[PATCH] can: check for null sk before deferencing it via the call to sock_net

[PATCH] can: check for null sk before deferencing it via the call to sock_net

[Colin King]

From: Colin Ian King <colin.king@canonical.com>

The assignment of net via call sock_net will dereference sk. This
is performed before a sanity null check on sk, so there could be
a potential null dereference on the sock_net call if sk is null.
Fix this by assigning net after the sk null check. Also replace
the sk == NULL with the more usual !sk idiom.

Detected by CoverityScan CID#1431862 ("Dereference before null check")

Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
 net/can/bcm.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/net/can/bcm.c b/net/can/bcm.c
index 47a8748d953a..a3791674b8ce 100644
--- a/net/can/bcm.c
+++ b/net/can/bcm.c
@@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
 static int bcm_release(struct socket *sock)
 {
 	struct sock *sk = sock->sk;
-	struct net *net = sock_net(sk);
+	struct net *net;
 	struct bcm_sock *bo;
 	struct bcm_op *op, *next;
 
-	if (sk == NULL)
+	if (!sk)
 		return 0;
 
+	net = sock_net(sk);
 	bo = bcm_sk(sk);
 
 	/* remove bcm_ops, timer, rx_unregister(), etc. */
-- 
2.14.1

[PATCH] can: check for null sk before deferencing it via the call to sock_net

[Colin King]

From: Colin Ian King <colin.king@canonical.com>

The assignment of net via call sock_net will dereference sk. This
is performed before a sanity null check on sk, so there could be
a potential null dereference on the sock_net call if sk is null.
Fix this by assigning net after the sk null check. Also replace
the sk = NULL with the more usual !sk idiom.

Detected by CoverityScan CID#1431862 ("Dereference before null check")

Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
 net/can/bcm.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/net/can/bcm.c b/net/can/bcm.c
index 47a8748d953a..a3791674b8ce 100644
--- a/net/can/bcm.c
+++ b/net/can/bcm.c
@@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
 static int bcm_release(struct socket *sock)
 {
 	struct sock *sk = sock->sk;
-	struct net *net = sock_net(sk);
+	struct net *net;
 	struct bcm_sock *bo;
 	struct bcm_op *op, *next;
 
-	if (sk = NULL)
+	if (!sk)
 		return 0;
 
+	net = sock_net(sk);
 	bo = bcm_sk(sk);
 
 	/* remove bcm_ops, timer, rx_unregister(), etc. */
-- 
2.14.1

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Oliver Hartkopp]

On 09/08/2017 05:02 PM, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> The assignment of net via call sock_net will dereference sk. This
> is performed before a sanity null check on sk, so there could be
> a potential null dereference on the sock_net call if sk is null.
> Fix this by assigning net after the sk null check. Also replace
> the sk == NULL with the more usual !sk idiom.
> 
> Detected by CoverityScan CID#1431862 ("Dereference before null check")
> 
> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
> Signed-off-by: Colin Ian King <colin.king@canonical.com>

Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>


Thanks Collin!

> ---
>   net/can/bcm.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/net/can/bcm.c b/net/can/bcm.c
> index 47a8748d953a..a3791674b8ce 100644
> --- a/net/can/bcm.c
> +++ b/net/can/bcm.c
> @@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
>   static int bcm_release(struct socket *sock)
>   {
>   	struct sock *sk = sock->sk;
> -	struct net *net = sock_net(sk);
> +	struct net *net;
>   	struct bcm_sock *bo;
>   	struct bcm_op *op, *next;
>   
> -	if (sk == NULL)
> +	if (!sk)
>   		return 0;
>   
> +	net = sock_net(sk);
>   	bo = bcm_sk(sk);
>   
>   	/* remove bcm_ops, timer, rx_unregister(), etc. */
> 

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Oliver Hartkopp]

On 09/08/2017 05:02 PM, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> The assignment of net via call sock_net will dereference sk. This
> is performed before a sanity null check on sk, so there could be
> a potential null dereference on the sock_net call if sk is null.
> Fix this by assigning net after the sk null check. Also replace
> the sk = NULL with the more usual !sk idiom.
> 
> Detected by CoverityScan CID#1431862 ("Dereference before null check")
> 
> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
> Signed-off-by: Colin Ian King <colin.king@canonical.com>

Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>


Thanks Collin!

