[tpmdd-devel,v4,1/2] tpm: Issue a TPM2_Shutdown for TPM2 devices.

Message ID CAHSjozDnf5Nm9Nw=kKBQRRYYmEozT-m=XN-bxwLbk8Rs+=pduA@mail.gmail.com
State New
Headers show

Commit Message

Josh Zimmerman May 25, 2017, 11:20 p.m.
If a TPM2 loses power without a TPM2_Shutdown command being issued (a
"disorderly reboot"), it may lose some state that has yet to be
persisted to NVRam, and will increment the DA counter. After the DA
counter gets sufficiently large, the TPM will lock the user out.

NOTE: This only changes behavior on TPM2 devices. Since TPM1 uses sysfs,
and sysfs relies on implicit locking on chip->ops, it is not safe to
allow this code to run in TPM1, or to add sysfs support to TPM2, until
that locking is made explicit.

Signed-off-by: Josh Zimmerman <joshz@google.com>
Reviewed-by: Jarko Sakkinen <jarkko.sakkinen@linux.intel.com>
Cc: stable@vger.kernel.org

----
v2:
  - Properly split changes between this and another commit
  - Use proper locking primitive.
  - Fix commenting style
v3:
  - Re-fix commenting style
v4:
  - Update description and tags (Reviewed-by, Cc).
---
---
 drivers/char/tpm/tpm-chip.c  | 20 ++++++++++++++++++++
 drivers/char/tpm/tpm-sysfs.c |  3 +++
 2 files changed, 23 insertions(+)

--
2.13.0.219.gdb65acc882-goog

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot

Comments

Jarkko Sakkinen May 30, 2017, 5:07 a.m. | #1
On Thu, May 25, 2017 at 04:20:28PM -0700, Josh Zimmerman wrote:
> If a TPM2 loses power without a TPM2_Shutdown command being issued (a
> "disorderly reboot"), it may lose some state that has yet to be
> persisted to NVRam, and will increment the DA counter. After the DA
> counter gets sufficiently large, the TPM will lock the user out.
> 
> NOTE: This only changes behavior on TPM2 devices. Since TPM1 uses sysfs,
> and sysfs relies on implicit locking on chip->ops, it is not safe to
> allow this code to run in TPM1, or to add sysfs support to TPM2, until
> that locking is made explicit.
> 
> Signed-off-by: Josh Zimmerman <joshz@google.com>
> Reviewed-by: Jarko Sakkinen <jarkko.sakkinen@linux.intel.com>
> Cc: stable@vger.kernel.org

Still have some remarks.

> ----
> v2:
>   - Properly split changes between this and another commit
>   - Use proper locking primitive.
>   - Fix commenting style
> v3:
>   - Re-fix commenting style
> v4:
>   - Update description and tags (Reviewed-by, Cc).
> ---
> ---
>  drivers/char/tpm/tpm-chip.c  | 20 ++++++++++++++++++++
>  drivers/char/tpm/tpm-sysfs.c |  3 +++
>  2 files changed, 23 insertions(+)
> 
> diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
> index 9dec9f551b83..272a42e77574 100644
> --- a/drivers/char/tpm/tpm-chip.c
> +++ b/drivers/char/tpm/tpm-chip.c
> @@ -142,6 +142,25 @@ static void tpm_devs_release(struct device *dev)
>         put_device(&chip->dev);
>  }
> 
> +static void tpm_shutdown(struct device *dev)
> +{
> +       struct tpm_chip *chip = container_of(dev, struct tpm_chip, dev);
> +       /* TPM 2.0 requires that the TPM2_Shutdown() command be issued prior to
> +        * loss of power. If it is not, the DA counter will be incremented and,
> +        * eventually, the user will be locked out of their TPM.
> +        * XXX: This codepath relies on the fact that sysfs is not enabled for
> +        * TPM2: sysfs uses an implicit lock on chip->ops, so this use could
> +        * race if TPM2 has sysfs support enabled before TPM sysfs's implicit
> +        * locking is fixed.
> +        */

The comment should be either deleted or a kdoc. 

> +       if (chip->flags & TPM_CHIP_FLAG_TPM2) {
> +               down_write(&chip->ops_sem);
> +               tpm2_shutdown(chip, TPM_SU_CLEAR);
> +               chip->ops = NULL;
> +               up_write(&chip->ops_sem);
> +       }
> +}

Would be a better idea to rename tpm2_shutdown as tpm_shutdown and call
it unconditionally in tpm_del_char_device.



