Re: [tpmdd-devel] [PATCH RFC v3 5/5] tpm2: expose resource manager via a device link /dev/tpms<n>

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Fri, 2017-01-20 at 23:05 +0200, Jarkko Sakkinen wrote:
> On Fri, Jan 20, 2017 at 03:39:14PM +0200, Jarkko Sakkinen wrote:
> > On Thu, Jan 19, 2017 at 07:19:40AM -0500, James Bottomley wrote:
> > > On Thu, 2017-01-19 at 12:49 +0200, Jarkko Sakkinen wrote:
> > > > On Wed, Jan 18, 2017 at 10:01:03AM -0500, James Bottomley
> > > > wrote:
> > > > > On Mon, 2017-01-16 at 15:12 +0200, Jarkko Sakkinen wrote:
> > > > > > From: James Bottomley <
> > > > > > James.Bottomley@HansenPartnership.com>
> > > > > > 
> > > > > > Currently the Resource Manager (RM) is not exposed to
> > > > > > userspace. 
> > > > > >  Make this exposure via a separate device, which can now be
> > > > > > opened multiple times because each read/write transaction
> > > > > > goes 
> > > > > > separately via the RM.
> > > > > > 
> > > > > > Concurrency is protected by the chip->tpm_mutex for each 
> > > > > > read/write transaction separately.  The TPM is cleared of
> > > > > > all 
> > > > > > transient objects by the time the mutex is dropped, so
> > > > > > there 
> > > > > > should be no interference between the kernel and userspace.
> > > > > 
> > > > > There's actually a missing kfree of context_buf on the
> > > > > tpms_release
> > > > > path as well.  This patch fixes it up.
> > > > 
> > > > Can you send me a fresh version of the whole patch so that I
> > > > can 
> > > > include to v4 that includes also changes that I requested in my
> > > > recent comments + all the fixes?
> > > 
> > > Sure, I think the attached is basically it
> > > 
> > > James
> > 
> > Thank you!
> 
> 'tabrm4' branch has been now rebased. It's now on top of master 
> branch that contains Stefan's latest patch (min body length check) 
> that I've reviewed and tested. It also contains your updated
> /dev/tpms patch.
> 
> I guess the 5 commits that are there now are such that we have fairly
> good consensus, don't we? If so, can I add your reviewed-by and
> tested-by to my commits and vice versa?

It looks like there's another problem: you need a continue after the
transient object is garbage collected otherwise the code falls through,
does a flush which fails and then adds a ~0 as the handle meaning we'll
have a mismatch between the saved contexts and the handles.

James

---

commit 0da3f83ce889379bd1741a11b07a30818a223924
Author: James Bottomley <James.Bottomley@HansenPartnership.com>
Date:   Sat Jan 21 12:19:06 2017 -0800

    continue after lazy reclaim

diff --git a/drivers/char/tpm/tpm2-space.c b/drivers/char/tpm/tpm2-space.c
index 8713d7f..9d87537 100644
--- a/drivers/char/tpm/tpm2-space.c
+++ b/drivers/char/tpm/tpm2-space.c
@@ -288,9 +288,10 @@ static int tpm2_save_space(struct tpm_chip *chip)
 		rc = tpm2_save_context(chip, space->context_tbl[i],
 				       space->context_buf, PAGE_SIZE,
 				       &offset);
-		if (rc == -ENOENT)
+		if (rc == -ENOENT) {
 			space->context_tbl[i] = 0;
-		else if (rc) {
+			continue;
+		} else if (rc) {
 			tpm2_flush_space(chip);
 			return rc;
 		}