[Pascal Quantin <pascal.quantin@xxxxxxxxx>]

2014-11-28 12:40 GMT+01:00 Pascal Quantin <pascal.quantin@xxxxxxxxx>:

2014-11-27 16:37 GMT+01:00 Evan Huus <eapache@xxxxxxxxx>:

On Thu, Nov 27, 2014 at 10:21 AM, Alexis La Goutte
<alexis.lagoutte@xxxxxxxxx> wrote:
> Hi,
>
> On Thu, Nov 27, 2014 at 4:13 AM, 蓝常珍 <lanczyx@xxxxxxxxx> wrote:
>> In the function "dissect_ipv6" of the ipv6
>> dissector(packet-ipv6.c),the ip6_hdr struct is allocated on the
>> stack,then it's address is passed to tap_queue_packet.I notice that
>> the other dissectors do not look like this.
>>
>> The code snippet of the dissect_ipv6 function,from dev-version 1.99.0:
>>
>> static void
>> dissect_ipv6(tvbuff_t *tvb, packet_info *pinfo, proto_tree *tree)
>> {
>> struct ip6_hdr ipv6;
>> ...
>> tap_queue_packet(ipv6_tap, pinfo, &ipv6);
>> ...
>> }
> What the bug ?

I think the implication is that when the tap runs, the pointer will be
invalid because the stack frame will have been destroyed. So any tap
trying to access that structure will cause an invalid memory access?

Hi,

yes I agree. Fortunately it looks like there is no listener for this tap (unless I missed it). Should we fix it or remove it completely?

Actually there are users of this tap. I fixed it in https://code.wireshark.org/review/#/c/6276/

Regards,
Pascal.