> +
>  /**
>   * tpm_chip_alloc() - allocate a new struct tpm_chip instance
>   * @pdev: device to which the chip is associated
> @@ -181,6 +200,7 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>         device_initialize(&chip->devs);
> 
>         chip->dev.class = tpm_class;
> +       chip->dev.class.shutdown = tpm_shutdown;
>         chip->dev.release = tpm_dev_release;
>         chip->dev.parent = pdev;
>         chip->dev.groups = chip->groups;
> diff --git a/drivers/char/tpm/tpm-sysfs.c b/drivers/char/tpm/tpm-sysfs.c
> index 55405dbe43fa..5e5ff7eb6f7e 100644
> --- a/drivers/char/tpm/tpm-sysfs.c
> +++ b/drivers/char/tpm/tpm-sysfs.c
> @@ -294,6 +294,9 @@ static const struct attribute_group tpm_dev_group = {
> 
>  void tpm_sysfs_add_device(struct tpm_chip *chip)
>  {
> +       /* XXX: Before this restriction is removed, tpm_sysfs must be updated
> +        * to explicitly lock chip->ops.
> +        */

Not sure about this remark. Most, if not all, attributes in tpm-sysfs.c
are useless attributes as you can use /dev/tpm0 to retrieve their
values.

>         if (chip->flags & TPM_CHIP_FLAG_TPM2)
>                 return;
> 
> --
> 2.13.0.219.gdb65acc882-goog

/Jarkko

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Josh Zimmerman May 30, 2017, 7 p.m. | #2
On Mon, May 29, 2017 at 10:07 PM, Jarkko Sakkinen
<jarkko.sakkinen@linux.intel.com> wrote:
> On Thu, May 25, 2017 at 04:20:28PM -0700, Josh Zimmerman wrote:
>> If a TPM2 loses power without a TPM2_Shutdown command being issued (a
>> "disorderly reboot"), it may lose some state that has yet to be
>> persisted to NVRam, and will increment the DA counter. After the DA
>> counter gets sufficiently large, the TPM will lock the user out.
>>
>> NOTE: This only changes behavior on TPM2 devices. Since TPM1 uses sysfs,
>> and sysfs relies on implicit locking on chip->ops, it is not safe to
>> allow this code to run in TPM1, or to add sysfs support to TPM2, until
>> that locking is made explicit.
>>
>> Signed-off-by: Josh Zimmerman <joshz@google.com>
>> Reviewed-by: Jarko Sakkinen <jarkko.sakkinen@linux.intel.com>
>> Cc: stable@vger.kernel.org
>
> Still have some remarks.
>
>> ----
>> v2:
>>   - Properly split changes between this and another commit
>>   - Use proper locking primitive.
>>   - Fix commenting style
>> v3:
>>   - Re-fix commenting style
>> v4:
>>   - Update description and tags (Reviewed-by, Cc).
>> ---
>> ---
>>  drivers/char/tpm/tpm-chip.c  | 20 ++++++++++++++++++++
>>  drivers/char/tpm/tpm-sysfs.c |  3 +++
>>  2 files changed, 23 insertions(+)
>>
>> diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
>> index 9dec9f551b83..272a42e77574 100644
>> --- a/drivers/char/tpm/tpm-chip.c
>> +++ b/drivers/char/tpm/tpm-chip.c
>> @@ -142,6 +142,25 @@ static void tpm_devs_release(struct device *dev)
>>         put_device(&chip->dev);
>>  }
>>
>> +static void tpm_shutdown(struct device *dev)
>> +{
>> +       struct tpm_chip *chip = container_of(dev, struct tpm_chip, dev);
>> +       /* TPM 2.0 requires that the TPM2_Shutdown() command be issued prior to
>> +        * loss of power. If it is not, the DA counter will be incremented and,
>> +        * eventually, the user will be locked out of their TPM.
>> +        * XXX: This codepath relies on the fact that sysfs is not enabled for
>> +        * TPM2: sysfs uses an implicit lock on chip->ops, so this use could
>> +        * race if TPM2 has sysfs support enabled before TPM sysfs's implicit
>> +        * locking is fixed.
>> +        */
>
> The comment should be either deleted or a kdoc.
Done.

