Rather than setting up and tearing down two identical namespaces,
create it once and use it for the duration of the test. (This is
probably a documentation issue with Infamy: there is no requirement
that test.steps are direct descendants of the test, but until this
point there was no example of this in the source base)
In the process of this, it was also discovered that no positive ping
test was done over v6, so that has been added.
Avoid having log messages stuck in a pipe when stdout is not a tty,
e.g. when running a suite of tests using 9pm.
Two benefits:
- Nicer experience when interactively running a suite
- Timestamps inserted by 9pm are more useful
It turns out that the semantics of ping's -c option does not match our
expectations. The manual says:
> Stop after sending count ECHO_REQUEST packets. With deadline option,
> ping waits for count ECHO_REPLY packets, until the timeout expires.
But in fact, something like an ICMP_DEST_UNREACH will also count as a
response, meaning that the process will terminate with non-zero
exitcode even though the deadline has not yet been crossed, rather
than keep waiting for an ECHO_REPLY.
To add some extra flavor: this is only the case for iputils' ping,
Busybox's implementation will hold out for a response until the
deadline expires.
Therefore, wrap the ping in an ugly retry loop.
Also, return the full available context from must{,_not}_reach on
failures, instead of just the output. This allows us to show the exit
code in the top-level exception handler.
When a test starts an isolated MACVLAN namespace, it is often useful
to interactively join that namespace to investigate issues. To do
this, we need to find the "sleeper" process backing the namespace, and
`nsenter` into that process's network namespace.
Add a script that automates this process.
To start a shell in the netns of the MACVLAN stacked on top of d1b,
start a shell in the test environment:
infix/test $ ./shell
13:37:00 infamy0:test # ./nsenter d1b
13:37:01 infamy0(netns:d1b):test #
It is not uncommon for subprocesses to fail when developing new tests,
but the default representation of CalledProcessError does not print
the stdout of the program, only that it failed.
Catch this common case when a test is about to fail, and make sure to
log any output it produced.
When a CLI user performs the following command, the DNS resolver was
left in an undefined state.
admin@infix-00-00-00:/config/> no system
admin@infix-00-00-00:/config/> leave
admin@infix:/>
The avahi-daemon reported the missing /etc/resolv.conf and the fix is to
ensure `resolvconf -u` runs on every major change ot ietf-system, even
if there is no new resolv.conf to roll in. At bootstrap this has been
handled by another mechanism so has not been seen by most users.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
The sr_get_data() API may return SR_ERR_OK and still set the cfg pointer
to NULL! This happens when, e.g., executing 'no system' from the CLI.
This patch adds a check for this in all occurrences of sr_get_data().
Also, in ietf-routing.c there was *no* error handling for this API, this
required some restructuring to ensure a FILE *fp was closed properly in
all cases.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
With the sysrepo patch from the previous commit, we can now properly
detect if a callback failed to apply its changes in SR_EV_DONE. When
this occurs the system may be in an undefined state, so we must try
to recover it before loading failure-config.
This patch tries to perform a factory-default RPC, which is an Infix
specifc RPC that does "copy factory-config running-config". We give
sysrepocfg some time to clean up any stale SHM connections before we
do a hard scratch of the db state and restart sysrepo-plugind.
Fixes#429
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This local patch of sysrepo adds support for detecting non-zero return
code from sysrepo callbacks during SR_EV_DONE, which upstream sysrepo
currently, as of v2.2.105, discards.
With this change failure to load startup-config results in the system
detecting the failure applying failure-config instead.
Fixes#429
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Fix regression introduced in 3f87945, which tries to map a user to an
existing home directory. The root cause for the problem is basically an
invalid check of the /home/admin UID being in use by any system account
or not. This patch uses the same check used by the BusyBox adduser
command.
In addition to the uid-already-in-use fix, several other checks have
been added to ensure we do not end up with an invalid system state:
- If reusing an existing /home/admin fails in the 'adduser admin'
command, wipe out /home/admin and retry adding the user cleanly
- Always delete any group with the same name before trying to add
the user, regardless if /home/$user exists or not, this prevents
issues with stale group records if we end up failing to load the
startup-config and need to load failure-config
- For the same reason (failure-config), make sure to always clean
up any stale user in the retry step. The same step is used when
there is a uid:gid mismapping (the original bug)
Fixes#428
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of having a chown process started by Finit in runlevel S, we can
just chown in the confd load script instead.
Issue #415
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Error messages from system() calls, like adduser, should be logged with
severity LOG_ERR to not arrive out of sequence with ERROR() messages.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Instead of hard-coding the fallback hostname we can now use the one
generated to /etc/os-release (for netbrowse).
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Since we don't have any VPD defined on this board, supply a default
password hash via the device tree instead.
Also, specify the path to external USB port, so that Infix will pick
it up in ietf-hardware.
Provide a patched device tree that:
- Uses the correct phy-mode for eth2, allowing it to be probed
properly
- Wires up the interrupt signal to the GPIO expander, so that the SFP
cage's I/O lines can be configured as interrupts, rather than having
to poll them
Assume that the caller always want to create a new partition table,
even in scenarios where one already exists. This allows devices to be
reprovisioned with a newer image, for example.
Fixes#394
This is not a production target, so we choose the more developer
friendly option of allowing the boot process to be interrupted.
Also, remove the old update cruft and rely on Marvell's `bubt`
instead.
The BootROM wants the bootloader to be located on LBA1, which is not
ideal when using a GPT to describe the partition table.
Make sure the `bubt` command targets SPI by default.
Buildroot's default ATF version changed from 2.9 to 2.10 when we
upgraded to 24.02.x. This meant that our patches were not applied,
which broke the build.
Take a conservative approach for now, and just pin the ATF version to
the known-good 2.9 series.
During the addition of multi-key support to RAUC and U-Boot, the
bootloader defconfig for CN9130 CRB was overlooked, which meant that
the Infix specific U-Boot hooks never ran, thereby ultimately breaking
the build.
Restore the required options to build in the dev-key by default.
The Buildroot 2024.02.x-kkit branch recently had uboot-tools upgraded
and the patches needed require bison to rebuild.
The recommendation is to also install flex, they always go together.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
