[PATCH v2] tpm: fix potential NULL pointer access in tpm_del_char_device

Some SPI controller drivers unregister the controller in the shutdown
handler (e.g. BCM2835). If such a controller is used with a TPM 2 slave
chip->ops may be accessed when it is already NULL:

At system shutdown the pre-shutdown handler tpm_class_shutdown() shuts down
TPM 2 and sets chip->ops to NULL. Then at SPI controller unregistration
tpm_tis_spi_remove() is called and eventually calls tpm_del_char_device()
which tries to shut down TPM 2 again. Thereby it accesses chip->ops again:
(tpm_del_char_device calls tpm_chip_start which calls tpm_clk_enable which
calls chip->ops->clk_enable).

Avoid the NULL pointer access by testing if chip->ops is valid and skipping
the TPM 2 shutdown procedure in case it is NULL.

Fixes: dcbeab1946454 ("tpm: fix crash in tpm_tis deinitialization")
Cc: stable@vger.kernel.org
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
---

Changes to v2:
- rephrased the commit message to clarify the circumstances under which
  this bug triggers (as requested by Jarkko)


I was able to reproduce this issue with a SLB 9670 TPM chip controlled by 
a BCM2835 SPI controller. 

The approach to fix this issue in the BCM2835 driver was rejected after a
discussion on the mailing list:

https://marc.info/?l=linux-integrity&m=163285906725367&w=2

The reason for the rejection was the realization, that this issue should rather
be fixed in the TPM code:

https://marc.info/?l=linux-spi&m=163311087423271&w=2

So this is the reworked version of a patch that is supposed to do that.


 drivers/char/tpm/tpm-chip.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
index ddaeceb7e109..7960da490e72 100644
--- a/drivers/char/tpm/tpm-chip.c
+++ b/drivers/char/tpm/tpm-chip.c

@@ -474,13 +474,19 @@ static void tpm_del_char_device(struct tpm_chip *chip)
 
 	/* Make the driver uncallable. */
 	down_write(&chip->ops_sem);
-	if (chip->flags & TPM_CHIP_FLAG_TPM2) {
-		if (!tpm_chip_start(chip)) {
-			tpm2_shutdown(chip, TPM2_SU_CLEAR);
-			tpm_chip_stop(chip);
+	/* Check if chip->ops is still valid: In case that the controller
+	 * drivers shutdown handler unregisters the controller in its
+	 * shutdown handler we are called twice and chip->ops to NULL.
+	 */
+	if (chip->ops) {
+		if (chip->flags & TPM_CHIP_FLAG_TPM2) {
+			if (!tpm_chip_start(chip)) {
+				tpm2_shutdown(chip, TPM2_SU_CLEAR);
+				tpm_chip_stop(chip);
+			}
 		}
+		chip->ops = NULL;
 	}
-	chip->ops = NULL;
 	up_write(&chip->ops_sem);
 }
 

base-commit: a7904a538933c525096ca2ccde1e60d0ee62c08e
-- 
2.34.1

Hi,

the patch below should also fix the issue that Stefan reported here:

https://marc.info/?l=linux-integrity&m=163927245509564&w=2


Regards,
Lino


On 20.12.21 at 16:06, Lino Sanfilippo wrote:

On 12/20/21 10:06, Lino Sanfilippo wrote:

Fixes: 39d0099f9439 ("powerpc/pseries: Add shutdown() to vio_driver and 
vio_bus")

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Tested-by: Stefan Berger <stefanb@linux.ibm.com>

Hi Stefan,


On 22.12.21 at 05:53, Stefan Berger wrote:

quoted

  drivers/char/tpm/tpm-chip.c | 16 +++++++++++-----
  1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
index ddaeceb7e109..7960da490e72 100644
--- a/drivers/char/tpm/tpm-chip.c
+++ b/drivers/char/tpm/tpm-chip.c

@@ -474,13 +474,19 @@ static void tpm_del_char_device(struct tpm_chip *chip)

        /* Make the driver uncallable. */
      down_write(&chip->ops_sem);
-    if (chip->flags & TPM_CHIP_FLAG_TPM2) {
-        if (!tpm_chip_start(chip)) {
-            tpm2_shutdown(chip, TPM2_SU_CLEAR);
-            tpm_chip_stop(chip);
+    /* Check if chip->ops is still valid: In case that the controller
+     * drivers shutdown handler unregisters the controller in its
+     * shutdown handler we are called twice and chip->ops to NULL.
+     */
+    if (chip->ops) {
+        if (chip->flags & TPM_CHIP_FLAG_TPM2) {
+            if (!tpm_chip_start(chip)) {
+                tpm2_shutdown(chip, TPM2_SU_CLEAR);
+                tpm_chip_stop(chip);
+            }
          }
+        chip->ops = NULL;
      }
-    chip->ops = NULL;
      up_write(&chip->ops_sem);
  }
 
base-commit: a7904a538933c525096ca2ccde1e60d0ee62c08e

Fixes: 39d0099f9439 ("powerpc/pseries: Add shutdown() to vio_driver and vio_bus")

Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>

Tested-by: Stefan Berger <stefanb@linux.ibm.com>

Thanks a lot for testing this.

