Re: [PATCH v5 09/12] evm: Allow setxattr() and setattr() for unmodified metadata

[Mimi Zohar]

On Mon, 2021-05-03 at 14:48 +0000, Roberto Sassu wrote:
> > From: Mimi Zohar [mailto:zohar@xxxxxxxxxxxxx]
> > Sent: Monday, May 3, 2021 3:00 PM
> > On Wed, 2021-04-07 at 12:52 +0200, Roberto Sassu wrote:
> > 
> > > diff --git a/security/integrity/evm/evm_main.c
> > b/security/integrity/evm/evm_main.c
> > > @@ -389,6 +473,11 @@ static int evm_protect_xattr(struct
> > user_namespace *mnt_userns,
> > >  	if (evm_status == INTEGRITY_FAIL_IMMUTABLE)
> > >  		return 0;
> > >
> > > +	if (evm_status == INTEGRITY_PASS_IMMUTABLE &&
> > > +	    !evm_xattr_change(mnt_userns, dentry, xattr_name, xattr_value,
> > > +			      xattr_value_len))
> > > +		return 0;
> > > +
> > 
> > If the purpose of evm_protect_xattr() is to prevent allowing an invalid
> > security.evm xattr from being re-calculated and updated, making it
> > valid, INTEGRITY_PASS_IMMUTABLE shouldn't need to be conditional.  Any
> > time there is an attr or xattr change, including setting it to the
> > existing value, the status flag should be reset.
> 
> The status is always reset if evm_protect_xattr() returns 0. This does not
> change.
> 
> Not making INTEGRITY_PASS_IMMUTABLE conditional would cause issues.
> Suppose that the status is INTEGRITY_FAIL. Writing the same xattr would
> cause evm_protect_xattr() to return 0 and the HMAC to be updated.

This example is mixing security.evm types.  Please clarify.

> > I'm wondering if making INTEGRITY_PASS_IMMUTABLE conditional would
> > prevent the file from being resigned.
> 
> INTEGRITY_FAIL_IMMUTABLE should be enough to continue the
> operation.

Agreed.

Mimi