doc: expand on static vs dynamic factory-config

Infix now supports static /etc/factory-config.cfg from a rootfs overlay
or /usr/share/product/<VPD.PRODUCT>/etc/factory-config.cfg in addition
to the autogenerated /run/confd/*-config.gen files.

Update the Branding & Releases document with a bit more information on
this rather complex topic.

[skip ci]

Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
This commit is contained in:
Joachim Wiberg
2024-10-01 11:43:45 +02:00
parent 808a04f125
commit d37656e347
2 changed files with 209 additions and 16 deletions
+192 -6
View File
@@ -41,13 +41,92 @@ Verify the result after a build by inspecting:
printed on a label on the device.
Factory Defaults
----------------
Factory & Failure Config
------------------------
The Infix default configuration, factory-config, is part static files
and part per-device generated files, e.g., SSH hostkey and hostname.
The latter is constructed from the file `/etc/hostname`, appended with
the last three octets of the system's base MAC address. To override the
To support booting the same image (CPU architecture) on multiple boards,
Infix by default generates the device's initial configuration every time
at boot. This also ensures the device can always be restored to a known
state after a factory reset, since the `factory-config` is guaranteed to
be compatible with the YANG models for the given software version. (For
more information on how the system boots, please see the section [Key
Concepts](introduction.md#key-concepts) in the Introduction document.)
However, for custom builds of Infix it is possible to override this with
a single static `/etc/factory-config.cfg` (and failure-config) in your
rootfs overlay -- with a [VPD](vpd.md) you can even support several!
### Variables & Format Specifiers
Parts of the configuration you likely always want to generated, like the
SSH hostkey used by NETCONF, a unique hostname, or the `admin` user's
unique (per-device with a VPD) password hash. This section lists the
available keywords, see the next section for examples of how to use
them:
- **Default password hash:** `$factory$` (from VPD, .dtb, or built-in)
XPath: `/ietf-system:system/authentication/user/password`
- **Default NETCONF hostkey:** `genkey` (regenerated at factory reset)
XPath: `/ietf-keystore:keystore/asymmetric-keys/asymmetric-key[name='genkey']`
- **Hostname format specifiers:**
XPath: `/ietf-system:system/hostname`
- `%i`: OS ID, from `/etc/os-release`, from Menuconfig branding
- `%h`: Default hostname, from `/etc/os-release`, from branding
- `%m`: NIC specific part of base MAC, e.g., to `c0-ff-ee`
- `%%`: Literal %
### Static Files
> **Caveat:** maintaining static a factory-config and failure-config may
> seem like an obvious choice, but as YANG models evolve (even the IETF
> models get upgraded), you may need to upgrade your static files.
First, for one-off builds (one image per product), the simplest way is
to override the location where the system looks for the files, `/etc`
already at build time. This can be done using a Buildroot rootfs
overlay providing, e.g., `/etc/factory-config.cfg`. Example: [NanoPi
R2S][] in `${INFIX}/board/aarch64/r2s/rootfs/etc/factory-config.cfg`.
Second, to support multiple products in a single image, we can employ
another method to install a `/etc/factory-config.cfg` override -- at
runtime. This relies on the very early system `probe` that detects the
specific product from VPD data.
The `probe` consists of several sequential steps that currently run from
`${INFIX}/board/common/rootfs/usr/libexec/infix/init.d/`. One of them
check if `/usr/share/product/<PRODUCT>` exists, and if so attempts to
copy the entire contents to `/`. Here, `<PRODUCT>` is determined from
the VPD, which is available in `/run/system.json` as `"product-name"`,
after `00-probe` has run. The lower case version of the string is used.
I.e., create a rootfs overlay that provides any combination of:
- `/usr/share/product/<PRODUCT>/etc/factory-config.cfg`
- `/usr/share/product/<PRODUCT>/etc/failure-config.cfg`
### Dynamically Generated
The generated `factory-config` and `failure-config` files consist of
both static JSON files and part generated files at runtime for each
device. The resulting files are written to the RAM disk in `/run`:
- `/run/confd/factory-config.gen`
- `/run/confd/failure-config.gen`
Provided no custom overrides (see above) have been installed already,
these files are then copied to:
- `/etc/factory-config.cfg`
- `/etc/failure-config.cfg`
... where the bootstrap process expects them to be in the next step.
Examples of generated contents are the SSH hostkey and hostname. The
latter is constructed from the file `/etc/hostname`, appended with the
last three octets of the system's base MAC address. To override the
base hostname, set `BR2_TARGET_GENERIC_HOSTNAME` in your defconfig.
The static files are installed by Infix `confd` in `/usr/share/confd/`
@@ -97,6 +176,112 @@ bootstrap script for more help, and up-to-date information.
