Re: Shameless plug for the FS Track at LPC next week!

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Sep 17, 2021 at 12:38 PM Jan Kara <jack@xxxxxxx> wrote:
>
> On Fri 17-09-21 10:36:08, Jan Kara wrote:
> > Let me also post Amir's thoughts on this from a private thread:
>
> And now I'm actually replying to Amir :-p
>
> > On Fri 17-09-21 10:30:43, Jan Kara wrote:
> > > We did a small update to the schedule:
> > >
> > > > Christian Brauner will run the second session, discussing what idmapped
> > > > filesystem mounts are for and the current status of supporting more
> > > > filesystems.
> > >
> > > We have extended this session as we'd like to discuss and get some feedback
> > > from users about project quotas and project ids:
> > >
> > > Project quotas were originally mostly a collaborative feature and later got
> > > used by some container runtimes to implement limitation of used space on a
> > > filesystem shared by multiple containers. As a result current semantics of
> > > project quotas are somewhat surprising and handling of project ids is not
> > > consistent among filesystems. The main two contending points are:
> > >
> > > 1) Currently the inode owner can set project id of the inode to any
> > > arbitrary number if he is in init_user_ns. It cannot change project id at
> > > all in other user namespaces.
> > >
> > > 2) Should project IDs be mapped in user namespaces or not? User namespace
> > > code does implement the mapping, VFS quota code maps project ids when using
> > > them. However e.g. XFS does not map project IDs in its calls setting them
> > > in the inode. Among other things this results in some funny errors if you
> > > set project ID to (unsigned)-1.
> > >
> > > In the session we'd like to get feedback how project quotas / ids get used
> > > / could be used so that we can define the common semantics and make the
> > > code consistently follow these rules.
> >
> > I think that legacy projid semantics might not be a perfect fit for
> > container isolation requirements. I added project quota support to docker
> > at the time because it was handy and it did the job of limiting and
> > querying disk usage of containers with an overlayfs storage driver.
> >
> > With btrfs storage driver, subvolumes are used to create that isolation.
> > The TREE_ID proposal [1] got me thinking that it is not so hard to
> > implement "tree id" as an extention or in addition to project id.
> >
> > The semantics of "tree id" would be:
> > 1. tree id is a quota entity accounting inodes and blocks
> > 2. tree id can be changed only on an empty directory
> > 3. tree id can be set to TID only if quota inode usage of TID is 0
> > 4. tree id is always inherited from parent
> > 5. No rename() or link() across tree id (clone should be possible)
> >
> > AFAIK btrfs subvol meets all the requirements of "tree id".
> >
> > Implementing tree id in ext4/xfs could be done by adding a new field to
> > inode on-disk format and a new quota entity to quota on-disk format and
> > quotatools.
> >
> > An alternative simpler way is to repurpose project id and project quota:
> > * Add filesystem feature projid-is-treeid
> > * The feature can be enabled on fresh mkfs or after fsck verifies "tree id"
> >    rules are followed for all usage of projid
> > * Once the feature is enabled, filesystem enforces the new semantics
> >   about setting projid and projid_inherit
> >
> > This might be a good option if there is little intersection between
> > systems that need to use the old project semantics and systems
> > that would rather have the tree id semantics.
>
> Yes, I actually think that having both tree-id and project-id on a
> filesystem would be too confusing. And I'm not aware of realistic usecases.
> I've heard only of people wanting current semantics (although these we more
> of the kind: "sometime in the past people used the feature like this") and
> the people complaining current semantics is not useful for them. This was
> discussed e.g. in ext4 list [2].
>
> > I think that with the "tree id" semantics, the user_ns/idmapped
> > questions become easier to answer.
> > Allocating tree id ranges per userns to avoid exhausting the tree id
> > namespace is a very similar problem to allocating uids per userns.
>
> It still depends how exactly tree ids get used - if you want to use them to
> limit space usage of a container, you still have to forbid changing of tree
> ids inside the container, don't you?
>

Yes.
This is where my view of userns becomes hazy (so pulling Christain into
the discussion), but in general I think that this use case would be similar
to the concept of single uid container - the range of allowed tree ids that
is allocated for the container in that case is a single tree id.

I understand that the next question would be about nesting subtree quotas
and I don't have a good answer to that question.

Are btrfs subvolume nested w.r.t. capacity limit? I don't think that they are.

Thanks,
Amir.



[Index of Archives]     [Reiser Filesystem Development]     [Ceph FS]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite National Park]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]     [Linux Media]

  Powered by Linux