> ---
>   net/can/bcm.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/net/can/bcm.c b/net/can/bcm.c
> index 47a8748d953a..a3791674b8ce 100644
> --- a/net/can/bcm.c
> +++ b/net/can/bcm.c
> @@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
>   static int bcm_release(struct socket *sock)
>   {
>   	struct sock *sk = sock->sk;
> -	struct net *net = sock_net(sk);
> +	struct net *net;
>   	struct bcm_sock *bo;
>   	struct bcm_op *op, *next;
>   
> -	if (sk = NULL)
> +	if (!sk)
>   		return 0;
>   
> +	net = sock_net(sk);
>   	bo = bcm_sk(sk);
>   
>   	/* remove bcm_ops, timer, rx_unregister(), etc. */
> 

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Oliver Hartkopp]

Hi Marc,

this fix from Colin still didn't make it into upstream.

Can you please push it to Dave along with this one:
https://marc.info/?l=linux-can&m=150575760930022&w=2

Many thanks,
Oliver

On 10/11/2017 09:03 AM, Chunyu Hu wrote:
> I hit this oops in real world when running trinity test in ppc64le
> with a 4.14-rc3 kernel.
> 
> [  154.507777] NIP [d0000000065919fc] bcm_release+0x4c/0x320 [can_bcm]
> [  154.508300] LR [c000000000a0e070] sock_close+0x50/0x110
> [  154.508421] Call Trace:
> [  154.508481] [c0000001ebe93c20] [00000001fffffbf7] 0x1fffffbf7 (unreliable)
> [  154.508808] [c0000001ebe93cd0] [c000000000a0e070] sock_close+0x50/0x110
> [  154.509015] [c0000001ebe93d40] [c000000000438104] __fput+0xe4/0x2f0
> [  154.509168] [c0000001ebe93da0] [c000000000163790] task_work_run+0x140/0x1a0
> [  154.509320] [c0000001ebe93e00] [c00000000001e8a8]
> do_notify_resume+0x128/0x130
> [  154.509504] [c0000001ebe93e30] [c00000000000b5c4]
> ret_from_except_lite+0x70/0x74
> [  154.509702] Instruction dump:
> [  154.509793] fb61ffd8 fb81ffe0 fba1ffe8 fbc1fff0 fbe1fff8 f8010010
> f821ff51 7c791b78
> [  154.509978] 48000008 e8410018 ebb90020 2fbd0000 <eb9d0030> 419e0238
> 387d02d0 7fbeeb78
> 
> On 9 September 2017 at 01:46, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>>
>>
>> On 09/08/2017 05:02 PM, Colin King wrote:
>>>
>>> From: Colin Ian King <colin.king@canonical.com>
>>>
>>> The assignment of net via call sock_net will dereference sk. This
>>> is performed before a sanity null check on sk, so there could be
>>> a potential null dereference on the sock_net call if sk is null.
>>> Fix this by assigning net after the sk null check. Also replace
>>> the sk == NULL with the more usual !sk idiom.
>>>
>>> Detected by CoverityScan CID#1431862 ("Dereference before null check")
>>>
>>> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM
>>> protocol")
>>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>>
>>
>> Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
>>
>>
>> Thanks Collin!
>>
>>
>>> ---
>>>    net/can/bcm.c | 5 +++--
>>>    1 file changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/net/can/bcm.c b/net/can/bcm.c
>>> index 47a8748d953a..a3791674b8ce 100644
>>> --- a/net/can/bcm.c
>>> +++ b/net/can/bcm.c
>>> @@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
>>>    static int bcm_release(struct socket *sock)
>>>    {
>>>          struct sock *sk = sock->sk;
>>> -       struct net *net = sock_net(sk);
>>> +       struct net *net;
>>>          struct bcm_sock *bo;
>>>          struct bcm_op *op, *next;
>>>    -     if (sk == NULL)
>>> +       if (!sk)
>>>                  return 0;
>>>    +     net = sock_net(sk);
>>>          bo = bcm_sk(sk);
>>>          /* remove bcm_ops, timer, rx_unregister(), etc. */
>>>
>>

