Wire up the runtime hot-swap Varlink method. The signature mirrors
AddStorage minus 'config': the device frontend (virtio-blk,
virtio-scsi, nvme, scsi-cd) doesn't change, only the backing file
behind it. Read-only/read-write may flip based on the new fd's
O_ACCMODE; scsi-cd is forced read-only to match the boot-time policy.
QMP sequence (entry: vmspawn_qmp_replace_block_device):
add-fd → on_replace_observe_stage
blockdev-add (new file) → on_replace_blockdev_add_complete
remove-fd (new fdset) → on_replace_observe_stage
blockdev-reopen (format) → on_replace_blockdev_reopen_complete
[commit + fire trailing del]
blockdev-del (old file) → on_replace_old_blockdev_del_complete
The reopen options must be a superset of every option that
qmp_build_blockdev_add_format() may emit, otherwise reopen rejects
'Cannot reset option X to default'. The 'file' field is a string
reference to the new file node — case 3 of the schema in
qemu/qapi/block-core.json:5034-5040 ("the current child is replaced
with that other node"). The format node's qmp_node_name is preserved
so the device frontend's drive=<X> binding does not move.
ReplaceCtx tracks the per-call state with a refcount mirroring the
add-stage drive-info pattern. On any pre-commit failure replace_fail
tears down whatever new-side state we created on the wire and replies
on drive->link via reply_qmp_error (disconnect → NotConnected). On
post-commit del failure we log a warning, leak the orphan, and reply
success — the swap itself succeeded and the leak resolves at VM exit.
file_generation is bumped before issuing blockdev-add so failed
attempts cannot collide on node-name when the user retries.
Errors:
NoSuchStorage - drive not in the registry
StorageImmutable - drive lacks QMP_DRIVE_REMOVABLE (boot-time)
EBUSY - add still pending or another replace/remove in flight
NotConnected - QMP transport disconnect during the chain
EIO - QEMU rejected blockdev-reopen
Also gates RemoveStorage on REPLACE_PENDING so a device_del cannot
race a mid-flight blockdev-reopen on the same drive.
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
Define the IDL for io.systemd.MachineInstance.ReplaceStorage, a
runtime hot-swap of an already-attached storage volume's backing
file. The signature mirrors AddStorage minus the 'config' field
because the device frontend (virtio-blk, virtio-scsi, nvme, scsi-cd)
does not change — only the backing file behind it.
The implementation lives in vmspawn (next commit) and uses QMP
blockdev-reopen to swap the file child of the existing format node.
The reused error vocabulary (NoSuchStorage, StorageImmutable,
NotConnected, plus the generic errno path) covers every failure
mode; no new errors are added.
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
The current vmspawn_qmp_add_block_device() emits a single blockdev-add
that combines the format-level node ("vmspawn-N-storage") with an
inline file child. QEMU's qmp_blockdev_add() only marks the top-level
returned BDS as monitor-owned (qemu/blockdev.c:3440); inline children
are NOT, so qmp_blockdev_del() rejects them with "Node X is not owned
by the monitor" (qemu/blockdev.c:3513-3517).
To prepare for ReplaceStorage — which needs to swap the file child of
an existing format node via blockdev-reopen, and then blockdev-del the
old file node — make the file node monitor-owned by issuing it as its
own blockdev-add call. The 4-stage add chain becomes 5 stages:
add-fd
blockdev-add (file) → on_add_file_node_stage sets FILE_NODE_ADDED
blockdev-add (format) → on_add_format_node_stage sets BLOCKDEV_ADDED
remove-fd
device_add
DriveInfo gains qmp_file_node_name ("vmspawn-N-file-G", G a generation
counter bumped on every replace), file_generation, and a stashed
fdset_id so future ReplaceStorage can target both for cleanup.
vmspawn_qmp_block_device_teardown() now deletes both nodes in order —
format first, then file — because the format holds a strong reference
to its file child and a file-first del is rejected with "Node X is
busy: node is used as 'file' of Y".
Folds bridge->features VMSPAWN_QMP_FEATURE_IO_URING into the file
node's flags so the new path inherits io_uring just like the old
inline form did. The format-level options (read-only, discard,
discard-no-unref) are unchanged.
The ephemeral path is structurally already separate file+format with
monitor-owned children; no behavioural change there beyond the
on_add_blockdev_stage → on_add_format_node_stage rename.
Drops the now-unused qmp_build_blockdev_add_inline() helper.
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
$PATH might point to a systemd checkout containing
a newer version of systemd-journal-remote which we
should use, hence prefer an executable from $PATH
over the one from /usr/lib/systemd.
