From: Tom Lane <tgl@sss.pgh.pa.us>
To: Daniele Varrazzo <daniele.varrazzo@gmail.com>
cc: pgsql-bugs <pgsql-bugs@postgresql.org>
Subject: Re: [PATCH] Fix segfault calling PQflush on invalid connection
In-reply-to: 
 <CA+mi_8Zm_mVVyW1iNFgyMd9Oh0Nv8-F+7Y3-BqwMgTMHuo_h2Q@mail.gmail.com>
References: 
 <CA+mi_8Zm_mVVyW1iNFgyMd9Oh0Nv8-F+7Y3-BqwMgTMHuo_h2Q@mail.gmail.com>
Comments: In-reply-to Daniele Varrazzo <daniele.varrazzo@gmail.com>
	message dated "Mon, 15 Aug 2022 02:44:55 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <819196.1660527847.1@sss.pgh.pa.us>
Date: Sun, 14 Aug 2022 21:44:07 -0400
Message-ID: <819197.1660527847@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/819197.1660527847%40sss.pgh.pa.us>
Precedence: bulk

Daniele Varrazzo <daniele.varrazzo@gmail.com> writes:
> PQflush calls pqFlush, which performs struct access to the connection
> without checking if it's valid, resulting in a segfault if called with
> a null pointer.

> Please find attached a patch adding a guard to PQflush().

Seems reasonable, but this tickled a thought that's been in my
hindbrain for awhile: just checking for a null pointer is not
much of a check for being passed a valid PGconn pointer.  Should
we add a magic number to struct PGconn, and modify all libpq's
entry points along the lines of

	if (!conn || conn->magic != PGCONN_MAGIC)
		return failure;

I'm honestly not entirely sure if this is worth the trouble;
I've not heard of many application bugs that this would've caught.
But the lack of any such check does seem like it's not up to
modern standards.

			regards, tom lane