Re: [RFC PATCH v2] ima,fuse: introduce new fs flag FS_NO_IMA_CACHE

[Mimi Zohar <zohar@linux.vnet.ibm.com>]

On Fri, 2018-01-19 at 11:35 +0100, Alban Crequy wrote:
> On Thu, Jan 18, 2018 at 10:25 PM, Mimi Zohar <zohar@linux.vnet.ibm.com> wrote:
> > On Tue, 2018-01-16 at 16:10 +0100, Alban Crequy wrote:
> >> From: Alban Crequy <alban@kinvolk.io>
> >>
> >> This patch forces files to be re-measured, re-appraised and re-audited
> >> on file systems with the feature flag FS_NO_IMA_CACHE. In that way,
> >> cached integrity results won't be used.
> >>
> >> For now, this patch adds the new flag only FUSE filesystems. This is
> >> needed because the userspace FUSE process can change the underlying
> >> files at any time.
> >
> > Thanks, it's working nicely.
> >
> >
> >> diff --git a/include/linux/fs.h b/include/linux/fs.h
> >> index 511fbaabf624..2bd7e73ebc2a 100644
> >> --- a/include/linux/fs.h
> >> +++ b/include/linux/fs.h
> >> @@ -2075,6 +2075,7 @@ struct file_system_type {
> >>  #define FS_BINARY_MOUNTDATA  2
> >>  #define FS_HAS_SUBTYPE               4
> >>  #define FS_USERNS_MOUNT              8       /* Can be mounted by userns root */
> >> +#define FS_NO_IMA_CACHE              16      /* Force IMA to re-measure, re-appraise, re-audit files */
> >>  #define FS_RENAME_DOES_D_MOVE        32768   /* FS will handle d_move() during rename() internally. */
> >>       struct dentry *(*mount) (struct file_system_type *, int,
> >>                      const char *, void *);
> >>
> >
> > Since IMA is going to need another flag, we probably should have a
> > consistent prefix (eg. "FS_IMA").  Maybe rename this flag to
> > FS_IMA_NO_CACHE.
> 
> Ok, I can rename it.
> 
> Is there a discussion about the other IMA flag?

There's not a single thread that I can point to, but more of an on
going discussion as to what it means for a filesystem to support IMA
and how that decision is made.

- Initial measuring, verifying, auditing files
- Safely detecting when a file changes
- Not applicable/supported

With Sascha Hauer's patch "ima: Use i_version only when filesystem
supports it" and this patch, the second issue is addressed, but will
cause files to be re-validated, perhaps unnecessarily, impacting
performance.

Some filesystems should not be evaluated, such as pseudo filesystems
(eg. cgroups, sysfs, devpts, pstorefs, efivarfs, debugfs, selinux,
smack).  Instead of defining a flag indicating whether or not IMA is
applicable/supported, we should define a new flag, indicating whether
it is a pseudo filesystem.  This would eliminate a large portion of at
least the builtin IMA policy rules.

> > I'm also wondering if this change should be
> > separated from the IMA change.
> 
> Do you mean one patch for adding the flag and the IMA change and
> another patch for using the flag in FUSE?

The flag and FUSE usage of the flag, separately from IMA.

Mimi