Обсуждение: Schema Usage Patterns vs. dbowner or CREATEROLE

When I updated
https://www.postgresql.org/docs/devel/ddl-schemas.html#DDL-SCHEMAS-PATTERNS
for CVE-2018-1058, I wrote that some schema usage patterns are secure against
CREATEROLE users and database owners.  That was incorrect.  Even with the
first pattern, a database owner can attack the database's users via "CREATE
SCHEMA trojan; ALTER DATABASE $mydb SET search_path = trojan, public;".  A
CREATEROLE user can issue "GRANT $dbowner TO $me" and then use the database
owner attack.  I plan to update this section to assert that all three usage
patterns are equally vulnerable to dbowner and CREATEROLE attack.  (The
technique described under "If untrusted users have access to a database ..."
in libpq.sgml remains secure.)

On Sat, Oct 12, 2019 at 06:35:12PM -0700, Noah Misch wrote:
> When I updated
> https://www.postgresql.org/docs/devel/ddl-schemas.html#DDL-SCHEMAS-PATTERNS
> for CVE-2018-1058, I wrote that some schema usage patterns are secure against
> CREATEROLE users and database owners.  That was incorrect.  Even with the
> first pattern, a database owner can attack the database's users via "CREATE
> SCHEMA trojan; ALTER DATABASE $mydb SET search_path = trojan, public;".  A
> CREATEROLE user can issue "GRANT $dbowner TO $me" and then use the database
> owner attack.  I plan to update this section to assert that all three usage
> patterns are equally vulnerable to dbowner and CREATEROLE attack.  (The
> technique described under "If untrusted users have access to a database ..."
> in libpq.sgml remains secure.)

I'm attaching a patch for that.  Two of the patterns became so similar that I
merged them.