Best regards,
Lino

On Mon, Dec 20, 2021 at 04:06:35PM +0100, Lino Sanfilippo wrote:

Some SPI controller drivers unregister the controller in the shutdown
handler (e.g. BCM2835). If such a controller is used with a TPM 2 slave
chip->ops may be accessed when it is already NULL:

At system shutdown the pre-shutdown handler tpm_class_shutdown() shuts down
TPM 2 and sets chip->ops to NULL. Then at SPI controller unregistration
tpm_tis_spi_remove() is called and eventually calls tpm_del_char_device()
which tries to shut down TPM 2 again. Thereby it accesses chip->ops again:
(tpm_del_char_device calls tpm_chip_start which calls tpm_clk_enable which
calls chip->ops->clk_enable).

Avoid the NULL pointer access by testing if chip->ops is valid and skipping
the TPM 2 shutdown procedure in case it is NULL.

Fixes: dcbeab1946454 ("tpm: fix crash in tpm_tis deinitialization")
Cc: stable@vger.kernel.org
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>

Thank you.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR,
Jarkko

Hi,

On 29.12.21 at 01:13, Jarkko Sakkinen wrote:

On Mon, Dec 20, 2021 at 04:06:35PM +0100, Lino Sanfilippo wrote:

quoted

Some SPI controller drivers unregister the controller in the shutdown
handler (e.g. BCM2835). If such a controller is used with a TPM 2 slave
chip->ops may be accessed when it is already NULL:

At system shutdown the pre-shutdown handler tpm_class_shutdown() shuts down
TPM 2 and sets chip->ops to NULL. Then at SPI controller unregistration
tpm_tis_spi_remove() is called and eventually calls tpm_del_char_device()
which tries to shut down TPM 2 again. Thereby it accesses chip->ops again:
(tpm_del_char_device calls tpm_chip_start which calls tpm_clk_enable which
calls chip->ops->clk_enable).

Avoid the NULL pointer access by testing if chip->ops is valid and skipping
the TPM 2 shutdown procedure in case it is NULL.

Fixes: dcbeab1946454 ("tpm: fix crash in tpm_tis deinitialization")
Cc: stable@vger.kernel.org
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>

Thank you.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR,
Jarkko

Thanks a lot for the review. Please note that the latest version is v3 which
contains one more Fixes tag and also a tag for the review by Stefan. I also
adjusted the source code comment in that version which I somehow messed up in v2.

Regards,
Lino

On Wed, Dec 29, 2021 at 02:41:52AM +0100, Lino Sanfilippo wrote:

Hi,

On 29.12.21 at 01:13, Jarkko Sakkinen wrote:

quoted

On Mon, Dec 20, 2021 at 04:06:35PM +0100, Lino Sanfilippo wrote:

quoted

Some SPI controller drivers unregister the controller in the shutdown
handler (e.g. BCM2835). If such a controller is used with a TPM 2 slave
chip->ops may be accessed when it is already NULL:

At system shutdown the pre-shutdown handler tpm_class_shutdown() shuts down
TPM 2 and sets chip->ops to NULL. Then at SPI controller unregistration
tpm_tis_spi_remove() is called and eventually calls tpm_del_char_device()
which tries to shut down TPM 2 again. Thereby it accesses chip->ops again:
(tpm_del_char_device calls tpm_chip_start which calls tpm_clk_enable which
calls chip->ops->clk_enable).

Avoid the NULL pointer access by testing if chip->ops is valid and skipping
the TPM 2 shutdown procedure in case it is NULL.

Fixes: dcbeab1946454 ("tpm: fix crash in tpm_tis deinitialization")
Cc: stable@vger.kernel.org
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>

Thank you.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR,
Jarkko

Thanks a lot for the review. Please note that the latest version is v3 which
contains one more Fixes tag and also a tag for the review by Stefan. I also
adjusted the source code comment in that version which I somehow messed up in v2.

Since there is no functional difference, I rather do not swap it.

I fixed a glitch:

+
+       /*

A multi-line commit needs to have this as the very first line.
You can check if the all tags were pulled by b4.

/Jarkko

On 29.12.21 at 02:46, Jarkko Sakkinen wrote:

Since there is no functional difference, I rather do not swap it.

I fixed a glitch:

+
+       /*

A multi-line commit needs to have this as the very first line.
You can check if the all tags were pulled by b4.

/Jarkko

I see, thanks for fixing the comment then.

Regards,
Lino