From: | Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> |
---|---|
To: | Heikki Linnakangas <hlinnaka(at)iki(dot)fi> |
Cc: | Ranier Vilela <ranier(dot)vf(at)gmail(dot)com>, Pg Hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Fix possible dereference null pointer (src/backend/replication/logical/reorderbuffer.c) |
Date: | 2024-04-15 06:30:48 |
Message-ID: | CAA4eK1Ky8jLD+rF4xh0w9spmYPMkzr=ZMf18UYi59KpGohsAKg@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
On Sat, Apr 13, 2024 at 12:46 PM Heikki Linnakangas <hlinnaka(at)iki(dot)fi> wrote:
>
> I don't much like adding extra runtime checks for "can't happen"
> scenarios either. Assertions would be more clear, but in this case the
> code would just segfault trying to dereference the NULL pointer, which
> is fine for a "can't happen" scenario.
>
Isn't the existing assertion (Assert(!create || txn != NULL);) in
ReorderBufferTXNByXid() sufficient to handle this case?
> Looking closer, when we identify an XID as a subtransaction, we:
> - assign toplevel_xid,
> - set RBTXN_IS_SUBXACT, and
> - assign toptxn pointer.
>
> ISTM the 'toplevel_xid' and RBTXN_IS_SUBXACT are redundant.
> 'toplevel_xid' is only used in those two calls that do
> ReorderBufferTXNByXid(rb, txn->toplevel_xid,...), and you could replace
> those by following the 'toptxn' pointer directly. And RBTXN_IS_SUBXACT
> is redundant with (toptxn != NULL). So here's a patch to remove
> 'toplevel_xid' and RBTXN_IS_SUBXACT altogether.
>
Good idea. I don't see a problem with this idea.
@@ -1135,8 +1133,6 @@ static void
ReorderBufferTransferSnapToParent(ReorderBufferTXN *txn,
ReorderBufferTXN *subtxn)
{
- Assert(subtxn->toplevel_xid == txn->xid);
Is there a benefit in converting this assertion using toptxn?
--
With Regards,
Amit Kapila.
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2024-04-15 06:31:59 | Re: Why is parula failing? |
Previous Message | Michael Paquier | 2024-04-15 06:12:19 | Re: UUID v7 |