Issue with some calls to GetMultiXactIdMembers()

From: Greg Nancarrow <gregn4422(at)gmail(dot)com>
To: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Issue with some calls to GetMultiXactIdMembers()
Date: 2021-06-16 10:22:46
Message-ID: CAJcOf-dO5kVosn+bS4EnYKUnjSZOa7f6duE2TRraP92qvk1piw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

There's a couple of calls to GetMultiXactIdMembers() in heapam.c which
subsequently pfree() the returned "members" pointer (pass-by-reference
parameter) if it's non-NULL.
However, there's an error return within GetMultiXactIdMembers() that
returns -1 without NULLing out "members", and the callers have simply
allocated that pointer on the stack without initializing it to NULL.
If that error condition were to ever happen, pfree() would likely be
called with a junk value.
Also note that there's another error return (about 15 lines further
down) in GetMultiXactIdMembers() that returns -1 and does NULL out
"members", so the handling is inconsistent.
The attached patch adds the NULLing out of the "members" pointer in
the first error case, to fix that and guard against possible pfree()
on error by such callers.

I also note that there are other callers which pfree() "members" based
on the returned "nmembers" value, and this is also inconsistent.
Some pfree() "members" if nmembers>= 0, while others pfree() it if nmembers>0.
After looking at the code for a while, it looks like the "nmembers ==
0" case can't actually happen (right?). I decided not to mess with any
of the calling code.

Regards,
Greg Nancarrow
Fujitsu Australia

Attachment Content-Type Size
0001-Fix-possible-pfree-of-junk-members-value.patch application/octet-stream 932 bytes

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Yagiz Nizipli 2021-06-16 10:48:02 Re: [PATCH] rename column if exists
Previous Message Amit Kapila 2021-06-16 10:20:57 Re: locking [user] catalog tables vs 2pc vs logical rep