Patchwork [net] sctp: jsctp_sf_eat_sack: fix kernel NULL ptr dereference

login
register
mail settings
Submitter Daniel Borkmann
Date Dec. 14, 2012, 2:27 p.m.
Message ID <2813cf01509e495fee13ff1fab309f1186c0e57a.1355494956.git.dborkman@redhat.com>
Download mbox | patch
Permalink /patch/206511/
State Rejected
Delegated to: David Miller
Headers show

Comments

Daniel Borkmann - Dec. 14, 2012, 2:27 p.m.
During debugging a sctp problem, I hit a kernel NULL pointer
dereference in the jprobes callback jsctp_sf_eat_sack(). This
small patch fixes the panic.

Cc: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
---
 net/sctp/probe.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Vlad Yasevich - Dec. 14, 2012, 3:12 p.m.
On 12/14/2012 09:27 AM, Daniel Borkmann wrote:
> During debugging a sctp problem, I hit a kernel NULL pointer
> dereference in the jprobes callback jsctp_sf_eat_sack(). This
> small patch fixes the panic.
>
> Cc: Vlad Yasevich <vyasevich@gmail.com>
> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
> ---
>   net/sctp/probe.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/sctp/probe.c b/net/sctp/probe.c
> index bc6cd75..0a4e9d6 100644
> --- a/net/sctp/probe.c
> +++ b/net/sctp/probe.c
> @@ -134,7 +134,7 @@ sctp_disposition_t jsctp_sf_eat_sack(const struct sctp_endpoint *ep,
>
>   	sp = asoc->peer.primary_path;
>
> -	if ((full || sp->cwnd != lcwnd) &&
> +	if (sp && (full || sp->cwnd != lcwnd) &&
>   	    (!port || asoc->peer.port == port ||
>   	     ep->base.bind_addr.port == port)) {
>   		lcwnd = sp->cwnd;
>


Sorry, but this patch isn't making much sense.  @sp points to the 
primary path of the association and that can not be null if we receiving
SACKs on this association.

sctp_sf_eat_sack_6_2(), which we are probing, is only called while the 
association is valid and all the transports still exist.  It is also 
called under lock, so the transports can not go away while processing of 
the SACK.  So unless there is some kind of jprobe issue, the pointer
you are looking at should always be valid.

-vlad
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Daniel Borkmann - Dec. 14, 2012, 4:57 p.m.
On 12/14/2012 04:12 PM, Vlad Yasevich wrote:
> On 12/14/2012 09:27 AM, Daniel Borkmann wrote:
>> During debugging a sctp problem, I hit a kernel NULL pointer
>> dereference in the jprobes callback jsctp_sf_eat_sack(). This
>> small patch fixes the panic.
>>
>> Cc: Vlad Yasevich <vyasevich@gmail.com>
>> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
>> ---
>>   net/sctp/probe.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/net/sctp/probe.c b/net/sctp/probe.c
>> index bc6cd75..0a4e9d6 100644
>> --- a/net/sctp/probe.c
>> +++ b/net/sctp/probe.c
>> @@ -134,7 +134,7 @@ sctp_disposition_t jsctp_sf_eat_sack(const struct sctp_endpoint *ep,
>>
>>       sp = asoc->peer.primary_path;
>>
>> -    if ((full || sp->cwnd != lcwnd) &&
>> +    if (sp && (full || sp->cwnd != lcwnd) &&
>>           (!port || asoc->peer.port == port ||
>>            ep->base.bind_addr.port == port)) {
>>           lcwnd = sp->cwnd;
>
> Sorry, but this patch isn't making much sense.  @sp points to the primary path of the association and that can not be null if we receiving
> SACKs on this association.
>
> sctp_sf_eat_sack_6_2(), which we are probing, is only called while the association is valid and all the transports still exist.  It is also called under lock, so the transports can not go away while processing of the SACK.  So unless there is some kind of jprobe issue, the pointer
> you are looking at should always be valid.

