Bug in is_setting_search_path

From: Grant Shirreffs <GShirreffs(at)stayinfront(dot)com>
To: "pgsql-odbc(at)lists(dot)postgresql(dot)org" <pgsql-odbc(at)lists(dot)postgresql(dot)org>
Subject: Bug in is_setting_search_path
Date: 2018-01-11 00:17:07
Message-ID: BN6PR07MB3009577DBC0019D3C0438064CF160@BN6PR07MB3009.namprd07.prod.outlook.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-odbc

Hello,

I have found a bug in the is_setting_search_path function (connection.c line 1597).

The search loop is currently:

for (; *q; q++)
{
if (IS_NOT_SPACE(*q))
{
if (strnicmp(q, "search_path", 11) == 0)
return TRUE;
q++;
while (IS_NOT_SPACE(*q))
q++;
}
}

The inner while(IS_NOT_SPACE(*q)) loop will terminate if a null is reached. The loop variable will then be further incremented by the "for" loop, to point beyond the null terminator, and so the loop will continue, until by chance two nulls are encountered. If two nulls are not found, then eventually the loop will reach the end of the memory page, and cause an access violation. Note that if the string "search_path" exists in memory beyond the end of the statement, a false positive results from this function.

The fix is to remove the increment from the "for" loop, and move it instead to the false path of the "if":

for (; *q;)
{
if (IS_NOT_SPACE(*q))
{
if (strnicmp(q, "search_path", 11) == 0)
return TRUE;
q++;
while (IS_NOT_SPACE(*q))
q++;
}
else
q++
}

This issue has been causing occasional access violations in our code (when calling SET LC_TIME=''). We are currently testing with a fixed version, which is giving no other problems so far.

Please advise me if there is some other way I should submit this change for review and inclusion.

Thankyou

Grant Shirreffs
Principal Developer
StayinFront Inc

Responses

Browse pgsql-odbc by date

  From Date Subject
Next Message Inoue, Hiroshi 2018-01-11 21:47:53 Re: Bug in is_setting_search_path
Previous Message Dave Cramer 2018-01-10 20:44:30 Re: psqlODBC 10.01.0000 Released -- what about 9.6.6?