On 30 May 2017 at 10:09, Ben Pfaff <[email protected]> wrote:
> On Tue, May 30, 2017 at 10:01:55AM -0700, Joe Stringer wrote:
>> On 26 May 2017 at 22:07, Ben Pfaff <[email protected]> wrote:
>> > Coverity says that, in this test at the end of revalidate_ukey__(),
>> > 'netflow' is always NULL:
>> >
>> >     if (netflow && result == UKEY_DELETE) {
>> >         netflow_flow_clear(netflow, &ctx.flow);
>> >     }
>> >
>> > I suspect that it is right.
>> >
>> > This is from
>> > https://scan3.coverity.com/reports.htm#v16889/p10449/fileInstanceId=14762999&defectInstanceId=4304056&mergedDefectId=179567,
>> > which I hope is more or less publicly accessible.
>>
>> Looks like it works on an account whitelist system. I've requested to
>> look at it.
>
> I sent you an invitation.
>
>> > Any thoughts?
>>
>> The address of 'netflow' is stored into the reval_context, which goes
>> through xlate_key() and xlate_lookup() and presumably is populated
>> there, so that if the translation determines that the flow should be
>> deleted, then this code should execute to clear the netflow flow. Its
>> usage is definitely more convoluted than I'd prefer.
>
> OK.  If you think it's correct, then that's fine.  It's even better, of
> course, if we can make it more obviously correct.

I believe it's correct. I don't have any ideas to make it more obvious
other than perhaps to add a comment somewhere to highlight that this
is how it's populated.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to