5 ms·

Would be cool if there was a tool that could detect that configured path was attempted to be opened, and redirect the open() syscall to the real path... But tha

Would be cool if there was a tool that could detect that configured path was attempted to be opened, and redirect the open() syscall to the real path... But that would most likely require kernel support. Boxxy seems good enough for userspace.

notamy 4y ago
I actually attempted using ptrace to rewrite syscalls first! It was... horribly painful, and didn't work anywhere near as well as the bind-mount version.
- oneplane 4y ago
  In the past I used something like inotab to use an inotify-based trigger to pipe data from SSH to a different system that didn't have any NAS or SAN support, it might also work to detect 'who' is touching any files that boxxy has previously seen rules for. Perhaps still too tricky to make it do magic things and break programs in the process, but it could be used to audit who's working with what paths and let the user print a report so they know what apps to boxx up and make them behave.
  - notamy 4y ago
    That is an excellent idea! Something like could definitely be worth adding. It's why there's a "remount rootfs as ro" flag; that way anything not specified in rules is ro and misbehaving programs will explode.
    - oneplane 4y ago
      Love exploding apps. That's what they get for eating my filesystem.
- yegle 4y ago
  Fun fact, one layer of App Engine's sandbox from a couple years ago was implemented using ptrace. It will redirect filesystem IO to in-memory files.
  - notamy 4y ago
    Interesting! That makes perfect sense, I just don't think I'm smart enough to use ptrace properly right now :P
    - efreak 4y ago
      Try checking out proot?
    - gary_0 4y ago
      Were you mostly going off the ptrace man page? I tried reading it to figure out ptrace and it made me feel not smart enough too.
      - notamy 4y ago
        The man page + Google! There's very few good examples of it, and I've accepted that I'm just not familiar enough with that specific problem space.
- lytigas 4y ago
  For a structured approach to ptrace/syscall rewriting, you could try FB's reverie. I worked on and used it during an internship a few years back; it's pretty amazing at what it does. https://github.com/facebookexperimental/reverie https://github.com/facebookexperimental/reverie
sjaak 4y ago
Lo and behold http://ordiluc.net/fs/libetc/ http://ordiluc.net/fs/libetc/

notamy 4y ago

I actually attempted using ptrace to rewrite syscalls first! It was... horribly painful, and didn't work anywhere near as well as the bind-mount version.

oneplane 4y ago
In the past I used something like inotab to use an inotify-based trigger to pipe data from SSH to a different system that didn't have any NAS or SAN support, it might also work to detect 'who' is touching any files that boxxy has previously seen rules for. Perhaps still too tricky to make it do magic things and break programs in the process, but it could be used to audit who's working with what paths and let the user print a report so they know what apps to boxx up and make them behave.
- notamy 4y ago
  That is an excellent idea! Something like could definitely be worth adding. It's why there's a "remount rootfs as ro" flag; that way anything not specified in rules is ro and misbehaving programs will explode.
  - oneplane 4y ago
    Love exploding apps. That's what they get for eating my filesystem.
yegle 4y ago
Fun fact, one layer of App Engine's sandbox from a couple years ago was implemented using ptrace. It will redirect filesystem IO to in-memory files.
- notamy 4y ago
  Interesting! That makes perfect sense, I just don't think I'm smart enough to use ptrace properly right now :P
  - efreak 4y ago
    Try checking out proot?
  - gary_0 4y ago
    Were you mostly going off the ptrace man page? I tried reading it to figure out ptrace and it made me feel not smart enough too.
    - notamy 4y ago
      The man page + Google! There's very few good examples of it, and I've accepted that I'm just not familiar enough with that specific problem space.
lytigas 4y ago
For a structured approach to ptrace/syscall rewriting, you could try FB's reverie. I worked on and used it during an internship a few years back; it's pretty amazing at what it does. https://github.com/facebookexperimental/reverie https://github.com/facebookexperimental/reverie

sjaak 4y ago

Lo and behold http://ordiluc.net/fs/libetc/ http://ordiluc.net/fs/libetc/