From: Alexander Kuzmenkov <akuzmenkov@tigerdata.com>
Message-ID: <b607e42e-34ff-414a-b727-dd5ec70babe3@timescale.com>
Date: Tue, 10 Mar 2026 22:51:50 +0100
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: Fix uninitialized xl_running_xacts padding
To: Andres Freund <andres@anarazel.de>,
 Anthonin Bonnefoy <anthonin.bonnefoy@datadoghq.com>
Cc: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>,
 Michael Paquier <michael@paquier.xyz>, Thomas Munro
 <thomas.munro@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
References: 
 <CAO6_Xqoxp7C+y0L==xZXH5V=9PjpBx4T9vJYs87EbxFp_9nwOA@mail.gmail.com>
 <CA+hUKG++LE6P6g4n+-QPHBwAnvcVRyG1tUnzscUriWAFHc6s6Q@mail.gmail.com>
 <aZJuoggMUaXjwoDO@paquier.xyz>
 <aZLHYtCsEldmm8Eu@ip-10-97-1-34.eu-west-3.compute.internal>
 <CAO6_XqpGYneOe357s4QhOTYpU6UEgk3JJrBiyeuk-9n=f7p=Vw@mail.gmail.com>
 <aoaj45foewpjtu6r5cs67yrx4en3pkurs23e4azv6tbikpw6c3@h3pnaqaksoeg>
Content-Language: en-US
In-Reply-To: <aoaj45foewpjtu6r5cs67yrx4en3pkurs23e4azv6tbikpw6c3@h3pnaqaksoeg>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: 
 <https://www.postgresql.org/message-id/b607e42e-34ff-414a-b727-dd5ec70babe3%40timescale.com>
Precedence: bulk

On 16/02/2026 21:10, Andres Freund wrote:
> I don't think it makes a whole lot of sense to tackle this specifically for
> xl_running_xacts. Until now we just accepted that WAL insertions can contain
> random padding. If we don't want that, we should go around and make sure that
> there is no padding (or padding is initialized) for *all* WAL records,
> document that as the rule, and remove the relevant valgrind suppressions.

That's not random, that's server memory, right? Probably not another 
Heartbleed, but I'd rather initialize a few locals than find out.


Happy to see this being worked on, these uninitialized WAL records are a 
major obstacle to enabling MemorySanitizer. I ran into this again today 
and this is how I found this thread. Unfortunately the MemorySanitizer 
can't even use the same suppressions as Valgrind, because the 
suppression architecture is different (can only remove the checks from a 
given function, not all stack traces that have this function like 
Valgrind does).


Best regards
Alexander Kuzmenkov
TigerData