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Josh Boyer]

On Fri, Sep 8, 2017 at 1:46 PM, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>
>
> On 09/08/2017 05:02 PM, Colin King wrote:
>>
>> From: Colin Ian King <colin.king@canonical.com>
>>
>> The assignment of net via call sock_net will dereference sk. This
>> is performed before a sanity null check on sk, so there could be
>> a potential null dereference on the sock_net call if sk is null.
>> Fix this by assigning net after the sk null check. Also replace
>> the sk == NULL with the more usual !sk idiom.
>>
>> Detected by CoverityScan CID#1431862 ("Dereference before null check")
>>
>> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM
>> protocol")
>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>
>
> Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>

I don't see this one queued up in the net or net-next trees.  Did it
fall through the cracks or did it get queued up elsewhere?  Seems like
it's a good candidate to get into 4.14?

josh

>
>
> Thanks Collin!
>
>
>> ---
>>   net/can/bcm.c | 5 +++--
>>   1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/net/can/bcm.c b/net/can/bcm.c
>> index 47a8748d953a..a3791674b8ce 100644
>> --- a/net/can/bcm.c
>> +++ b/net/can/bcm.c
>> @@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
>>   static int bcm_release(struct socket *sock)
>>   {
>>         struct sock *sk = sock->sk;
>> -       struct net *net = sock_net(sk);
>> +       struct net *net;
>>         struct bcm_sock *bo;
>>         struct bcm_op *op, *next;
>>   -     if (sk == NULL)
>> +       if (!sk)
>>                 return 0;
>>   +     net = sock_net(sk);
>>         bo = bcm_sk(sk);
>>         /* remove bcm_ops, timer, rx_unregister(), etc. */
>>
>

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Josh Boyer]

On Fri, Sep 8, 2017 at 1:46 PM, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>
>
> On 09/08/2017 05:02 PM, Colin King wrote:
>>
>> From: Colin Ian King <colin.king@canonical.com>
>>
>> The assignment of net via call sock_net will dereference sk. This
>> is performed before a sanity null check on sk, so there could be
>> a potential null dereference on the sock_net call if sk is null.
>> Fix this by assigning net after the sk null check. Also replace
>> the sk = NULL with the more usual !sk idiom.
>>
>> Detected by CoverityScan CID#1431862 ("Dereference before null check")
>>
>> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM
>> protocol")
>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>
>
> Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>

I don't see this one queued up in the net or net-next trees.  Did it
fall through the cracks or did it get queued up elsewhere?  Seems like
it's a good candidate to get into 4.14?

josh

>
>
> Thanks Collin!
>
>
>> ---
>>   net/can/bcm.c | 5 +++--
>>   1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/net/can/bcm.c b/net/can/bcm.c
>> index 47a8748d953a..a3791674b8ce 100644
>> --- a/net/can/bcm.c
>> +++ b/net/can/bcm.c
>> @@ -1493,13 +1493,14 @@ static int bcm_init(struct sock *sk)
>>   static int bcm_release(struct socket *sock)
>>   {
>>         struct sock *sk = sock->sk;
>> -       struct net *net = sock_net(sk);
>> +       struct net *net;
>>         struct bcm_sock *bo;
>>         struct bcm_op *op, *next;
>>   -     if (sk = NULL)
>> +       if (!sk)
>>                 return 0;
>>   +     net = sock_net(sk);
>>         bo = bcm_sk(sk);
>>         /* remove bcm_ops, timer, rx_unregister(), etc. */
>>
>

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Oliver Hartkopp]