>> +       if (chip->flags & TPM_CHIP_FLAG_TPM2) {
>> +               down_write(&chip->ops_sem);
>> +               tpm2_shutdown(chip, TPM_SU_CLEAR);
>> +               chip->ops = NULL;
>> +               up_write(&chip->ops_sem);
>> +       }
>> +}
>
> Would be a better idea to rename tpm2_shutdown as tpm_shutdown and call
> it unconditionally in tpm_del_char_device.
I'm not sure quite what you mean here. Are you suggesting that
tpm_del_char_device should unconditionally call the tpm_shutdown that
this patch introduces?  Or that the tpm2_shutdown function from
drivers/char/tpm/tpm2-cmd.c (which right now just sends the
TPM2_Shutdown command) be renamed to tpm_shutdown?

>> +
>>  /**
>>   * tpm_chip_alloc() - allocate a new struct tpm_chip instance
>>   * @pdev: device to which the chip is associated
>> @@ -181,6 +200,7 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>>         device_initialize(&chip->devs);
>>
>>         chip->dev.class = tpm_class;
>> +       chip->dev.class.shutdown = tpm_shutdown;
>>         chip->dev.release = tpm_dev_release;
>>         chip->dev.parent = pdev;
>>         chip->dev.groups = chip->groups;
>> diff --git a/drivers/char/tpm/tpm-sysfs.c b/drivers/char/tpm/tpm-sysfs.c
>> index 55405dbe43fa..5e5ff7eb6f7e 100644
>> --- a/drivers/char/tpm/tpm-sysfs.c
>> +++ b/drivers/char/tpm/tpm-sysfs.c
>> @@ -294,6 +294,9 @@ static const struct attribute_group tpm_dev_group = {
>>
>>  void tpm_sysfs_add_device(struct tpm_chip *chip)
>>  {
>> +       /* XXX: Before this restriction is removed, tpm_sysfs must be updated
>> +        * to explicitly lock chip->ops.
>> +        */
>
> Not sure about this remark. Most, if not all, attributes in tpm-sysfs.c
> are useless attributes as you can use /dev/tpm0 to retrieve their
> values.
This is again in reference to
https://patchwork.kernel.org/patch/9516631/; if at some point in the
future a developer wishes to enable sysfs support for TPM2.0, the
implicit locking must be fixed.

I've attempted to clarify the phrasing here.

