Re: [PATCH v3 2/2] tpm: in tpm2_del_space check if ops pointer is still valid

[Jarkko Sakkinen]

On Thu, Feb 04, 2021 at 04:34:11PM -0800, James Bottomley wrote:
> On Fri, 2021-02-05 at 00:50 +0100, Lino Sanfilippo wrote:
> > From: Lino Sanfilippo <l.sanfilippo@xxxxxxxxxx>
> > 
> > In tpm2_del_space() chip->ops is used for flushing the sessions.
> > However
> > this function may be called after tpm_chip_unregister() which sets
> > the chip->ops pointer to NULL.
> > Avoid a possible NULL pointer dereference by checking if chip->ops is
> > still
> > valid before accessing it.
> > 
> > Fixes: a3fbfae82b4c ("tpm: take TPM chip power gating out of
> > tpm_transmit()")
> > Signed-off-by: Lino Sanfilippo <l.sanfilippo@xxxxxxxxxx>
> > ---
> >  drivers/char/tpm/tpm2-space.c | 15 ++++++++++-----
> >  1 file changed, 10 insertions(+), 5 deletions(-)
> > 
> > diff --git a/drivers/char/tpm/tpm2-space.c b/drivers/char/tpm/tpm2-
> > space.c
> > index 784b8b3..9a29a40 100644
> > --- a/drivers/char/tpm/tpm2-space.c
> > +++ b/drivers/char/tpm/tpm2-space.c
> > @@ -58,12 +58,17 @@ int tpm2_init_space(struct tpm_space *space,
> > unsigned int buf_size)
> >  
> >  void tpm2_del_space(struct tpm_chip *chip, struct tpm_space *space)
> >  {
> > -	mutex_lock(&chip->tpm_mutex);
> > -	if (!tpm_chip_start(chip)) {
> > -		tpm2_flush_sessions(chip, space);
> > -		tpm_chip_stop(chip);
> > +	down_read(&chip->ops_sem);
> > +	if (chip->ops) {
> > +		mutex_lock(&chip->tpm_mutex);
> > +		if (!tpm_chip_start(chip)) {
> > +			tpm2_flush_sessions(chip, space);
> > +			tpm_chip_stop(chip);
> > +		}
> > +		mutex_unlock(&chip->tpm_mutex);
> >  	}
> > -	mutex_unlock(&chip->tpm_mutex);
> > +	up_read(&chip->ops_sem);
> > +
> >  	kfree(space->context_buf);
> >  	kfree(space->session_buf);
> >  }
> 
> 
> Actually, this still isn't right.  As I said to the last person who
> reported this, we should be doing a get/put on the ops, not rolling our
> own here:
> 
> https://lore.kernel.org/linux-integrity/e7566e1e48f5be9dca034b4bfb67683b5d3cb88f.camel@xxxxxxxxxxxxxxxxxxxxx/
> 
> The reporter went silent before we could get this tested, but could you
> try, please, because your patch is still hand rolling the ops get/put,
> just slightly better than it had been done previously.
> 
> James

Thanks for pointing this out. I'd strongly support Jason's proposal:

https://lore.kernel.org/linux-integrity/20201215175624.GG5487@xxxxxxxx/

It's the best long-term way to fix this.

Honestly, I have to admit that this thread leaked from me. It happened
exactly at the time when I was on vacation. Something must have gone wrong
when I pulled emails after that.

/Jarkko