Falco depends on a driver that taps into the stream of system calls on a machine and passes those system calls to user space.
The kernel module called
falco is the default driver. Alternatively, an eBPF probe can be used.
The Falco project has three different kind of drivers.
- Kernel module
- eBPF probe
- Userspace instrumentation
By default, the kernel module will be installed when installing the Falco debian/rpm package, when running the
falco-driver-loader script shipped within the binary package, or when running the
falcosecurity/falco-driver-loader docker image (that just wraps the aforementioned script).
To install the kernel module, please refer to the installation page.
Falco tries to use the kernel module driver by default.
The eBPF probe is an alternative to the one described above. Note that eBPF is a feature supported only by recent kernels.
To install the eBPF probe, please refer to the installation page.
To enable the eBPF support in Falco set the
FALCO_BPF_PROBE environment variable to an empty value (ie.
FALCO_BPF_PROBE="") or otherwise explicitly set it to the path where the eBPF probe resides.
Differently from the other drivers, as the name suggests, userspace instrumentations happens 100% in userspace.
The Falco community, in 0.24.0 promoted the userspace instrumentation feature to official support to be included in Falco.
However, there's a difference between userspace instrumentation and the other drivers. At the moment of writing, the Falco project does not have any officially supported userspace instrumentation implementation.
To summarize: the code that defines the contracts to do userspace instrumentation in Falco itself is stable and under official support. While there's no implementation that rached the status of official support yet
The community is working on an implementation based on PTRACE(2) that you can find falcosecurity/pdig.
How to enable userspace instrumentation in Falco:
- Start Falco with the
--userspaceflag. This will tell Falco to look at userspace instrumentation instead of looking at the Kernel module (the default).
- Select an userspace instrumentation implementation you want to use (let's take pdig as it is the only one available now)
- Now, since pdig is still incubating - a very early stage - it's not included in our release process. This, for you means that to install it, you will need to compile it yourself. Follow the instructions here.
- Now that you have pdig installed, you will need to start it. Remember, in the case of pdig, it does not know the root process tree you want to instrument by itself. You have to specify that via the
-pflag or start the process with pdig itself. It's very similar to running a debugger against a binary. You can run a process directly using it or instrument an existing process. Full instructions here.
There are attempts at making the installation process easier, the community didn't pick one yet and it's likely that many iterations will go into making any one of the below GA. However, here's a list of projects you can look at if you want to get Falco with userspace instrumentation in Kubernetes.
- Falco Trace - a convenient container image to use as a base to add userspace instrumentation to your images.
- Falco Inject - a tool that uses Falco Trace artifacts to inject Falco and userspace instrumentation into your containers via the kubernetes API.
As you probably already understood, the userspace instrumentation drivers are a bit different, handle with care!
Was this page helpful?
Glad to hear it! Please tell us how we can improve.
Sorry to hear that. Please tell us how we can improve.