Josh

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Jarkko Sakkinen May 31, 2017, 12:01 p.m. | #3
On Tue, May 30, 2017 at 12:00:53PM -0700, Josh Zimmerman wrote:
> On Mon, May 29, 2017 at 10:07 PM, Jarkko Sakkinen
> <jarkko.sakkinen@linux.intel.com> wrote:
> > On Thu, May 25, 2017 at 04:20:28PM -0700, Josh Zimmerman wrote:
> >> If a TPM2 loses power without a TPM2_Shutdown command being issued (a
> >> "disorderly reboot"), it may lose some state that has yet to be
> >> persisted to NVRam, and will increment the DA counter. After the DA
> >> counter gets sufficiently large, the TPM will lock the user out.
> >>
> >> NOTE: This only changes behavior on TPM2 devices. Since TPM1 uses sysfs,
> >> and sysfs relies on implicit locking on chip->ops, it is not safe to
> >> allow this code to run in TPM1, or to add sysfs support to TPM2, until
> >> that locking is made explicit.
> >>
> >> Signed-off-by: Josh Zimmerman <joshz@google.com>
> >> Reviewed-by: Jarko Sakkinen <jarkko.sakkinen@linux.intel.com>
> >> Cc: stable@vger.kernel.org
> >
> > Still have some remarks.
> >
> >> ----
> >> v2:
> >>   - Properly split changes between this and another commit
> >>   - Use proper locking primitive.
> >>   - Fix commenting style
> >> v3:
> >>   - Re-fix commenting style
> >> v4:
> >>   - Update description and tags (Reviewed-by, Cc).
> >> ---
> >> ---
> >>  drivers/char/tpm/tpm-chip.c  | 20 ++++++++++++++++++++
> >>  drivers/char/tpm/tpm-sysfs.c |  3 +++
> >>  2 files changed, 23 insertions(+)
> >>
> >> diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
> >> index 9dec9f551b83..272a42e77574 100644
> >> --- a/drivers/char/tpm/tpm-chip.c
> >> +++ b/drivers/char/tpm/tpm-chip.c
> >> @@ -142,6 +142,25 @@ static void tpm_devs_release(struct device *dev)
> >>         put_device(&chip->dev);
> >>  }
> >>
> >> +static void tpm_shutdown(struct device *dev)
> >> +{
> >> +       struct tpm_chip *chip = container_of(dev, struct tpm_chip, dev);
> >> +       /* TPM 2.0 requires that the TPM2_Shutdown() command be issued prior to
> >> +        * loss of power. If it is not, the DA counter will be incremented and,
> >> +        * eventually, the user will be locked out of their TPM.
> >> +        * XXX: This codepath relies on the fact that sysfs is not enabled for
> >> +        * TPM2: sysfs uses an implicit lock on chip->ops, so this use could
> >> +        * race if TPM2 has sysfs support enabled before TPM sysfs's implicit
> >> +        * locking is fixed.
> >> +        */
> >
> > The comment should be either deleted or a kdoc.
> Done.
> 
> >> +       if (chip->flags & TPM_CHIP_FLAG_TPM2) {
> >> +               down_write(&chip->ops_sem);
> >> +               tpm2_shutdown(chip, TPM_SU_CLEAR);
> >> +               chip->ops = NULL;
> >> +               up_write(&chip->ops_sem);
> >> +       }
> >> +}
> >
> > Would be a better idea to rename tpm2_shutdown as tpm_shutdown and call
> > it unconditionally in tpm_del_char_device.
> I'm not sure quite what you mean here. Are you suggesting that
> tpm_del_char_device should unconditionally call the tpm_shutdown that
> this patch introduces?  Or that the tpm2_shutdown function from
> drivers/char/tpm/tpm2-cmd.c (which right now just sends the
> TPM2_Shutdown command) be renamed to tpm_shutdown?

The second option. In addition can make that your patch set applies to
security/next so I can merge both. I realized that the first patch does
not apply so that needs a resend too.

> >> +
> >>  /**
> >>   * tpm_chip_alloc() - allocate a new struct tpm_chip instance
> >>   * @pdev: device to which the chip is associated
> >> @@ -181,6 +200,7 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
> >>         device_initialize(&chip->devs);
> >>
> >>         chip->dev.class = tpm_class;
> >> +       chip->dev.class.shutdown = tpm_shutdown;
> >>         chip->dev.release = tpm_dev_release;
> >>         chip->dev.parent = pdev;
> >>         chip->dev.groups = chip->groups;
> >> diff --git a/drivers/char/tpm/tpm-sysfs.c b/drivers/char/tpm/tpm-sysfs.c
> >> index 55405dbe43fa..5e5ff7eb6f7e 100644
> >> --- a/drivers/char/tpm/tpm-sysfs.c
> >> +++ b/drivers/char/tpm/tpm-sysfs.c
> >> @@ -294,6 +294,9 @@ static const struct attribute_group tpm_dev_group = {
> >>
> >>  void tpm_sysfs_add_device(struct tpm_chip *chip)
> >>  {
> >> +       /* XXX: Before this restriction is removed, tpm_sysfs must be updated
> >> +        * to explicitly lock chip->ops.
> >> +        */
> >
> > Not sure about this remark. Most, if not all, attributes in tpm-sysfs.c
> > are useless attributes as you can use /dev/tpm0 to retrieve their
> > values.
> This is again in reference to
> https://patchwork.kernel.org/patch/9516631/; if at some point in the
> future a developer wishes to enable sysfs support for TPM2.0, the
> implicit locking must be fixed.
> 
> I've attempted to clarify the phrasing here.
> 
> Josh

OK lets keep it!

