4 ms·

Filesystem mounts in user namespaces

47 points by vezzy-fnord 11y ago

angersock 11y ago
Those who don't run Plan 9 are doomed to re-implement it. Poorly. http://doc.cat-v.org/plan_9/misc/ubiquitous_fileserver/ubiquitous_fileserver.pdf http://doc.cat-v.org/plan_9/misc/ubiquitous_fileserver/ubiqu...
- vezzy-fnord 11y ago
  Actually, I think the chief initial architect of Linux namespaces (and bind mounts) was Al Viro [1], an ex-Plan 9 developer. So he's well aware of that, but evidently a poor concession is better than nothing at all. [1] Citation: http://www.ibm.com/developerworks/library/l-mount-namespaces/ http://www.ibm.com/developerworks/library/l-mount-namespaces...
  - angersock 11y ago
    Good to know! I've spent the weekend reading Plan 9 papers, and it's just all such a shame. They tried to save us!
    - vezzy-fnord 11y ago
      Yep, that's why Al Viro is such an abrasive yet highly entertaining grump. He makes Linus Torvalds look like a sensitivity training counselor in comparison, but his argumentative style is great (query "site:lwn.net al viro udev" for a fun way to waste some time). A veteran Linux kernel hacker who hates udev and cgroups. Yep.
      - e12e 11y ago
        Do you know if he's back on the lists/development in full force? Looks like he had some very serious medical issues last year: http://lwn.net/Articles/439635/ http://lwn.net/Articles/439635/
- Animats 11y ago
  "In the end, the only safe way to allow unprivileged mounts may be the filesystem in user space (FUSE) mechanism, which isolates the filesystem code into a separate user-space process." Which is how QNX, the message-passing real-time OS, does it. All file systems are in user space. ZIP file systems were added easily, for example. QNX has a process which starts at boot time to manage pathname connections to "resource manager" programs. Programs can register for a portion of the pathname space. Pathnames are a level of abstraction below file systems. Linux keeps adding the ability to put more in user space, with interprocess message passing, drivers in user space, and file systems in user space. But the old in-kernel code for the same things never gets thrown out, and the bolted-on user space capabilities are slower than the kernel ones. Message passing support and CPU dispatching have to be tightly integrated for this to work fast, and because message passing is an afterthought in UNIX/Linux, it never is.
viraptor 11y ago
Are there any (still maintained) microkernel systems which do unprivileged user mounts? I know Minix3 at least has filesystem services, but I'm not sure if it goes as far as allowing someone to build (for example) sshfs that allows mounts in /home/...
- vezzy-fnord 11y ago
  GNU Hurd.
  - e12e 11y ago
    Just in case someone thinks this was a joke (maybe it was! :-) -- there are few people trying hard to actually make Hurd work in some meaningful sense of the word, see eg: https://lists.debian.org/debian-hurd/2015/04/msg00047.html https://lists.debian.org/debian-hurd/2015/04/msg00047.html
    - viraptor 11y ago
      I don't think it was a joke. If people are interested in checking this out, the Hurd name for this functionality is "translators". And it seems to do what I expected.
      - vezzy-fnord 11y ago
        Indeed, it was no joke.

47 points by vezzy-fnord 11y ago

Those who don't run Plan 9 are doomed to re-implement it. Poorly. http://doc.cat-v.org/plan_9/misc/ubiquitous_fileserver/ubiquitous_fileserver.pdf http://doc.cat-v.org/plan_9/misc/ubiquitous_fileserver/ubiqu...

vezzy-fnord 11y ago
Actually, I think the chief initial architect of Linux namespaces (and bind mounts) was Al Viro [1], an ex-Plan 9 developer. So he's well aware of that, but evidently a poor concession is better than nothing at all. [1] Citation: http://www.ibm.com/developerworks/library/l-mount-namespaces/ http://www.ibm.com/developerworks/library/l-mount-namespaces...
- angersock 11y ago
  Good to know! I've spent the weekend reading Plan 9 papers, and it's just all such a shame. They tried to save us!
  - vezzy-fnord 11y ago
    Yep, that's why Al Viro is such an abrasive yet highly entertaining grump. He makes Linus Torvalds look like a sensitivity training counselor in comparison, but his argumentative style is great (query "site:lwn.net al viro udev" for a fun way to waste some time). A veteran Linux kernel hacker who hates udev and cgroups. Yep.
    - e12e 11y ago
      Do you know if he's back on the lists/development in full force? Looks like he had some very serious medical issues last year: http://lwn.net/Articles/439635/ http://lwn.net/Articles/439635/
Animats 11y ago
"In the end, the only safe way to allow unprivileged mounts may be the filesystem in user space (FUSE) mechanism, which isolates the filesystem code into a separate user-space process." Which is how QNX, the message-passing real-time OS, does it. All file systems are in user space. ZIP file systems were added easily, for example. QNX has a process which starts at boot time to manage pathname connections to "resource manager" programs. Programs can register for a portion of the pathname space. Pathnames are a level of abstraction below file systems. Linux keeps adding the ability to put more in user space, with interprocess message passing, drivers in user space, and file systems in user space. But the old in-kernel code for the same things never gets thrown out, and the bolted-on user space capabilities are slower than the kernel ones. Message passing support and CPU dispatching have to be tightly integrated for this to work fast, and because message passing is an afterthought in UNIX/Linux, it never is.

viraptor 11y ago

Are there any (still maintained) microkernel systems which do unprivileged user mounts? I know Minix3 at least has filesystem services, but I'm not sure if it goes as far as allowing someone to build (for example) sshfs that allows mounts in /home/...

vezzy-fnord 11y ago
GNU Hurd.
- e12e 11y ago
  Just in case someone thinks this was a joke (maybe it was! :-) -- there are few people trying hard to actually make Hurd work in some meaningful sense of the word, see eg: https://lists.debian.org/debian-hurd/2015/04/msg00047.html https://lists.debian.org/debian-hurd/2015/04/msg00047.html
  - viraptor 11y ago
    I don't think it was a joke. If people are interested in checking this out, the Hurd name for this functionality is "translators". And it seems to do what I expected.
    - vezzy-fnord 11y ago
      Indeed, it was no joke.