /mnt/nfs/dir2
# mount -t nfs 192.168.0.1:/srv/nfs/dir2 /mnt/nfs/dir2
# touch /mnt/nfs/dir2/file
# systemd-run -pKillMode=none -pSendSIGKILL=no tail -f /mnt/nfs/dir2/file
Running as unit: run-r367825c967ca4d88a793ae4793c02f8b.service
# systemctl poweroff
The invoked tail command escapes normal termination by systemd and
prevents stopping mnt-nfs.mount and mnt-nfs-dir2.mount as it makes the
mounts busy. Systemd then again attempts to unmount these file systems
in systemd-shutdown but this fails as well. The utility tries to unmount
/mnt/nfs/dir2 but the kernel waits indefinitely doing a path lookup for
/mnt/nfs because network is no longer available at that point. The
systemd-shutdown gives up after 90 seconds. Finally, the control is
transferred to dracut which tries to unmount the file systems in the
same way and ends up indefinitely waiting on umount to finish.
This situation causes that the system hangs during shutdown. The patch
improves the shutdown module to add a timeout of 90 seconds for the
umount operation and continue with the shutdown if it gets reached,
similarly to what systemd-shutdown does.
fix(kernel-modules): add reset controllers for arm
Reset controllers might be needed by some of the devices used in the
initrd. Particularly on the Raspberry Pi 4, 'xhci-pci' depends on a
platform specific reset controller.
Signed-off-by: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
If plymouth is used then its own installation script is preferred.
While plymouth's own variant of inst_binary resolves the library
dependencies properly, its inst_library shell function doesn't
actually install them in some cases properly.
Make sure to install both the required binaries and their
dependencies. In the worst case, it's a NOP.
fix(network-manager): allow override network manager version
With dracutsysrootdir set and foreign binaries in sysroot,
running NetworkManager --version is not possible and it may be
different than the one installed on the host.
NM_VERSION can be used to override it.
feat(dracut.sh): allow overriding the systemctl command for sysroot
Allow overriding the systemctl command for sysroot with $SYSTEMCTL
Modified every modules' module-setup.sh to use the envvar instead
of the hardcoded command name.
foopub [Wed, 3 Feb 2021 17:41:08 +0000 (17:41 +0000)]
Fix bad ls parsing
This used to cause some echo write/invalid argument errors. Simplest fix is to use readlink with -f for absolute path and -n to remove trailing newline.
See previous discussion here:
https://github.com/void-linux/void-packages/issues/13024
Alexander Tsoy [Thu, 17 Dec 2020 21:01:32 +0000 (00:01 +0300)]
fix: move ldconfig after library workaround
Commit de3cb0e3214c moved libpthread workaround after the code that runs
ldconfig. As a result, if installed in a non-default path, libgcc_s.so.1
could not be found by dynamic linker.
This modules introduces systemd-repart which can be used
to generate partition(s), via a simple systemd-repart
drop-in(s), which can be useful in wide varity of usecase.
feat(systemd-repart): adding module to spec file
feat(systemd-repart): removing unessary requirement on type unit
Harald Hoyer [Fri, 5 Feb 2021 15:25:14 +0000 (16:25 +0100)]
ci: get rid of "mount -o loop"
This finally allows running the test suite completely in a rootless container:
```
❯ podman run \
--user 0 \
-v /dev:/dev \
-v ./:/dracut \
-it \
quay.io/haraldh/dracut-fedora:33 \
bash -c 'cd /dracut; make DRACUT_NO_XATTR=1 check'
```
Harald Hoyer [Mon, 8 Feb 2021 12:27:58 +0000 (13:27 +0100)]
docs: remove TODO file
TODO file copied to https://github.com/dracutdevs/dracut/issues/1068
and also filed as individual issues on
https://github.com/dracutdevs/dracut/issues
Kairui Song [Sun, 7 Feb 2021 09:20:16 +0000 (17:20 +0800)]
95nfs: fix rpc.statd installation
commit 5840c466d tried to make it create the right directory according
to the directory layout on current host, but added "$initdir" prefix
to the if statement by accident, so the installation code have no
effect now. Just fix it.
Matthew Thode [Fri, 29 Jan 2021 04:40:59 +0000 (22:40 -0600)]
fix: do not set cmdline for uefi images unless asked
The kernel does not allow the cmdline to be overridden if one is defined
in the uefi image. This means that bootloaders like systemd-boot or
grub cannot manage boot options (editing at boot for debug being a use
case). This patchset allows for no cmdline to be set (instead of force
setting one if none is provided).
Fixes: https://github.com/dracutdevs/dracut/issues/1012 Signed-off-by: Matthew Thode <mthode@mthode.org>
Anjali Kulkarni [Thu, 4 Feb 2021 01:15:15 +0000 (17:15 -0800)]
feat(network-legacy): send dhcp in parallel on all devices
We can drastically improve boot times on some machines, or BM or VM
shapes, if we send DHCP requests in parallel on all interfaces
Add ip=single-dhcp as a kernel boot parameter, to send DHCP requests in
parallel on all interfaces. Thus, DHCP on each interface is overlapped
thereby reducing boot times for multiple interface machines.
This can help reduce the boot time by 1 minute per interface.
In addition, we may also save another 40 secs delay per interface, if
the link on which we try to send DHCP is in DOWN state. This can be
reduced by giving the following kernel command line parameter while
booting - rd.net.timeout.carrier=<timeout>.
Added manpage for ip=single-dhcp with the caveat that it does not
apply to Network Manager or SUSE wicked.
Mariusz Tkaczyk [Thu, 26 Nov 2020 11:51:47 +0000 (12:51 +0100)]
fix(mdraid): add grow continue service
During raid assembly mdadm will call mdadm-grow-continue.service if
reshape is in progress.
If service doesn't exist then it is done via fork and systemd will kill
it unconditionally after processing all events.
Add proper service to allow systemd to manage reshape correctly.
fix(kernel-network-modules): also install modules from mdio subdirectory
A recent change in Linux kernel v5.10 split the MDIO drivers from the PHY
subdirectory into a new MDIO subdirectory [0]. This caused MDIO drivers
to not be included in the initramfs.
Also include the modules of this new subdirectory to prevent this issue.
Ondrej Mosnacek [Fri, 4 Dec 2020 16:21:05 +0000 (17:21 +0100)]
fix(90kernel-modules): install generic crypto modules with hostonly unset
Otherwise e.g. the xts(aes) implementation provided by the vmx_crypto
module (which does usually get included on ppc64le) fails to initialize
when xts is built as a module (CONFIG_CRYPTO_XTS=m), because it can't
instantiate the fallback generic xts(aes) implementation (needs the
generic xts module).
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>