/Jarkko

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Josh Zimmerman May 31, 2017, 10:08 p.m. | #4
On Wed, May 31, 2017 at 5:01 AM, Jarkko Sakkinen
<jarkko.sakkinen@linux.intel.com> wrote:
> On Tue, May 30, 2017 at 12:00:53PM -0700, Josh Zimmerman wrote:
>> On Mon, May 29, 2017 at 10:07 PM, Jarkko Sakkinen
>> <jarkko.sakkinen@linux.intel.com> wrote:
>> > On Thu, May 25, 2017 at 04:20:28PM -0700, Josh Zimmerman wrote:
>> >> If a TPM2 loses power without a TPM2_Shutdown command being issued (a
>> >> "disorderly reboot"), it may lose some state that has yet to be
>> >> persisted to NVRam, and will increment the DA counter. After the DA
>> >> counter gets sufficiently large, the TPM will lock the user out.
>> >>
>> >> NOTE: This only changes behavior on TPM2 devices. Since TPM1 uses sysfs,
>> >> and sysfs relies on implicit locking on chip->ops, it is not safe to
>> >> allow this code to run in TPM1, or to add sysfs support to TPM2, until
>> >> that locking is made explicit.
>> >>
>> >> Signed-off-by: Josh Zimmerman <joshz@google.com>
>> >> Reviewed-by: Jarko Sakkinen <jarkko.sakkinen@linux.intel.com>
>> >> Cc: stable@vger.kernel.org
>> >
>> > Still have some remarks.
>> >
>> >> ----
>> >> v2:
>> >>   - Properly split changes between this and another commit
>> >>   - Use proper locking primitive.
>> >>   - Fix commenting style
>> >> v3:
>> >>   - Re-fix commenting style
>> >> v4:
>> >>   - Update description and tags (Reviewed-by, Cc).
>> >> ---
>> >> ---
>> >>  drivers/char/tpm/tpm-chip.c  | 20 ++++++++++++++++++++
>> >>  drivers/char/tpm/tpm-sysfs.c |  3 +++
>> >>  2 files changed, 23 insertions(+)
>> >>
>> >> diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
>> >> index 9dec9f551b83..272a42e77574 100644
>> >> --- a/drivers/char/tpm/tpm-chip.c
>> >> +++ b/drivers/char/tpm/tpm-chip.c
>> >> @@ -142,6 +142,25 @@ static void tpm_devs_release(struct device *dev)
>> >>         put_device(&chip->dev);
>> >>  }
>> >>
>> >> +static void tpm_shutdown(struct device *dev)
>> >> +{
>> >> +       struct tpm_chip *chip = container_of(dev, struct tpm_chip, dev);
>> >> +       /* TPM 2.0 requires that the TPM2_Shutdown() command be issued prior to
>> >> +        * loss of power. If it is not, the DA counter will be incremented and,
>> >> +        * eventually, the user will be locked out of their TPM.
>> >> +        * XXX: This codepath relies on the fact that sysfs is not enabled for
>> >> +        * TPM2: sysfs uses an implicit lock on chip->ops, so this use could
>> >> +        * race if TPM2 has sysfs support enabled before TPM sysfs's implicit
>> >> +        * locking is fixed.
>> >> +        */
>> >
>> > The comment should be either deleted or a kdoc.
>> Done.
>>
>> >> +       if (chip->flags & TPM_CHIP_FLAG_TPM2) {
>> >> +               down_write(&chip->ops_sem);
>> >> +               tpm2_shutdown(chip, TPM_SU_CLEAR);
>> >> +               chip->ops = NULL;
>> >> +               up_write(&chip->ops_sem);
>> >> +       }
>> >> +}
>> >
>> > Would be a better idea to rename tpm2_shutdown as tpm_shutdown and call
>> > it unconditionally in tpm_del_char_device.
>> I'm not sure quite what you mean here. Are you suggesting that
>> tpm_del_char_device should unconditionally call the tpm_shutdown that
>> this patch introduces?  Or that the tpm2_shutdown function from
>> drivers/char/tpm/tpm2-cmd.c (which right now just sends the
>> TPM2_Shutdown command) be renamed to tpm_shutdown?
>
> The second option.
I'm afraid I don't quite understand.  I believe that tpm2_shutdown is
currently quite specific to the TPM2 devices. It can also be called
when preparing for hibernation, in which case we may not want to NULL
out chip->ops. Can you please explain again what you'd like me to
accomplish by making this change?

> In addition can make that your patch set applies to
> security/next so I can merge both. I realized that the first patch does
> not apply so that needs a resend too.
Replied in the other thread.  This patch appears to apply cleanly on
the branch I mentioned there.

