created_at2022-09-17 16:47:13.511945
updated_at2022-09-25 16:48:57.177521
descriptionAn on demand publisher of sysfs/procfs
Michael Bacarella (mbac-architect)



Netidx System File Publisher

netidx-sysfs publishes the contents of small system files, such as those found in sysfs and procfs, to netidx, and allows writing values back subject to permissions. Because of the nature of sysfs and procfs netidx-sysfs behaves in the following potentially unexpected ways,

  • pure polling of structure and file contents, no use of e.g. inotify. This is because inotify either doesn't work at all, or only partly works on the target filesystems.
  • linear backoff of polling frequency for both structure and files to reduce overhead for parts of the filesystem that don't change often. Maximum poll interval is currently 120 seconds, while the minimum is 1 second. Only a changed file will produce an update, even though it polls at least every 120 seconds no update will be produced if the file's contents didn't change.
  • on demand polling for both structure and files. Only files that are subscribed are polled. Structure is only polled near subscribed files (e.g. directories are only read if they contain a subscribed file, or a recent attempt to subscribe to a file). As a result netidx-sysfs should consume no cpu time and only a little memory if it isn't being used.
  • cross platform (unix, maybe windows). Does not use io uring or other patform specific apis, only tokio and std file operations.
  • only reads the first 1k of each file, regardless of how big it is. Ideal for small files, tolerant of but maybe useless for large ones. Not at all limited to sysfs/procfs, just tailored to that use case.


To setup cargo install netidx-sysfs and make sure you have a netidx resolver either on the local machine or somewhere on the network. To run, either run as a regular user, in which case you won't be able to see everything, or run as root for full access. E.G.

# netidx-sysfs -a local -b local --netidx-base /local/system/sysfs --path /sys

will publish sysfs to /local/system/sysfs

Of course you are not limited to publishing locally, you could publish every machine on your network's sysfs to netidx so you can read and manipulate values over the network while enjoying strong encryption, authentication, and authorization provided by netidx and your kerberos v5 infrastructure.


Because the resolver and browser automatically infer table structure from regular tree structure, many of the things in sysfs and procfs will be automatically formatted as tables. For example, here is a look at my /sys/block


Commit count: 58

cargo fmt