Okay, I'll dig a bit deeper into that over the weekend.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
danborkmann@iogearbox.net - Dec. 14, 2012, 9:49 p.m.
On Fri, Dec 14, 2012 at 5:57 PM, Daniel Borkmann <dborkman@redhat.com> wrote:
> On 12/14/2012 04:12 PM, Vlad Yasevich wrote:
>> On 12/14/2012 09:27 AM, Daniel Borkmann wrote:
>>>
>>> During debugging a sctp problem, I hit a kernel NULL pointer
>>> dereference in the jprobes callback jsctp_sf_eat_sack(). This
>>> small patch fixes the panic.
>>>
>>> Cc: Vlad Yasevich <vyasevich@gmail.com>
>>> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
>>> ---
>>>   net/sctp/probe.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/net/sctp/probe.c b/net/sctp/probe.c
>>> index bc6cd75..0a4e9d6 100644
>>> --- a/net/sctp/probe.c
>>> +++ b/net/sctp/probe.c
>>> @@ -134,7 +134,7 @@ sctp_disposition_t jsctp_sf_eat_sack(const struct
>>> sctp_endpoint *ep,
>>>
>>>       sp = asoc->peer.primary_path;
>>>
>>> -    if ((full || sp->cwnd != lcwnd) &&
>>> +    if (sp && (full || sp->cwnd != lcwnd) &&
>>>           (!port || asoc->peer.port == port ||
>>>            ep->base.bind_addr.port == port)) {
>>>           lcwnd = sp->cwnd;
>>
>> Sorry, but this patch isn't making much sense.  @sp points to the primary
>> path of the association and that can not be null if we receiving
>> SACKs on this association.
>>
>> sctp_sf_eat_sack_6_2(), which we are probing, is only called while the
>> association is valid and all the transports still exist.  It is also called
>> under lock, so the transports can not go away while processing of the SACK.
>> So unless there is some kind of jprobe issue, the pointer
>> you are looking at should always be valid.
>
> Okay, I'll dig a bit deeper into that over the weekend.

That's just for the record, I'll do further tests / debugging tomorrow ...

I compiled the net kernel with kprobes selftest / sanity test
(kernel/test_kprobes.c) and it passed:

[    2.978082] Kprobe smoke test started
[    3.034014] Kprobe smoke test passed successfully

There, also jprobes are tested.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
danborkmann@iogearbox.net - Dec. 15, 2012, 7:58 p.m.
On Fri, Dec 14, 2012 at 10:49 PM, Daniel Borkmann
<danborkmann@iogearbox.net> wrote:
> On Fri, Dec 14, 2012 at 5:57 PM, Daniel Borkmann <dborkman@redhat.com> wrote:
>> On 12/14/2012 04:12 PM, Vlad Yasevich wrote:
>>> On 12/14/2012 09:27 AM, Daniel Borkmann wrote:
>>>>
>>>> During debugging a sctp problem, I hit a kernel NULL pointer
>>>> dereference in the jprobes callback jsctp_sf_eat_sack(). This
>>>> small patch fixes the panic.
>>>>
>>>> Cc: Vlad Yasevich <vyasevich@gmail.com>
>>>> Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
>>>> ---
>>>>   net/sctp/probe.c | 2 +-
>>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/net/sctp/probe.c b/net/sctp/probe.c
>>>> index bc6cd75..0a4e9d6 100644
>>>> --- a/net/sctp/probe.c
>>>> +++ b/net/sctp/probe.c
>>>> @@ -134,7 +134,7 @@ sctp_disposition_t jsctp_sf_eat_sack(const struct
>>>> sctp_endpoint *ep,
>>>>
>>>>       sp = asoc->peer.primary_path;
>>>>
>>>> -    if ((full || sp->cwnd != lcwnd) &&
>>>> +    if (sp && (full || sp->cwnd != lcwnd) &&
>>>>           (!port || asoc->peer.port == port ||
>>>>            ep->base.bind_addr.port == port)) {
>>>>           lcwnd = sp->cwnd;
>>>
>>> Sorry, but this patch isn't making much sense.  @sp points to the primary
>>> path of the association and that can not be null if we receiving
>>> SACKs on this association.
>>>
>>> sctp_sf_eat_sack_6_2(), which we are probing, is only called while the
>>> association is valid and all the transports still exist.  It is also called
>>> under lock, so the transports can not go away while processing of the SACK.
>>> So unless there is some kind of jprobe issue, the pointer
>>> you are looking at should always be valid.
>>
>> Okay, I'll dig a bit deeper into that over the weekend.
>
> That's just for the record, I'll do further tests / debugging tomorrow ...
>
> I compiled the net kernel with kprobes selftest / sanity test
> (kernel/test_kprobes.c) and it passed:
>
> [    2.978082] Kprobe smoke test started
> [    3.034014] Kprobe smoke test passed successfully
>
> There, also jprobes are tested.

Found the problem, will send a new patch for the net tree in a couple
of minutes.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/net/sctp/probe.c b/net/sctp/probe.c
index bc6cd75..0a4e9d6 100644
--- a/net/sctp/probe.c
+++ b/net/sctp/probe.c
@@ -134,7 +134,7 @@  sctp_disposition_t jsctp_sf_eat_sack(const struct sctp_endpoint *ep,
 
 	sp = asoc->peer.primary_path;
 
-	if ((full || sp->cwnd != lcwnd) &&
+	if (sp && (full || sp->cwnd != lcwnd) &&
 	    (!port || asoc->peer.port == port ||
 	     ep->base.bind_addr.port == port)) {
 		lcwnd = sp->cwnd;