Обсуждение: Bug in pg_get_aios()

Hi,

If I understand correctly, we want to set nulls[0] to true if owner_pid equals to 0:

```
diff --git a/src/backend/storage/aio/aio_funcs.c b/src/backend/storage/aio/aio_funcs.c
index 8997c762062..bcdd82318f7 100644
--- a/src/backend/storage/aio/aio_funcs.c
+++ b/src/backend/storage/aio/aio_funcs.c
@@ -149,7 +149,7 @@ retry:
                if (owner_pid != 0)
                        values[0] = Int32GetDatum(owner_pid);
                else
-                       nulls[0] = false;
+                       nulls[0] = true;
 
                /* column: IO's id */
                values[1] = Int32GetDatum(ioh_id);
```

--
Regards,
ChangAo Chen

> On Mar 23, 2026, at 14:23, cca5507 <cca5507@qq.com> wrote:
>
> Hi,
>
> If I understand correctly, we want to set nulls[0] to true if owner_pid equals to 0:
>
> ```
> diff --git a/src/backend/storage/aio/aio_funcs.c b/src/backend/storage/aio/aio_funcs.c
> index 8997c762062..bcdd82318f7 100644
> --- a/src/backend/storage/aio/aio_funcs.c
> +++ b/src/backend/storage/aio/aio_funcs.c
> @@ -149,7 +149,7 @@ retry:
>                if (owner_pid != 0)
>                        values[0] = Int32GetDatum(owner_pid);
>                else
> -                       nulls[0] = false;
> +                       nulls[0] = true;
>
>                /* column: IO's id */
>                values[1] = Int32GetDatum(ioh_id);
> ```
>
> --
> Regards,
> ChangAo Chen

Indeed an obvious bug.

If we looking downwards, there is a correct usage:
```
        /* column: raw result (i.e. some form of syscall return value) */
        if (start_state == PGAIO_HS_COMPLETED_IO
            || start_state == PGAIO_HS_COMPLETED_SHARED
            || start_state == PGAIO_HS_COMPLETED_LOCAL)
            values[9] = Int32GetDatum(ioh_copy.result);
        else
            nulls[9] = true;
```

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/

On Mon, Mar 23, 2026 at 02:30:28PM +0800, Chao Li wrote:
>> On Mar 23, 2026, at 14:23, cca5507 <cca5507@qq.com> wrote:
>> If I understand correctly, we want to set nulls[0] to true if owner_pid equals to 0:
>
> Indeed an obvious bug.

Or the intention of the code could have been to allow a PID value of 0
to be exposed, but I don't really buy it.  Will fix, thanks for the
report.
--
Michael

Hi,

On March 23, 2026 4:44:21 AM EDT, Michael Paquier <michael@paquier.xyz> wrote:
>On Mon, Mar 23, 2026 at 02:30:28PM +0800, Chao Li wrote:
>>> On Mar 23, 2026, at 14:23, cca5507 <cca5507@qq.com> wrote:
>>> If I understand correctly, we want to set nulls[0] to true if owner_pid equals to 0:
>>
>> Indeed an obvious bug.
>
>Or the intention of the code could have been to allow a PID value of 0
>to be exposed, but I don't really buy it.

That certainly wasn't the intent :).

I suspect this is unreachable, we wait for ios before exiting, but of course that doesn't excuse broken code.


>Will fix, thanks for the report.

Thanks!

Andres


--
Sent from my Android device with K-9 Mail. Please excuse my brevity.