Re: [PATCH v2] ima: drop vla in ima_audit_measurement()

[Tycho Andersen]

Hi Mimi,

On Thu, Mar 08, 2018 at 05:05:40PM -0500, Mimi Zohar wrote:
> On Thu, 2018-03-08 at 14:45 -0700, Tycho Andersen wrote:
> > Hi Mimi,
> > 
> > On Thu, Mar 08, 2018 at 03:36:14PM -0500, Mimi Zohar wrote:
> > > On Thu, 2018-03-08 at 13:23 -0700, Tycho Andersen wrote:
> > > 
> > > >  /*
> > > > diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c
> > > > index 2cfb0c714967..356faae6f09c 100644
> > > > --- a/security/integrity/ima/ima_main.c
> > > > +++ b/security/integrity/ima/ima_main.c
> > > > @@ -288,8 +288,11 @@ static int process_measurement(struct file *file, char *buf, loff_t size,
> > > >  					      xattr_value, xattr_len, opened);
> > > >  		inode_unlock(inode);
> > > >  	}
> > > > -	if (action & IMA_AUDIT)
> > > > -		ima_audit_measurement(iint, pathname);
> > > > +	if (action & IMA_AUDIT) {
> > > > +		rc = ima_audit_measurement(iint, pathname);
> > > > +		if (rc < 0)
> > > > +			goto out_locked;
> > > > +	}
> > > > 
> > > >  	if ((file->f_flags & O_DIRECT) && (iint->flags & IMA_PERMIT_DIRECTIO))
> > > >  		rc = 0;
> > > 
> > > Only when IMA-appraisal is enforcing file data integrity should
> > > process_measurement() ever fail.  Other errors can be logged/audited.
> > 
> > Ok, so previously in ima_audit_measurement() when allocation failed,
> > there was nothing logged. If we just keep this behavior like below,
> > does that look good?
> 
> Before the IMA locking change that were just upstreamed, there were
> problems with measuring/appraising files that were opened with the
> O_DIRECT flag.  Unless the IMA policy specified permit_directio, the
> measurement/appraisal failed.  With the new locking, opening files
> with the O_DIRECTIO flag shouldn't be a problem.  It just needs to be
> fully tested before removing this code.
> 
> On failure, the code below tests the ima_audit_measurement() result
> and skips the IMA_PERMIT_DIRECTIO test.  Unless I'm missing something,
> I don't see the point.

It skips the IMA_PERMIT_DIRECTIO test because it's already going to
fail: we're in enforce mode and we got an allocation failure and so we
can't audit this access (note: there is another allocation failure in
ima_audit_measurement() which is still ignored after this patch, so
maybe ignoring failures is ok; seems like it's not, though).

I'm not sure I really understand the rest of your message though. Can
you suggest what the patch should do here? Should we just ignore all
failures as before?

Tycho