[PATCH 1/3] Revert "expand-on-first-boot: Ensure that /tmp is writable"

[Uladzimir Bely]

This reverts commit 49073b9e716b2fb3203d26d9bd46b3bbd60e6016.
---
.../expand-on-first-boot/files/expand-on-first-boot.service | 1 -
1 file changed, 1 deletion(-)

diff --git a/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service b/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
index 8e76998b..90c92a39 100644
--- a/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
+++ b/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
@@ -16,7 +16,6 @@ Type=oneshot
ExecStart=/usr/share/expand-on-first-boot/expand-last-partition.sh
ExecStartPost=-/bin/systemctl disable expand-on-first-boot.service
ExecStopPost=-/bin/systemctl disable expand-on-first-boot.service
-PrivateTmp=true

[Install]
WantedBy=sysinit.target
--
2.44.2

[Uladzimir Bely]

Merging v3 of "expand-on-first-boot: Ensure that /tmp is writable"
led to systemd services ordering cycle. Systemd randomly drops one
of sycled services and sometimes it's "expand-on-first-boot" one.

This patchset reverts version 3 of the patch and returns to
previous version.

Additionally, we extend CI so that qemu run tests could always detect
any systemd services ordering cycles.

Clara Kowalsky (1):
expand-on-first-boot: Add /tmp to ConditionPathIsReadWrite

Uladzimir Bely (2):
Revert "expand-on-first-boot: Ensure that /tmp is writable"
testsuite: Detect systemd services ordering cycle

.../expand-on-first-boot/files/expand-on-first-boot.service | 3 +--
testsuite/cibuilder.py | 5 +++++
2 files changed, 6 insertions(+), 2 deletions(-)

--
2.44.2

[Uladzimir Bely]

This allows to recognize any cyclic dependencies between
systemd services.

Signed-off-by: Uladzimir Bely <ub...@ilbers.de>
---
testsuite/cibuilder.py | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/testsuite/cibuilder.py b/testsuite/cibuilder.py
index 3f77cd42..1e2c265f 100755
--- a/testsuite/cibuilder.py
+++ b/testsuite/cibuilder.py
@@ -556,6 +556,8 @@ class CIBuilder(Test):
# the printk of recipes-kernel/example-module
module_output = b'Just an example'
resize_output = None
+ # systemd service ordering cycle
+ ordering_cycle = b'Found ordering cycle'
image_fstypes, wks_file, bbdistro = CIUtils.getVars(
'IMAGE_FSTYPES', 'WKS_FILE', 'DISTRO', target=multiconfig
)
@@ -582,6 +584,9 @@ class CIBuilder(Test):
else:
rc = 2
self.log.error("No example module output while expected")
+ if ordering_cycle in data:
+ rc = 3
+ self.log.error("Systemd services ordering cycle detected")
return rc

def vm_dump_dict(self, vm):
--
2.44.2

[Uladzimir Bely]

From: Clara Kowalsky <clara.k...@siemens.com>

Since mktemp is used and creates a temporary file, we need to ensure
that /tmp is writable.

Signed-off-by: Clara Kowalsky <clara.k...@siemens.com>
---
.../expand-on-first-boot/files/expand-on-first-boot.service | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service b/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
index 90c92a39..9daae719 100644
--- a/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
+++ b/meta/recipes-support/expand-on-first-boot/files/expand-on-first-boot.service
@@ -9,7 +9,7 @@ DefaultDependencies=no
Conflicts=shutdown.target
After=systemd-remount-fs.service
Before=local-fs-pre.target shutdown.target
-ConditionPathIsReadWrite=/etc
+ConditionPathIsReadWrite=/etc /tmp

[Service]
Type=oneshot
--
2.44.2

[Uladzimir Bely]

Hello all.

If there is no objection, we would like to merge this rather soon (in
~2 days) since currently CI is broken.

--
Best regards,
Uladzimir.

[Uladzimir Bely]

On Fri, 2024-08-16 at 17:38 +0300, Uladzimir Bely wrote:

Applied to next.

--
Best regards,
Uladzimir.

[Jan Kiszka]

Still broken:

root@isar:~# journalctl -u expand-on-first-boot
-- Boot 336ca3ef86c74a06ab8e11e40a321c84 --
Jun 16 09:44:32 isar systemd[1]: expand-on-first-boot.service - Expand last partition was skipped because of an unmet condition check (ConditionPathIsReadWrite=/etc /tmp).
root@isar:~# cat /proc/cmdline
BOOT_IMAGE=/vmlinuz-6.1.0-23-amd64 root=PARTLABEL=platform rw rootwait console=ttyS0,115200 console=tty0

Patch will follow.

Why didn't you test this?

Jan

--
Siemens AG, Technology
Linux Expert Center

[Uladzimir Bely]

Hello.

The patchset was autotested in CI several times, and, according to VM
logs (for example, with bullseye/amd64 it previusly was always failing)
everything is OK:

```
[ 8.433834] systemd[1]: Starting Expand last partition...
[ 8.467179] systemd[1]: Condition check resulted in Rebuild Hardware
Database being skipped.
[ 8.481797] systemd[1]: Condition check resulted in Platform
Persistent Storage Archival being skipped.
[ 8.492728] random: systemd: uninitialized urandom read (16 bytes
read)
[ 8.520001] systemd[1]: Starting Load/Save Random Seed...
[ 8.581835] systemd[1]: Starting Apply Kernel Variables...
[ 8.655061] systemd[1]: Starting Create System Users...
[ 8.738683] systemd[1]: Mounted FUSE Control File System.
[ 8.786544] systemd[1]: Mounted Kernel Configuration File System.
[ 8.837270] systemd[1]: Finished Apply Kernel Variables.
[ 8.934155] systemd[1]: Started Journal Service.
[ 9.096109] random: crng init done
[ 9.098673] random: 46 urandom warning(s) missed due to ratelimiting
[ 9.151174] systemd-journald[195]: Received client request to flush
runtime journal.
[ 9.175233] systemd-journald[195]: Creating journal file
/var/log/journal/2cada36492d746358164b901eda86292/system.journal on a
btrfs file system, and copy-on-write is enabled. This is likely to slow
down journal access substantially, please consider turning off the
copy-on-write file attribute on the journal directory, using chattr +C.
[ 10.292661] BTRFS info (device sda2): resize device /dev/sda2 (devid
1) from 1305477120 to 1573912576
```

Now I looked to all logs and found for "bookworm/arm" the following:

```
[ 24.219166] systemd[1]: expand-on-first-boot.service - Expand last

partition was skipped because of an unmet condition check
(ConditionPathIsReadWrite=/etc /tmp).

```

We probably should to extend checks in testsuite, while currenaly only
images using "sdimage-efi-sd" and "sdimage-efi-btrfs" only are tested.

Probably, this depend on systemd version and bullseye and lower
versions are not affected.

> Jan
>

--
Best regards,
Uladzimir.