On 10/16/2017 06:37 PM, Josh Boyer wrote:
> On Fri, Sep 8, 2017 at 1:46 PM, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>>
>>
>> On 09/08/2017 05:02 PM, Colin King wrote:
>>>
>>> From: Colin Ian King <colin.king@canonical.com>
>>>
>>> The assignment of net via call sock_net will dereference sk. This
>>> is performed before a sanity null check on sk, so there could be
>>> a potential null dereference on the sock_net call if sk is null.
>>> Fix this by assigning net after the sk null check. Also replace
>>> the sk == NULL with the more usual !sk idiom.
>>>
>>> Detected by CoverityScan CID#1431862 ("Dereference before null check")
>>>
>>> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM
>>> protocol")
>>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>>
>>
>> Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
> 
> I don't see this one queued up in the net or net-next trees.  Did it
> fall through the cracks or did it get queued up elsewhere?  Seems like
> it's a good candidate to get into 4.14?

It definitely is!

Marc is our responsible guy for CAN related upstreams - but he seems to 
be busy as I already poked him here:

https://marc.info/?l=linux-can&m=150771819505097&w=2

If he doesn't send a pull request by beginning of next week, I would ask 
Dave to grab these patches - to get them into 4.14.

Best regards,
Oliver

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Oliver Hartkopp]

On 10/16/2017 06:37 PM, Josh Boyer wrote:
> On Fri, Sep 8, 2017 at 1:46 PM, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>>
>>
>> On 09/08/2017 05:02 PM, Colin King wrote:
>>>
>>> From: Colin Ian King <colin.king@canonical.com>
>>>
>>> The assignment of net via call sock_net will dereference sk. This
>>> is performed before a sanity null check on sk, so there could be
>>> a potential null dereference on the sock_net call if sk is null.
>>> Fix this by assigning net after the sk null check. Also replace
>>> the sk = NULL with the more usual !sk idiom.
>>>
>>> Detected by CoverityScan CID#1431862 ("Dereference before null check")
>>>
>>> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM
>>> protocol")
>>> Signed-off-by: Colin Ian King <colin.king@canonical.com>
>>
>>
>> Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
> 
> I don't see this one queued up in the net or net-next trees.  Did it
> fall through the cracks or did it get queued up elsewhere?  Seems like
> it's a good candidate to get into 4.14?

It definitely is!

Marc is our responsible guy for CAN related upstreams - but he seems to 
be busy as I already poked him here:

https://marc.info/?l=linux-can&m\x150771819505097&w=2

If he doesn't send a pull request by beginning of next week, I would ask 
Dave to grab these patches - to get them into 4.14.

Best regards,
Oliver

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Marc Kleine-Budde]

[-- Attachment #1.1: Type: text/plain, Size: 950 bytes --]

On 09/08/2017 05:02 PM, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> The assignment of net via call sock_net will dereference sk. This
> is performed before a sanity null check on sk, so there could be
> a potential null dereference on the sock_net call if sk is null.
> Fix this by assigning net after the sk null check. Also replace
> the sk == NULL with the more usual !sk idiom.
> 
> Detected by CoverityScan CID#1431862 ("Dereference before null check")
> 
> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
> Signed-off-by: Colin Ian King <colin.king@canonical.com>

Applied to can.

Tnx,
Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] can: check for null sk before deferencing it via the call to sock_net

[Marc Kleine-Budde]

[-- Attachment #1.1: Type: text/plain, Size: 950 bytes --]

On 09/08/2017 05:02 PM, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> The assignment of net via call sock_net will dereference sk. This
> is performed before a sanity null check on sk, so there could be
> a potential null dereference on the sock_net call if sk is null.
> Fix this by assigning net after the sk null check. Also replace
> the sk == NULL with the more usual !sk idiom.
> 
> Detected by CoverityScan CID#1431862 ("Dereference before null check")
> 
> Fixes: 384317ef4187 ("can: network namespace support for CAN_BCM protocol")
> Signed-off-by: Colin Ian King <colin.king@canonical.com>

Applied to can.

Tnx,
Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]