2013-10-19 11:38:46

by Geyslan G. Bem

Subject: net: sctp: possible dereference after freeing

Hi maintainers,

I would like to know if these are catches:

/net/sctp/endpointola.c (281)
static void sctp_endpoint_destroy(struct sctp_endpoint *ep)
{
struct sock *sk;
...
kfree(ep);
SCTP_DBG_OBJCNT_DEC(ep);
}

The 'ep' object counter is being decremented?! Is the kfree to be there indeed?
Let me know what was intended here.

Same here:
/net/sctp/endpointola.c (165)
static void sctp_transport_destroy_rcu(struct rcu_head *head)
{
struct sctp_transport *transport;
...
kfree(transport);
SCTP_DBG_OBJCNT_DEC(transport);
}

Regards,

Geyslan Gregório Bem
hackingbits.com

2013-10-19 11:49:14

by Geyslan G. Bem

Subject: Re: net: sctp: possible dereference after freeing

2013/10/19 Geyslan Gregório Bem <[email protected]>:
> Hi maintainers,
>
> I would like to know if these are catches:
>
> /net/sctp/endpointola.c (281)
> static void sctp_endpoint_destroy(struct sctp_endpoint *ep)
> {
> struct sock *sk;
> ...
> kfree(ep);
> SCTP_DBG_OBJCNT_DEC(ep);
> }
>
> The 'ep' object counter is being decremented?! Is the kfree to be there indeed?
> Let me know what was intended here.
>
> Same here:
> /net/sctp/endpointola.c (165)
> static void sctp_transport_destroy_rcu(struct rcu_head *head)
> {
> struct sctp_transport *transport;
> ...
> kfree(transport);
> SCTP_DBG_OBJCNT_DEC(transport);
> }
>
> Regards,
>
> Geyslan Gregório Bem
> hackingbits.com

Oops. I got it. ;)

extern atomic_t sctp_dbg_objcnt_ep;
http://lxr.free-electrons.com/source/include/net/sctp/sctp.h#L269
extern atomic_t sctp_dbg_objcnt_transport;
http://lxr.free-electrons.com/source/arch/x86/include/asm/atomic.h#L105

Cheers.

2013-10-19 17:36:42

by Vlad Yasevich

Subject: Re: net: sctp: possible dereference after freeing

On Oct 19, 2013, at 7:49 AM, Geyslan Gregório Bem <[email protected]> wrote:

> 2013/10/19 Geyslan Gregório Bem <[email protected]>:
>> Hi maintainers,
>>
>> I would like to know if these are catches:
>>
>> /net/sctp/endpointola.c (281)
>> static void sctp_endpoint_destroy(struct sctp_endpoint *ep)
>> {
>> struct sock *sk;
>> ...
>> kfree(ep);
>> SCTP_DBG_OBJCNT_DEC(ep);
>> }
>>
>> The 'ep' object counter is being decremented?! Is the kfree to be there indeed?
>> Let me know what was intended here.
>>
>> Same here:
>> /net/sctp/endpointola.c (165)
>> static void sctp_transport_destroy_rcu(struct rcu_head *head)
>> {
>> struct sctp_transport *transport;
>> ...
>> kfree(transport);
>> SCTP_DBG_OBJCNT_DEC(transport);
>> }
>>
>> Regards,
>>
>> Geyslan Gregório Bem
>> hackingbits.com
>
> Oops. I got it. ;)
>
> extern atomic_t sctp_dbg_objcnt_ep;
> http://lxr.free-electrons.com/source/include/net/sctp/sctp.h#L269
> extern atomic_t sctp_dbg_objcnt_transport;
> http://lxr.free-electrons.com/source/arch/x86/include/asm/atomic.h#L105
>
> Cheers.

There is no reference here since the macro turns variable name to string using ## name.
So the order doesn't matter really.

-vlad -