Re: [PATCH v11 21/48] ext4: Add richacl feature flag

[Austin S Hemmelgarn]

On 2015-10-17 19:17, Dave Chinner wrote:

On Fri, Oct 16, 2015 at 02:27:57PM -0400, Austin S Hemmelgarn wrote:

On 2015-10-16 13:41, Andreas Gruenbacher wrote:

On Fri, Oct 16, 2015 at 7:31 PM, Austin S Hemmelgarn
<ahferro...@gmail.com> wrote:

I would like to re-iterate, on both XFS and ext4, I _really_ think this
should be a ro_compat flag, and not an incompat one.  If a person has the
ability to mount the FS (even if it's a read-only mount), then they by
definition have read access to the file or partition that the filesystem is
contained in, which means that any ACL's stored on the filesystem are
functionally irrelevant,

It is unfortunately not safe to make such a file system accessible to
other users, so the feature is not strictly read-only compatible.

If it's not safe WRT data integrity, then the design needs to be
reworked, as that directly implies that isn't safe for even every
day usage on a writable filesystem.

This is exactly what we have *incompat feature flags for*: to
protect old code that doesn't know about potentially dangerous new
on-disk formats from trying to parse those formats and causing
unpredictable bad things from happening.

However, unless things have changed (I haven't had time to re-read the patch-set yet), then the only change will be a new set of xattrs, and that type of change _shouldn't_ break existing code that doesn't know about them. Andreas really needs to explain _exactly_ why it isn't safe to mount this on a kernel that doesn't support it and let other users access it, and if the answer is 'because the ACL's won't be honored' then that really isn't acceptable reason IMHO, because (as I outlined in the previous e-mail) being able to mount the filesystem implies that they have at least read access to the underlying storage, which means that the ACL's in the filesystem are irrelevant as far as any competent individual who is actively trying to illegitimately access the data is concerned.

Austin, your arguments hold no weight because they are no different
to the considerations for any new on-disk feature: the user needs to
have both kernel and userspace support to recover filesystems that
go bad. If you are using a brand new fs/kernel feature, then it is
expected that you know that your DR processes take this into
account.

Except that the given argument from Andreas as to why it's an incompat feature does not clarify whether it's to prevent breaking the existing filesystem code (which I understand and agree is a proper usage for such a flag), or to try and provide some thin facade of security when there really is none (which is what the bit about 'and expose it to other users' really sounds like to me). Yes the argument that I made which you have replied to was admittedly shortsighted and didn't need to be made to get the point that I was actually trying to make across, and I sincerely apologize for that.

smime.p7s
Description: S/MIME Cryptographic Signature