> you may want to look into `GEN_IFACE_OPTS`.
### Example Snippets
**IETF System:**
```hsib
"ietf-system:system": {
"hostname": "example-%m",
"ntp": {
"enabled": true,
"server": [
{
"name": "ntp.org",
"udp": {
"address": "pool.ntp.org"
}
}
]
},
"authentication": {
"user": [
{
"name": "admin",
"password": "$factory$",
"infix-system:shell": "bash"
}
]
},
"infix-system:motd-banner": "Li0tLS0tLS0uCnwgIC4gLiAgfCBJbmZpeCAtLSBhIE5ldHdvcmsgT3BlcmF0aW5nIFN5c3RlbQp8LS4gdiAuLXwgaHR0cHM6Ly9rZXJuZWxraXQuZ2l0aHViLmlvCictJy0tLSctJwo="
}, # <---- REMEMBER COMMA SEPARATORS IN SNIPPETS!
# <---- ... and no comments.
```
The `motd-banner` is a binary type, which is basically a Base64 encoded
text file without line breaks (`-w0`):
```bash
$ echo "Li0tLS0tLS0uCnwgIC4gLiAgfCBJbmZpeCAtLSBhIE5ldHdvcmsgT3BlcmF0aW5nIFN5c3RlbQp8LS4gdiAuLXwgaHR0cHM6Ly9rZXJuZWxraXQuZ2l0aHViLmlvCictJy0tLSctJwo=" |base64 -d
.-------.
| . . | Infix -- a Network Operating System
|-. v .-| https://kernelkit.github.io
'-'---'-'
```
**IETF Keystore**
Notice how both the public and private keys are left empty here. The
`genkey` is always automatically regenerated after each factory reset.
Keeping the `factory-config` snippet like this means we can use the same
file on multiple devices, without risking them sharing the same host
keys. Sometimes you may want the same host keys, but that is the easy
use-case and not documented here.
```json
"ietf-keystore:keystore": {
"asymmetric-keys": {
"asymmetric-key": [
{
"name": "genkey",
"public-key-format": "ietf-crypto-types:ssh-public-key-format",
"public-key": "",
"private-key-format": "ietf-crypto-types:rsa-private-key-format",
"cleartext-private-key": "",
"certificates": {}
}
]
}
},
```
The `genkey` is currently only used by the NETCONF SSH backend.
**IETF NETCONF Server**
```json
"ietf-netconf-server:netconf-server": {
"listen": {
"endpoints": {
"endpoint": [
{
"name": "default-ssh",
"ssh": {
"tcp-server-parameters": {
"local-address": "::"
},
"ssh-server-parameters": {
"server-identity": {
"host-key": [
{
"name": "default-key",
"public-key": {
"central-keystore-reference": "genkey"
}
}
]
}
}
}
}
]
}
}
},
```
Integration
-----------
@@ -203,3 +388,4 @@ variable `INFIX_OEM_PATH` to that of the br2-external. It is also
possible to set the `GIT_VERSION` variable in your `post-build.sh`
script to change how the VCS version is extracted.
[NanoPi R2S]: https://github.com/kernelkit/infix/blob/main/board/aarch64/r2s/rootfs/etc/factory-config.cfg
+17 -10
View File
@@ -62,9 +62,11 @@ The system has several datastores (or files):
configure context. Any changes made here can be discarded (`abort`,
`rollback`) or committed (`commit`, `leave`) to `running-config`.
> See the [Branding & Releases](branding.md) for information on how
> `factory-config` and `failure-config` can be adapted to different
> customer requirements.
> Please see the [Branding & Releases](branding.md) document for more
> in-depth information on how `factory-config` and `failure-config` can
> be adapted to different customer requirements. Including how you can
> override the generated versions of these files with plain per-product
> ones -- this may even protect against some of the failure modes below.
## System Boot
@@ -90,7 +92,8 @@ password in VPD, are detailed later in this section.
there is a working fail safe, or rather *fail secure*, mode
4. On first power-on, and after a factory reset, the system does not
have a `startup-config`, in which case `factory-config` is copied
to `startup-config`
to `startup-config` -- if a per-product specific version exists it
is preferred over the generated one
5. Provided the integrity of the `startup-config` is OK, a system
service loads and activates the configuration
@@ -123,13 +126,17 @@ syntax, failed validation against the system's YANG model, or a bug in
the system's `confd` service, the *Fail Secure Mode* is triggered and
`failure-config` is loaded (unless VPD Failure, see above).
> Again, please see the [Branding & Releases](branding.md) document for
> how to provide a per-product hard-coded `failure-config` to suit your
> products preferences.
*Fail Secure Mode* is a fail-safe mode provided for debugging the
system. The default[^3] is isolated interfaces with communication only to
the management CPU, SSH and console login using the device's factory
reset password, IP connectivity only using IPv6 link-local, and device
discovery protocols: LLDP, mDNS-SD. The login and shell prompt are set
to `failure-c0-ff-ee`, the last three octets of the device's base MAC
address.
system. The default[^3] creates a setup of isolated interfaces with
communication only to the management CPU, SSH and console login using
the device's factory reset password, IP connectivity only using IPv6
link-local, and device discovery protocols: LLDP, mDNS-SD. The login
and shell prompt are set to `failure-c0-ff-ee`, the last three octets of
the device's base MAC address.
[^1]: YANG is a modeling language from IETF, replacing that used for
SNMP (MIB), used to describe the subsystems and properties of