>
>> >> +
>> >>  /**
>> >>   * tpm_chip_alloc() - allocate a new struct tpm_chip instance
>> >>   * @pdev: device to which the chip is associated
>> >> @@ -181,6 +200,7 @@ struct tpm_chip *tpm_chip_alloc(struct device *pdev,
>> >>         device_initialize(&chip->devs);
>> >>
>> >>         chip->dev.class = tpm_class;
>> >> +       chip->dev.class.shutdown = tpm_shutdown;
>> >>         chip->dev.release = tpm_dev_release;
>> >>         chip->dev.parent = pdev;
>> >>         chip->dev.groups = chip->groups;
>> >> diff --git a/drivers/char/tpm/tpm-sysfs.c b/drivers/char/tpm/tpm-sysfs.c
>> >> index 55405dbe43fa..5e5ff7eb6f7e 100644
>> >> --- a/drivers/char/tpm/tpm-sysfs.c
>> >> +++ b/drivers/char/tpm/tpm-sysfs.c
>> >> @@ -294,6 +294,9 @@ static const struct attribute_group tpm_dev_group = {
>> >>
>> >>  void tpm_sysfs_add_device(struct tpm_chip *chip)
>> >>  {
>> >> +       /* XXX: Before this restriction is removed, tpm_sysfs must be updated
>> >> +        * to explicitly lock chip->ops.
>> >> +        */
>> >
>> > Not sure about this remark. Most, if not all, attributes in tpm-sysfs.c
>> > are useless attributes as you can use /dev/tpm0 to retrieve their
>> > values.
>> This is again in reference to
>> https://patchwork.kernel.org/patch/9516631/; if at some point in the
>> future a developer wishes to enable sysfs support for TPM2.0, the
>> implicit locking must be fixed.
>>
>> I've attempted to clarify the phrasing here.
>>
>> Josh
>
> OK lets keep it!
>
> /Jarkko

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot

Patch

diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
index 9dec9f551b83..272a42e77574 100644
--- a/drivers/char/tpm/tpm-chip.c
+++ b/drivers/char/tpm/tpm-chip.c
@@ -142,6 +142,25 @@  static void tpm_devs_release(struct device *dev)
        put_device(&chip->dev);
 }

+static void tpm_shutdown(struct device *dev)
+{
+       struct tpm_chip *chip = container_of(dev, struct tpm_chip, dev);
+       /* TPM 2.0 requires that the TPM2_Shutdown() command be issued prior to
+        * loss of power. If it is not, the DA counter will be incremented and,
+        * eventually, the user will be locked out of their TPM.
+        * XXX: This codepath relies on the fact that sysfs is not enabled for
+        * TPM2: sysfs uses an implicit lock on chip->ops, so this use could
+        * race if TPM2 has sysfs support enabled before TPM sysfs's implicit
+        * locking is fixed.
+        */
+       if (chip->flags & TPM_CHIP_FLAG_TPM2) {
+               down_write(&chip->ops_sem);
+               tpm2_shutdown(chip, TPM_SU_CLEAR);
+               chip->ops = NULL;
+               up_write(&chip->ops_sem);
+       }
+}
+
 /**
  * tpm_chip_alloc() - allocate a new struct tpm_chip instance
  * @pdev: device to which the chip is associated
@@ -181,6 +200,7 @@  struct tpm_chip *tpm_chip_alloc(struct device *pdev,
        device_initialize(&chip->devs);

        chip->dev.class = tpm_class;
+       chip->dev.class.shutdown = tpm_shutdown;
        chip->dev.release = tpm_dev_release;
        chip->dev.parent = pdev;
        chip->dev.groups = chip->groups;
diff --git a/drivers/char/tpm/tpm-sysfs.c b/drivers/char/tpm/tpm-sysfs.c
index 55405dbe43fa..5e5ff7eb6f7e 100644
--- a/drivers/char/tpm/tpm-sysfs.c
+++ b/drivers/char/tpm/tpm-sysfs.c
@@ -294,6 +294,9 @@  static const struct attribute_group tpm_dev_group = {

 void tpm_sysfs_add_device(struct tpm_chip *chip)
 {
+       /* XXX: Before this restriction is removed, tpm_sysfs must be updated
+        * to explicitly lock chip->ops.
+        */
        if (chip->flags & TPM_CHIP_FLAG_TPM2)
                return;