[PATCH] tpm: fix race condition in tpm_common_write()

[PATCH] tpm: fix race condition in tpm_common_write()

[Tadeusz Struk]

There is a race condition in tpm_common_write function allowing two
threads on the same /dev/tpm<N>, or two different applications on
the same /dev/tpmrm<N> to overwrite eachother requests/responses.

Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
---
 drivers/char/tpm/tpm-dev-common.c |   14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/drivers/char/tpm/tpm-dev-common.c b/drivers/char/tpm/tpm-dev-common.c
index 230b99288024..1ce6f989f40a 100644
--- a/drivers/char/tpm/tpm-dev-common.c
+++ b/drivers/char/tpm/tpm-dev-common.c
@@ -91,17 +91,19 @@ ssize_t tpm_common_write(struct file *file, const char __user *buf,
 	size_t in_size = size;
 	ssize_t out_size;
 
+	if (in_size > TPM_BUFSIZE)
+		return -E2BIG;
+
+	mutex_lock(&priv->buffer_mutex);
+
 	/* Cannot perform a write until the read has cleared either via
 	 * tpm_read or a user_read_timer timeout. This also prevents split
 	 * buffered writes from blocking here.
 	 */
-	if (atomic_read(&priv->data_pending) != 0)
+	if (atomic_read(&priv->data_pending) != 0) {
+		mutex_unlock(&priv->buffer_mutex);
 		return -EBUSY;
-
-	if (in_size > TPM_BUFSIZE)
-		return -E2BIG;
-
-	mutex_lock(&priv->buffer_mutex);
+	}
 
 	if (copy_from_user
 	    (priv->data_buffer, (void __user *) buf, in_size)) {

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Jason Gunthorpe]

On Tue, May 22, 2018 at 10:32:46AM -0700, Tadeusz Struk wrote:
> There is a race condition in tpm_common_write function allowing two
> threads on the same /dev/tpm<N>, or two different applications on
> the same /dev/tpmrm<N> to overwrite eachother requests/responses.
> 
> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
> ---
>  drivers/char/tpm/tpm-dev-common.c |   14 ++++++++------
>  1 file changed, 8 insertions(+), 6 deletions(-)

I didn't see any reasn for data_pending to be an atomic, ever use case
is near the buffer_mutex, can you respin this patch to just drop that
completely and only manipulate it within the lock?

Jason

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Tadeusz Struk]

On 05/22/2018 01:07 PM, Jason Gunthorpe wrote:
> I didn't see any reasn for data_pending to be an atomic, ever use case
> is near the buffer_mutex, can you respin this patch to just drop that
> completely and only manipulate it within the lock?

Yes, will do.
Thanks,
-- 
Tadeusz

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Jarkko Sakkinen]

On Tue, May 22, 2018 at 10:32:46AM -0700, Tadeusz Struk wrote:
> There is a race condition in tpm_common_write function allowing two
> threads on the same /dev/tpm<N>, or two different applications on
> the same /dev/tpmrm<N> to overwrite eachother requests/responses.
> 
> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>

Ouch o_O Do you have a fixes tag for this one?

Thank you.

Reviewed-by: Jarkko Sakkien <jarkko.sakkinen@linux.intel.com>

/Jarkko

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Tadeusz Struk]

On 05/23/2018 06:23 AM, Jarkko Sakkinen wrote:
> Ouch o_O Do you have a fixes tag for this one?
> 

This one is quite tricky.
The original bug was introduced by abce9ac292e13 (tpm: Propagate error from tpm_transmit to fix a timeout hang)
and the code back then was in drivers/char/tpm/tpm-interface.c file

Then there were two other commits that moved the code around:
afdba32e2a9ea (tpm: Pull everything related to /dev/tpmX into tpm-dev.c)
which moved it from drivers/char/tpm/tpm-interface.c into drivers/char/tpm/tpm-dev.c

and last one, ecb38e2f521b0 (tpm: split out tpm-dev.c into tpm-dev.c and tpm-common-dev.c)
which moved it from drivers/char/tpm/tpm-dev.c into tpm-common-dev.c

I have no idea how to tag it. Maybe we can use:
Fixes: ecb38e2f521b0 ("tpm: split out tpm-dev.c into tpm-dev.c and tpm-common-dev.c")

And then it probably needs to be back ported manually all the way back to abce9ac292e13.

Thanks,
-- 
Tadeusz

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Jason Gunthorpe]

On Wed, May 23, 2018 at 10:57:07AM -0700, Tadeusz Struk wrote:
> On 05/23/2018 06:23 AM, Jarkko Sakkinen wrote:
> > Ouch o_O Do you have a fixes tag for this one?
> > 
> 
> This one is quite tricky.
> The original bug was introduced by abce9ac292e13 (tpm: Propagate error from tpm_transmit to fix a timeout hang)
> and the code back then was in drivers/char/tpm/tpm-interface.c file

No, it has been wrong since before git history started, so just use

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")

All the other commits you listed are just moving the bad code around
to new files.

Jason

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Jarkko Sakkinen]

On Wed, May 23, 2018 at 10:57:07AM -0700, Tadeusz Struk wrote:
> On 05/23/2018 06:23 AM, Jarkko Sakkinen wrote:
> > Ouch o_O Do you have a fixes tag for this one?
> > 
> 
> This one is quite tricky.
> The original bug was introduced by abce9ac292e13 (tpm: Propagate error from tpm_transmit to fix a timeout hang)
> and the code back then was in drivers/char/tpm/tpm-interface.c file
> 
> Then there were two other commits that moved the code around:
> afdba32e2a9ea (tpm: Pull everything related to /dev/tpmX into tpm-dev.c)
> which moved it from drivers/char/tpm/tpm-interface.c into drivers/char/tpm/tpm-dev.c
> 
> and last one, ecb38e2f521b0 (tpm: split out tpm-dev.c into tpm-dev.c and tpm-common-dev.c)
> which moved it from drivers/char/tpm/tpm-dev.c into tpm-common-dev.c
> 
> I have no idea how to tag it. Maybe we can use:
> Fixes: ecb38e2f521b0 ("tpm: split out tpm-dev.c into tpm-dev.c and tpm-common-dev.c")
> 
> And then it probably needs to be back ported manually all the way back to abce9ac292e13.
> 
> Thanks,
> -- 
> Tadeusz

Thank you. I'll cc this to stable.

/Jarkko

Re: [PATCH] tpm: fix race condition in tpm_common_write()

[Jarkko Sakkinen]

On Wed, May 23, 2018 at 01:41:15PM -0600, Jason Gunthorpe wrote:
> On Wed, May 23, 2018 at 10:57:07AM -0700, Tadeusz Struk wrote:
> > On 05/23/2018 06:23 AM, Jarkko Sakkinen wrote:
> > > Ouch o_O Do you have a fixes tag for this one?
> > > 
> > 
> > This one is quite tricky.
> > The original bug was introduced by abce9ac292e13 (tpm: Propagate error from tpm_transmit to fix a timeout hang)
> > and the code back then was in drivers/char/tpm/tpm-interface.c file
> 
> No, it has been wrong since before git history started, so just use
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> 
> All the other commits you listed are just moving the bad code around
> to new files.
> 
> Jason

OK, thank you Jason. I'll use that.

/Jarkko