Even after switching the wait loop to a polling `journalctl --grep`, the
test still fails intermittently because the very first messages emitted by
the freshly-spawned systemd-networkd-wait-online process can carry stale
journald metadata. journald associates `_SYSTEMD_UNIT=` (and friends) with
each entry by reading `/proc/$pid/cgroup` of the originating PID; if those
messages are produced before journald notices the cgroup migration into the
new service, they get tagged with `_SYSTEMD_UNIT=init.scope`. The
`-u $unit` filter then fails to match them.
Capture a journal cursor before launching the unit, and grep using
`--after-cursor=` plus `SYSLOG_IDENTIFIER=systemd-networkd-wait-online`
instead of `-u $unit`. SYSLOG_IDENTIFIER is set by the program itself, so
it's not subject to the cgroup-discovery race. The cursor bounds the search
to entries produced by this invocation, so prior wait-online runs in
earlier testcases don't interfere.
Logs from the failing run showing the messages exist but are tagged with
the wrong unit:
[ 2570.948554] TEST-75-RESOLVED.sh[2178]: + unit=wait-online-dns-ede81407-b93b-459d-8e5d-69292b42d2ae.service
[ 2571.023162] TEST-75-RESOLVED.sh[2178]: + systemd-run -u wait-online-dns-ede81407-b93b-459d-8e5d-69292b42d2ae.service ...
[ 2571.049189] TEST-75-RESOLVED.sh[2178]: + timeout 30 bash -c 'until journalctl -b -u wait-online-dns-ede81407-b93b-459d-8e5d-69292b42d2ae.service --grep ...'
[ 2571.964986] systemd-networkd-wait-online[2190]: dns0: No DNS server is accessible.
[ 2601.051088] TEST-75-RESOLVED.sh[2178]: ++ cleanup
And for that 2571.964986 entry:
_SYSTEMD_CGROUP=/init.scope
_SYSTEMD_UNIT=init.scope
_EXE=/usr/lib/systemd/systemd-executor
_CMDLINE=/usr/lib/systemd/systemd-executor --deserialize 68 ...
SYSLOG_IDENTIFIER=systemd-networkd-wait-online
MESSAGE=dns0: No DNS server is accessible.
Follow-up for d4bc62713e
Co-developed-by: Claude Opus 4.7 <noreply@anthropic.com>
When an LSM (e.g. SELinux) denies an fd transfer or the receiver hits
RLIMIT_NOFILE, the kernel drops the fd(s) from the SCM_RIGHTS cmsg and
sets MSG_CTRUNC on the recvmsg(). recvmsg_safe() turns that into
-ECHRNG, which causes json_stream_read() to discard the data bytes
that were nevertheless received and the varlink server to silently
tear down the connection — leaving the caller waiting for a reply
that never comes.
Inline the recvmsg() call instead and, on MSG_CTRUNC, drop the partial
fds but keep the message data. The method handler will surface a clean
-ENXIO when it tries to peek the missing fd, which sd-varlink wraps as
io.systemd.System for the peer, instead of a hang. This matches the
recent sd-bus fix in 6c8de404c9 ('sd-bus: allow receiving messages with
MSG_CTRUNC set').
If a unit has active jobs, when it gets serialized there are job
subsections, each with their own empty line marker. The skipping
function ignores this and skips until the marker, but then leaves
the job in place, breaking deserialization.
Consume jobs subsections too.
This shows up now that there's TEST-07-PID1.alias-corruption,
which occasionally fails when the aliased unit happens to
still have a job when the reexec happens.
```
[ 967.551630] TEST-07-PID1.sh[179]: + echo 'Testing with: systemctl daemon-reexec'
[ 967.551630] TEST-07-PID1.sh[179]: Testing with: systemctl daemon-reexec
[ 968.405274] TEST-07-PID1.sh[179]: + echo '--- Attempt 1/3 ---'
[ 968.405274] TEST-07-PID1.sh[179]: --- Attempt 1/3 ---
[ 968.698641] TEST-07-PID1.sh[179]: + echo 'Running daemon-reexec...'
[ 968.698641] TEST-07-PID1.sh[179]: Running daemon-reexec...
[ 969.130261] TEST-07-PID1.sh[179]: + echo 'legit.service PID remains 1282. Attempt 1 passed.'
[ 969.130261] TEST-07-PID1.sh[179]: legit.service PID remains 1282. Attempt 1 passed.
[ 970.870456] TEST-07-PID1.sh[179]: + echo '--- Attempt 2/3 ---'
[ 970.870456] TEST-07-PID1.sh[179]: --- Attempt 2/3 ---
[ 971.267205] TEST-07-PID1.sh[179]: + echo 'Running daemon-reexec...'
