[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH] rpc: Fix crash on error paths of message dispatching



Eric Blake wrote:
> On 01/29/2013 07:05 PM, Jim Fehlig wrote:
>
>   
>>>> Mention CVE-2013-0170 in the commit message, now that it is public:
>>>> https://bugzilla.redhat.com/show_bug.cgi?id=893450
>>>>
>>>>         
>>>>> * rpc/virnetserverclient.c: virNetServerClientDispatchRead:
>>>>>      - avoid use after free of RPC messages
>>>>> ---
>>>>>   src/rpc/virnetserverclient.c | 3 +++
>>>>>   1 file changed, 3 insertions(+)
>>>>>           
>>>> ACK.  Looks like we need this on {v0.10.2,v0.9.11,v0.9.6}-maint as well.
>>>>         
>>> Thanks. I added the CVE notice and pushed to upstream and the v0.10.2
>>> and v0.9.11 maint branches. v0.9.6 is not vulnerable. The problem was
>>> introduced in 0.9.7
>>>       
>> Hi Peter,
>>
>> Looks like 0.9.6 was vulnerable since this made its way to the
>> v0.9.6-maint branch as well.  Do you happen to know when this was
>> introduced?
>>     
>
> I did some more research:
>   

Eric,

Thank you for the research and explanation, it was not expected but very
much appreciated.  I was actually having quite the RTFM moment after
asking this question...

Regards,
Jim

> The original problem was introduced in commit 4e00b1d (libvirt 0.9.3),
> when we switched over to new RPC handling; there, we only had one faulty
> error path.  Later, commit 3ae0ab67 (libvirt 0.9.7) exacerbated the
> problem, by adding two more faulty error paths.  Peter's test case when
> originally reporting the CVE was on one of the error paths added in
> 0.9.7, hence his claim that "the problem was introduced in 0.9.7"; but I
> still think it is possible to trigger the remaining faulty error path
> when targeting libvirt 0.9.3, and agree with Cole's backport to the
> v0.9.6-maint branch.
>
>   


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]