Re: BUG #14615: ReplicationOriginShmemInit Memory access cross-border

On Wed, Apr 12, 2017 at 12:20 AM, Andres Freund <andres(at)anarazel(dot)de> wrote:
> On 2017-04-10 15:38:56 +0900, Michael Paquier wrote:
>> diff --git a/src/backend/replication/logical/origin.c b/src/backend/replication/logical/origin.c
>> index 5eaf863e02..0aa468789c 100644
>> --- a/src/backend/replication/logical/origin.c
>> +++ b/src/backend/replication/logical/origin.c
>> @@ -473,7 +473,8 @@ ReplicationOriginShmemInit(void)
>>
>> replication_states_ctl->tranche_id = LWTRANCHE_REPLICATION_ORIGIN;
>>
>> - MemSet(replication_states, 0, ReplicationOriginShmemSize());
>> + MemSet(replication_states, 0,
>> + mul_size(max_replication_slots, sizeof(ReplicationState)));
>
> What's the benefit of using mul_size here? That's usually only
> beneficial in the original size computation - during use/initialization
> an actual error should be impossible.

Clarity in initializing only the replication states.

> To me the right fix seems to be to just do:
> - replication_states_ctl->tranche_id = LWTRANCHE_REPLICATION_ORIGIN;
> -
> - MemSet(replication_states, 0, ReplicationOriginShmemSize());
> + MemSet(replication_states_ctl, 0, ReplicationOriginShmemSize());
> +
> + replication_states_ctl->tranche_id = LWTRANCHE_REPLICATION_ORIGIN;
>
> No?

That's what Bret is proposing. I am fine either way.
--
Michael