Minor binary-search int overflow in timezone code
Hi,
a fellow Debian Developer found a minor glitch in
src/timezone/localtime.c, where binary search is used. Now I don't
think there is an actual problem (unless there's > 2^30 timezones),
but it would at least make sense to mark the code as okayish so that
people running code scanners won't stumble over the issue again.
The attached patch added comments to address this.
Date: Sun, 30 Nov 2014 22:06:42 +0100
From: Niels Thykier <niels@thykier.net>
Reply-To: Niels Thykier <niels@thykier.net>, 771580@bugs.debian.org
To: Debian Bug Tracking System <submit@bugs.debian.org>
Subject: [Pkg-postgresql-public] Bug#771580: postgresql-9.4: Minor binary-search
int overflow
Source: postgresql-9.4
Version: 9.4~rc1-1
Severity: minor
Hi,
I stumbled on the folowing snippet from src/timezone/localtime.c,
function pg_interpret_timezone_abbrev:
{
int lo = 0;
int hi = sp->timecnt;
while (lo < hi)
{
int mid = (lo + hi) >> 1;
^^^^^^^
This looks it is subject to a known int overflow, when (original) hi
is close to INT_MAX and the item being close to then end of the array.
~Niels
[The original report had a link here to the googleresearch blog , but
the PG list servers think it is spam :(]
Attachments:
timezone-binary-search.patchtext/x-diff; charset=us-asciiDownload
diff --git a/src/timezone/localtime.c b/src/timezone/localtime.c
new file mode 100644
index 19a24e1..878e471
*** a/src/timezone/localtime.c
--- b/src/timezone/localtime.c
*************** localsub(const pg_time_t *timep, long of
*** 1070,1076 ****
while (lo < hi)
{
! int mid = (lo + hi) >> 1;
if (t < sp->ats[mid])
hi = mid;
--- 1070,1076 ----
while (lo < hi)
{
! int mid = (lo + hi) >> 1; /* overflow unlikely */
if (t < sp->ats[mid])
hi = mid;
*************** pg_next_dst_boundary(const pg_time_t *ti
*** 1423,1429 ****
while (lo < hi)
{
! int mid = (lo + hi) >> 1;
if (t < sp->ats[mid])
hi = mid;
--- 1423,1429 ----
while (lo < hi)
{
! int mid = (lo + hi) >> 1; /* overflow unlikely */
if (t < sp->ats[mid])
hi = mid;
*************** pg_interpret_timezone_abbrev(const char
*** 1506,1512 ****
while (lo < hi)
{
! int mid = (lo + hi) >> 1;
if (t < sp->ats[mid])
hi = mid;
--- 1506,1512 ----
while (lo < hi)
{
! int mid = (lo + hi) >> 1; /* overflow unlikely */
if (t < sp->ats[mid])
hi = mid;
Christoph Berg <cb@df7cb.de> writes:
a fellow Debian Developer found a minor glitch in
src/timezone/localtime.c, where binary search is used. Now I don't
think there is an actual problem (unless there's > 2^30 timezones),
but it would at least make sense to mark the code as okayish so that
people running code scanners won't stumble over the issue again.
The attached patch added comments to address this.
This is totally silly. The timecnt couldn't be anywhere near INT_MAX (in
fact, it is not allowed to exceed TZ_MAX_TIMES, which is currently just
1200). And there are bunches of other instances of similar code in PG;
shall we put equally wishy-washy comments on them all?
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Re: Tom Lane 2014-12-15 <21813.1418655100@sss.pgh.pa.us>
This is totally silly. The timecnt couldn't be anywhere near INT_MAX (in
fact, it is not allowed to exceed TZ_MAX_TIMES, which is currently just
1200). And there are bunches of other instances of similar code in PG;
shall we put equally wishy-washy comments on them all?
Well, if it's not interesting, let's just forget it. Sorry.
Christoph
--
cb@df7cb.de | http://www.df7cb.de/
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
On 12/15/14, 1:39 PM, Christoph Berg wrote:
Re: Tom Lane 2014-12-15 <21813.1418655100@sss.pgh.pa.us>
This is totally silly. The timecnt couldn't be anywhere near INT_MAX (in
fact, it is not allowed to exceed TZ_MAX_TIMES, which is currently just
1200). And there are bunches of other instances of similar code in PG;
shall we put equally wishy-washy comments on them all?Well, if it's not interesting, let's just forget it. Sorry.
At the risk of sticking my head in the lions mouth... this is the kind of response that deters people from contributing anything to the project, including reviewing patches. A simple "thanks, but we feel it's already clear enough that there can't be anywhere close to INT_MAX timezones" would have sufficed.
--
Jim Nasby, Data Architect, Blue Treble Consulting
Data in Trouble? Get it in Treble! http://BlueTreble.com
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Jim Nasby <Jim.Nasby@BlueTreble.com> writes:
On 12/15/14, 1:39 PM, Christoph Berg wrote:
Well, if it's not interesting, let's just forget it. Sorry.
At the risk of sticking my head in the lions mouth... this is the kind of response that deters people from contributing anything to the project, including reviewing patches. A simple "thanks, but we feel it's already clear enough that there can't be anywhere close to INT_MAX timezones" would have sufficed.
Yeah, I need to apologize. I was a bit on edge today due to the release
wrap (which you may have noticed wasn't going too smoothly), and should
not have responded like that.
Having said that, though, the submission wasn't carefully thought through
either. That problem was either not-an-issue or a potential security bug,
and if the submitter hadn't taken the time to be sure which, reporting it
in a public forum wasn't the way to proceed.
I also remain curious as to what sort of tool would complain about this
particular code and not the N other nearly-identical binary-search loops
in the PG sources, most of which deal with data structures potentially
far larger than the timezone data ...
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
Re: Tom Lane 2014-12-16 <14615.1418694505@sss.pgh.pa.us>
Jim Nasby <Jim.Nasby@BlueTreble.com> writes:
On 12/15/14, 1:39 PM, Christoph Berg wrote:
Well, if it's not interesting, let's just forget it. Sorry.
At the risk of sticking my head in the lions mouth... this is the kind of response that deters people from contributing anything to the project, including reviewing patches. A simple "thanks, but we feel it's already clear enough that there can't be anywhere close to INT_MAX timezones" would have sufficed.
Yeah, I need to apologize. I was a bit on edge today due to the release
wrap (which you may have noticed wasn't going too smoothly), and should
not have responded like that.
Hi,
maybe I should apologize as well for submitting this right at the time
of the release...
I also remain curious as to what sort of tool would complain about this
particular code and not the N other nearly-identical binary-search loops
in the PG sources, most of which deal with data structures potentially
far larger than the timezone data ...
He said he found it in manual code review, not using a tool.
But anyway, I do agree this is a very minor issue and there's much
more interesting things to spend time on. I promise to send in more
severe security issues next time :)
Christoph
--
cb@df7cb.de | http://www.df7cb.de/
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers