Patchwork [Maverick,CVE-2011-0695,1/2] IB/cm: Bump reference count on cm_id before invoking callback, CVE-2011-0695

login
register
mail settings
Submitter Brad Figg
Date April 25, 2011, 7:29 p.m.
Message ID <1303759770-15708-1-git-send-email-brad.figg@canonical.com>
Download mbox | patch
Permalink /patch/92775/
State New
Headers show

Comments

Brad Figg - April 25, 2011, 7:29 p.m.
From: Sean Hefty <sean.hefty@intel.com>

CVE-2011-0695

BugLink: http://bugs.launchpad.net/bugs/770369

When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
refcount of the cm_id is initialized to 1.  However, cm_process_work
will decrement the refcount after invoking all callbacks.  The result
is that the cm_id will end up with refcount set to 0 by the end of the
sidr req handler.

If a user tries to destroy the cm_id, the destruction will proceed,
under the incorrect assumption that no other threads are referencing
the cm_id.  This can lead to a crash when the cm callback thread tries
to access the cm_id.

This problem was noticed as part of a larger investigation with kernel
crashes in the rdma_cm when running on a real time OS.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Roland Dreier <roland@purestorage.com>

(cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
Signed-off-by: Brad Figg <brad.figg@canonical.com>
---
 drivers/infiniband/core/cm.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
Tim Gardner - April 25, 2011, 8:15 p.m.
On 04/25/2011 01:29 PM, Brad Figg wrote:
> From: Sean Hefty<sean.hefty@intel.com>
>
> CVE-2011-0695
>
> BugLink: http://bugs.launchpad.net/bugs/770369
>
> When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
> refcount of the cm_id is initialized to 1.  However, cm_process_work
> will decrement the refcount after invoking all callbacks.  The result
> is that the cm_id will end up with refcount set to 0 by the end of the
> sidr req handler.
>
> If a user tries to destroy the cm_id, the destruction will proceed,
> under the incorrect assumption that no other threads are referencing
> the cm_id.  This can lead to a crash when the cm callback thread tries
> to access the cm_id.
>
> This problem was noticed as part of a larger investigation with kernel
> crashes in the rdma_cm when running on a real time OS.
>
> Signed-off-by: Sean Hefty<sean.hefty@intel.com>
> Acked-by: Doug Ledford<dledford@redhat.com>
> Cc:<stable@kernel.org>
> Signed-off-by: Roland Dreier<roland@purestorage.com>
>
> (cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
> Signed-off-by: Brad Figg<brad.figg@canonical.com>
> ---
>   drivers/infiniband/core/cm.c |    1 +
>   1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
> index 64e0903..1d9616b 100644
> --- a/drivers/infiniband/core/cm.c
> +++ b/drivers/infiniband/core/cm.c
> @@ -2989,6 +2989,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
>   		goto out; /* No match. */
>   	}
>   	atomic_inc(&cur_cm_id_priv->refcount);
> +	atomic_inc(&cm_id_priv->refcount);
>   	spin_unlock_irq(&cm.lock);
>
>   	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;

Acked-by: Tim Gardner <tim.gardner@canonical.com>
Leann Ogasawara - April 25, 2011, 8:50 p.m.
On Mon, 2011-04-25 at 12:29 -0700, Brad Figg wrote:
> From: Sean Hefty <sean.hefty@intel.com>
> 
> CVE-2011-0695
> 
> BugLink: http://bugs.launchpad.net/bugs/770369
> 
> When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
> refcount of the cm_id is initialized to 1.  However, cm_process_work
> will decrement the refcount after invoking all callbacks.  The result
> is that the cm_id will end up with refcount set to 0 by the end of the
> sidr req handler.
> 
> If a user tries to destroy the cm_id, the destruction will proceed,
> under the incorrect assumption that no other threads are referencing
> the cm_id.  This can lead to a crash when the cm callback thread tries
> to access the cm_id.
> 
> This problem was noticed as part of a larger investigation with kernel
> crashes in the rdma_cm when running on a real time OS.
> 
> Signed-off-by: Sean Hefty <sean.hefty@intel.com>
> Acked-by: Doug Ledford <dledford@redhat.com>
> Cc: <stable@kernel.org>
> Signed-off-by: Roland Dreier <roland@purestorage.com>
> 
> (cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
> Signed-off-by: Brad Figg <brad.figg@canonical.com>

Acked-by: Leann Ogasawara <leann.ogasawara@canonical.com>

> ---
>  drivers/infiniband/core/cm.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
> index 64e0903..1d9616b 100644
> --- a/drivers/infiniband/core/cm.c
> +++ b/drivers/infiniband/core/cm.c
> @@ -2989,6 +2989,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
>  		goto out; /* No match. */
>  	}
>  	atomic_inc(&cur_cm_id_priv->refcount);
> +	atomic_inc(&cm_id_priv->refcount);
>  	spin_unlock_irq(&cm.lock);
>  
>  	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;
> -- 
> 1.7.0.4
> 
>
John Johansen - April 25, 2011, 8:53 p.m.
On 04/25/2011 12:29 PM, Brad Figg wrote:
> From: Sean Hefty <sean.hefty@intel.com>
> 
> CVE-2011-0695
> 
> BugLink: http://bugs.launchpad.net/bugs/770369
> 
> When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
> refcount of the cm_id is initialized to 1.  However, cm_process_work
> will decrement the refcount after invoking all callbacks.  The result
> is that the cm_id will end up with refcount set to 0 by the end of the
> sidr req handler.
> 
> If a user tries to destroy the cm_id, the destruction will proceed,
> under the incorrect assumption that no other threads are referencing
> the cm_id.  This can lead to a crash when the cm callback thread tries
> to access the cm_id.
> 
> This problem was noticed as part of a larger investigation with kernel
> crashes in the rdma_cm when running on a real time OS.
> 
> Signed-off-by: Sean Hefty <sean.hefty@intel.com>
> Acked-by: Doug Ledford <dledford@redhat.com>
> Cc: <stable@kernel.org>
> Signed-off-by: Roland Dreier <roland@purestorage.com>
> 
> (cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
> Signed-off-by: Brad Figg <brad.figg@canonical.com>
Acked-by: John Johansen <john.johansen@canonical.com>



