Profiling Flatpak’d applications

One of the great powers of namespace APIs on Linux (mount namespaces, user namespaces, etc) is that you can create a new view into the world of your computer that is very different from the host. This can make traditional profiling tools difficult.

To begin with, we need to ensure that we have access to ptrace or perf infrastructure. Easy enough, just don’t drop those privileges before calling execve(). This is the --allow=devel option to flatpak run. But after that, we need to do the detailed phase of translating instruction pointers to a function name.

Generally, the translation between an instruction pointer and a function name requires looking up what file is mapped over the address. Then you open that file with an ELF reader and locate the file containing debug information (which may be the same file). Then open that file and locate what function contains that instruction pointer. (We subtract the beginning of the map from the instruction pointer to get a relative offset).

Now, here is the trouble with mount namespaces. The “path” of the map might be something like “/newroot/usr/lib/“. Of course, “/newroot/” doesn’t actually exist.

So… what to do.

Well, we can find information about the mounts in the process by looking at /proc/$pid/mountinfo. Just look for the longest common prefix to get the translation from the mount namespace into the host. One caveat though. The translated path is relative to the root of the partition mounted. So the next step is to locate the file-system mount on the host (via /proc/mounts).

If you put this all together, you can profile and extract symbols from applications running inside of containers.

Now, most applications aren’t going to ship debug symbols with them. But we’ll need those for proper symbol extraction. So if you install the .Debug variant of a Flatpak runtime (org.gnome.Sdk.Debug for example) then we can setup symbol resolution specifically for /usr/lib/debug. Builder now does this.

I’m not sure, but I think this makes Builder and Sysprof one of the first (if not the first) profiler on Linux to support symbol extraction while profiling containerized applications from outside the container.

I’ve only tested with Flatpak, but I don’t see why this code can’t work with other tooling using mount namespaces.

Valgrind Integration in Builder

I wanted a bit of a distraction this evening, so I put together a quick valgrind plugin. Just select “Run with Valgrind” from the drop down menu and it will make sure your program is run using the venerable valgrind.

If valgrind isn’t available in your runtime (host, jhbuild, flatpak, etc) you wont see the menu item. But thankfully the org.gnome.Sdk runtime we use to build GNOME apps has it, so for many of you it should Just Work™.

Builder’s Build Pipeline

One of the core features of Builder is, well, building. So we need to provide the best experience we can. That means we need wide support for build systems and languages. How we get from source code to a product can vary in a multitude of ways.

For example, you might need to alter how you launch programs to run inside a container. You might need access to headers that only exist in the SDK and not the application runtime. You might even need to download or build dependencies for your project at a given point. Furthermore, this could all change based on what plugins you’ve chosen to enable in your Builder environment.

When I set out to create Builder, I knew that I wouldn’t be able to see all the possibilities that Builder needs to handle. So I explicitly avoided designing an abstract solution until I understood the problem better. But I think we understand the problem well enough now. So over the winter holiday I started putting together the design of IdeBuildPipeline.

The pipeline is a one-directional series of build phases. Each phase can have stages attached to it. Builds progress in the order of their attached stages, each of which can be an asynchronous operation. We only need to advance the build up to the requested phase. If the IDE detects something has changed, it can invalidate various phases of the pipeline causing them to be executed again at the next build request.

Plugins should implement the IdeBuildPipelineAddin interface to register any necessary pipeline stages.

Generally, build stages are an IdeSubprocessLauncher, but you can implement custom IdeBuildStage subclasses to do more fancy things. For example, if you need to download something, you might add an IdeBuildStageTransfer and you’ll see the transfer show up in the transfers window.

Of course, some problems are complex enough where we don’t know if they are invalidated upfront and the build stage may query some external resource. For that, the IdeBuildStage class can implement the query signal. Of course we should strive to perform blocking operations asynchronously so consider pausing the stage until your asynchronous request has completed.

We know how difficult it is for newcomers to get started so we’ve taught Builder how to download Flatpak runtimes and SDKs based on your project configuration. That means that you can omit those flatpak install commands from your “project setup” guide. Just clone the project URL and click Build, we’ll take care of the rest.

Thanks to this new design, implementing a build system is much easier. To prove this, I finally implemented a basic cmake+ninja plugin in a matter of an hour. Check it out here. If you’d like to write CFLAG extraction for cmake, we could use that still!

We now have support for autotools, cargo, cmake, and meson build systems.