On 5/22/19 3:33 PM, Arthur Eubanks wrote:


On Wed, May 22, 2019 at 12:12 PM Sean Mullan <sean.mul...@oracle.com <mailto:sean.mul...@oracle.com>> wrote:

    On 5/22/19 1:28 PM, Arthur Eubanks wrote:
     > On Wed, May 22, 2019 at 7:13 AM Daniel Fuchs
    <daniel.fu...@oracle.com <mailto:daniel.fu...@oracle.com>
     > <mailto:daniel.fu...@oracle.com
    <mailto:daniel.fu...@oracle.com>>> wrote:
     >
     >     Hi Arthur,
     >
     >         18 // For IPSupport
     >         19 grant {
     >         20     permission java.net.SocketPermission "localhost:0",
     >     "listen,resolve";
     >         21     permission java.util.PropertyPermission
     >     "java.net.preferIPv4Stack", "read";
     >         22 };
     >
     >     It might be better if these permissions were granted to the
     >     library only.
     >
     > Done.

    Have you tested that with jtreg? I believe it may not work because of
    the way the SecurityManager is enabled inside the test (rather than
    using the jtreg java.security.policy option). You may find that you
    also
    need to grant those permissions to jtreg.jar since it is higher in the
    call stack. If that is the case, you are probably better off granting
    the permissions to all code, or restructuring the test to use the jtreg
    java.security.policy option, where jtreg installs its own
    SecurityManager to grant itself the proper permissions. However, that
    will require some code changes and granting some additional permissions
    to the test that are needed (for adding a security provider, etc)
    before
    it currently enables a SM. And that is probably more than you want
    to do
    for this fix.

    --Sean

Tried it directly with jtreg
$ ~/jtreg/build/images/jtreg/bin/jtreg -jdk:./images/jdk/ ../test/jdk/java/security/SecureClassLoader/DefineClass.java and it passes. Verified that removing the newly added permissions makes it fail again. There was some discussion on IPSupport and SecurityManagers when IPSupport was first introduced: https://markmail.org/message/vvemfm367ja3qllj

Ok, the fix looks good then to me.

--Sean

Reply via email to