> ---
>  drivers/infiniband/core/cm.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
> index 64e0903..1d9616b 100644
> --- a/drivers/infiniband/core/cm.c
> +++ b/drivers/infiniband/core/cm.c
> @@ -2989,6 +2989,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
>  		goto out; /* No match. */
>  	}
>  	atomic_inc(&cur_cm_id_priv->refcount);
> +	atomic_inc(&cm_id_priv->refcount);
>  	spin_unlock_irq(&cm.lock);
>  
>  	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;
Tim Gardner - April 26, 2011, 12:53 a.m.
On 04/25/2011 01:29 PM, Brad Figg wrote:
> From: Sean Hefty<sean.hefty@intel.com>
>
> CVE-2011-0695
>
> BugLink: http://bugs.launchpad.net/bugs/770369
>
> When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
> refcount of the cm_id is initialized to 1.  However, cm_process_work
> will decrement the refcount after invoking all callbacks.  The result
> is that the cm_id will end up with refcount set to 0 by the end of the
> sidr req handler.
>
> If a user tries to destroy the cm_id, the destruction will proceed,
> under the incorrect assumption that no other threads are referencing
> the cm_id.  This can lead to a crash when the cm callback thread tries
> to access the cm_id.
>
> This problem was noticed as part of a larger investigation with kernel
> crashes in the rdma_cm when running on a real time OS.
>
> Signed-off-by: Sean Hefty<sean.hefty@intel.com>
> Acked-by: Doug Ledford<dledford@redhat.com>
> Cc:<stable@kernel.org>
> Signed-off-by: Roland Dreier<roland@purestorage.com>
>
> (cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
> Signed-off-by: Brad Figg<brad.figg@canonical.com>
> ---
>   drivers/infiniband/core/cm.c |    1 +
>   1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
> index 64e0903..1d9616b 100644
> --- a/drivers/infiniband/core/cm.c
> +++ b/drivers/infiniband/core/cm.c
> @@ -2989,6 +2989,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
>   		goto out; /* No match. */
>   	}
>   	atomic_inc(&cur_cm_id_priv->refcount);
> +	atomic_inc(&cm_id_priv->refcount);
>   	spin_unlock_irq(&cm.lock);
>
>   	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;

I'm a bit curious how you got these 2 patches to apply 'cause they were 
released with 2.6.35-29.51. Did you start with master instead of 
master-next ?

rtg
Brad Figg - April 26, 2011, 3:13 a.m.
On 04/25/2011 05:53 PM, Tim Gardner wrote:
> On 04/25/2011 01:29 PM, Brad Figg wrote:
>> From: Sean Hefty<sean.hefty@intel.com>
>>
>> CVE-2011-0695
>>
>> BugLink: http://bugs.launchpad.net/bugs/770369
>>
>> When processing a SIDR REQ, the ib_cm allocates a new cm_id. The
>> refcount of the cm_id is initialized to 1. However, cm_process_work
>> will decrement the refcount after invoking all callbacks. The result
>> is that the cm_id will end up with refcount set to 0 by the end of the
>> sidr req handler.
>>
>> If a user tries to destroy the cm_id, the destruction will proceed,
>> under the incorrect assumption that no other threads are referencing
>> the cm_id. This can lead to a crash when the cm callback thread tries
>> to access the cm_id.
>>
>> This problem was noticed as part of a larger investigation with kernel
>> crashes in the rdma_cm when running on a real time OS.
>>
>> Signed-off-by: Sean Hefty<sean.hefty@intel.com>
>> Acked-by: Doug Ledford<dledford@redhat.com>
>> Cc:<stable@kernel.org>
>> Signed-off-by: Roland Dreier<roland@purestorage.com>
>>
>> (cherry-picked from commit 29963437a48475036353b95ab142bf199adb909e)
>> Signed-off-by: Brad Figg<brad.figg@canonical.com>
>> ---
>> drivers/infiniband/core/cm.c | 1 +
>> 1 files changed, 1 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
>> index 64e0903..1d9616b 100644
>> --- a/drivers/infiniband/core/cm.c
>> +++ b/drivers/infiniband/core/cm.c
>> @@ -2989,6 +2989,7 @@ static int cm_sidr_req_handler(struct cm_work *work)
>> goto out; /* No match. */
>> }
>> atomic_inc(&cur_cm_id_priv->refcount);
>> + atomic_inc(&cm_id_priv->refcount);
>> spin_unlock_irq(&cm.lock);
>>
>> cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;
>
> I'm a bit curious how you got these 2 patches to apply 'cause they were released with 2.6.35-29.51. Did you start with master instead of master-next ?
>
> rtg

That's a really good question. I patched an old local repo. I got distracted
working on these and thought I'd updated my repo but had not. Thanks for
catching this and sorry to waste everyone's time.

Brad

Patch

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 64e0903..1d9616b 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -2989,6 +2989,7 @@  static int cm_sidr_req_handler(struct cm_work *work)
 		goto out; /* No match. */
 	}
 	atomic_inc(&cur_cm_id_priv->refcount);
+	atomic_inc(&cm_id_priv->refcount);
 	spin_unlock_irq(&cm.lock);
 
 	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;