Re: [PATCH v6 0/6] define new fs integrity_read method

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

On Wed, 2017-08-16 at 19:52 +1000, James Morris wrote:
> On Wed, 16 Aug 2017, Christoph Hellwig wrote:
> 
> > On Wed, Aug 16, 2017 at 12:43:58PM +1000, James Morris wrote:
> > > On Tue, 15 Aug 2017, Mimi Zohar wrote:
> > > 
> > > > To resolve this locking problem, this patch set introduces a new
> > > > ->integrity_read file operation method.  Until all filesystems
> > > > define the new ->integrity_read method, files that were previously
> > > > measured might not be currently measured and files that were
> > > > previously appraised might fail to be appraised properly.
> > > 
> > > Are there any such filesystems in mainline which are not getting an 
> > > integrity_read method in this patchset?
> > 
> > There are a few, mostly because we're pretty sure the previous integrity
> > code did the wrong thing for them - e.g. ocfs2 and gfs2 where locking
> > vs operations on other cluster nodes was missing, or NFS where in addition
> > to the above deadlocks were 100% reprodicible with current code.
> 
> Should we do a warn_once for these filesystems when IMA is used?

I don't think it is necessary.  In terms of IMA-measurement, any file
in policy on an unsupported filesystem will be in the measurement
list, but the file hash will be 0's.  In terms of IMA-appraisal, any
file in policy on an unsupported filesystem will fail appraisal, since
the file hash is 0.

A separate patch set will emit a warning when a file system is mounted
without i_version.

Mimi