Linux’s failfs requires filesystem access to start from explicit handles
Linux documents failfs, an internal filesystem that rejects path operations rather than exposing an empty directory. A process using it as its root cannot resolve absolute paths, including interpreters needed by dynamically linked executables. Lookups anchored at explicit directory file descriptors still work. This creates a useful distinction for confinement: access can depend on handles the process already holds, rather than an ambient root directory.
Linux Kernel Documentation
Why it matters
Filesystem confinement becomes easier to reason about when permitted starting points are explicit instead of inherited through ordinary path lookup.
More context
The documentation describes entry through the FD_FAILFS_ROOT sentinel and fchroot, with privilege or specific unprivileged-entry conditions. Watch this low-level interface and study it in an isolated compatible kernel environment. The inspected page is next-tree documentation; its exact publication date was not established.
Keep in mind
failfs is not a complete sandbox. Existing handles and namespace privileges still matter, and the documentation discusses additional syscall restrictions needed to prevent leaving it. Application compatibility is limited.
Original read