[ 971.267205] TEST-07-PID1.sh[179]: Running daemon-reexec...
[ 971.715743] TEST-07-PID1.sh[179]: + echo 'legit.service PID changed from 1282 to 1643!'
[ 971.715743] TEST-07-PID1.sh[179]: legit.service PID changed from 1282 to 1643!
```
https://github.com/systemd/systemd/actions/runs/25376867873/job/74414201255
Without CAP_SYS_ADMIN, btrfs_ioctl_snap_destroy() runs an
inode_permission(MAY_WRITE) check against the target subvolume root, which
btrfs_permission() rejects with EROFS for a read-only subvolume. As a
result, unprivileged removal of a read-only subvolume fails — both via
btrfs_subvol_remove_at() directly and via the recursive cleanup path used
by rm_rf_subvolume_and_freep(), which propagates the EROFS up.
Detect EROFS after the destroy ioctl, clear the RDONLY flag (only inode
ownership is required for BTRFS_IOC_SUBVOL_SETFLAGS), and retry once.
While at it, fix the surrounding comments: BTRFS_IOC_SNAP_DESTROY drops the
entire subvolume tree, so regular files inside are irrelevant; ENOTEMPTY
from the ioctl indicates nested subvolumes (BTRFS_ROOT_REF_KEY entries) via
may_destroy_subvol(), not non-empty contents.
The help string for --smart-relinquish-var and --relinquish-var
were in reversed order because of the _fallthrough_.
We would resolve the conditions for "smart relinquish" immediately
in parse_argv() and call 'return 0' if the conditions were wrong,
terminating option parsing and the program. It seems nicer to delay
action until later. This makes the logic flow more standard. This
also allows the option parsing cases to be exchanged, fixing the
issue with --help.
Two namespaces are used: "journalctl" and "journalctl-varlink". Help for
--user/--system in the latter is added, even though it is not used yet.
I think it'll be good to have this for introspection.
The four FSS-related options (--interval, --verify-key, --force,
--setup-keys) unfortunately each gain an inline #if HAVE_GCRYPT / #else;
the EOPNOTSUPP fallback is duplicated four times.
The metavar for --identifier/--exclude-identifier is changed to "ID"
to make the layout nicer. (And because that seems to make more sense.)
Co-developed-by: Claude Opus 4.7 <noreply@anthropic.com>
We have different help strings for --user/--system in different places, so this
only covers a subset of --system/--user instances. But this particular help
seems to be the most widely used.
(In a few cases, the help string is fixed: it should be "system mode", not
"per-system mode".)
Pure reordering. ARG_SMART_RELINQUISH_VAR is kept immediately before
ARG_RELINQUISH_VAR because of the existing _fallthrough_; that's the
only deviation from strict --help order.
Co-developed-by: Claude Opus 4.7 <noreply@anthropic.com>
Note, previously we replaced the previous option with the same option code with
new one. But, DHCP message can have multiple options with same option code.
Hence, this make the conf parser not replace, but append new one.
There are four DHCP options with confusing names:
Option 43: Vendor-Specific Information
Option 60: Vendor Class Identifier
Option 124: Vendor-Identifying Vendor Class
Option 125: Vendor-Identifying Vendor-Specific Information
Let's use their full names for their corresponding enums.
BTRFS_IOC_SEARCH is only available to root in the
initial userns. This means we fail to recursively
snapshot even if a subvolume has no nested subvolumes
at the moment.
Let's fix this by using the newer btrfs ioctls which
do work even if we don't have CAP_SYS_ADMIN in the initial
userns.
Addition to PR https://github.com/systemd/systemd/pull/41181
Plasma-workspace OSD notifications about turning the touchpad on
and off are guided by f21. When this match is specified,
KDE notifies on this laptop that the on/off switch of the atchpad
state is pressed.
Fix dmesg:
atkbd serio0: Unknown key pressed (translated set 2, code 0xc1 on isa0060/serio0).
SUSE uses a different preset, so don't just assert in the test,
instead just start the socket in case it is not enabled
TEST-74-AUX-UTILS.sh[1594]: ++ systemctl is-enabled systemd-report-basic.socket
TEST-74-AUX-UTILS.sh[1540]: + [[ disabled == enabled ]]
TEST-74-AUX-UTILS.sh[120]: + echo 'Subtest /usr/lib/systemd/tests/testdata/units/TEST-74-AUX-UTILS.report.sh failed'
Follow-up for 4409e52494
Christian Brauner
Daan De Meyer
Yu Watanabe
Luca Boccassi
Lennart Poettering
Bone NI
Zbigniew Jędrzejewski-Szmek
drhydroxide
hschloss
Artem Proskurnev