Andrew,
Thank you for the fix, I'll take a look at it
-Dmitry (current SA owner)
On 2015-04-28 21:27, Andrew Haley wrote:
> I've been testing serviceability support for AArch64, and I keep getting
> bad pointers and NullPointerExceptions.
>
> I thought it was a bug in my back-end, but it's not. The problem is
> that we read a Symbol pointer out of the constant pool and then try to
> use it. Unfortunately, the bottom two bits of a symbol pointer in the
> constant pool have special meanings and we must mask them before use.
>
> This is my fix:
>
> diff -r 11b7f6b12521 agent/src/share/classes/sun/jvm/hotspot/oops/Symbol.java
> --- a/agent/src/share/classes/sun/jvm/hotspot/oops/Symbol.java Fri Apr 24
> 16:28:29 2015 +0100
> +++ b/agent/src/share/classes/sun/jvm/hotspot/oops/Symbol.java Tue Apr 28
> 19:23:18 2015 +0100
> @@ -43,6 +43,8 @@
> });
> }
>
> + static class TagBits { final static int _resolved_value = 0, _symbol_bit
> = 1, _pseudo_bit = 2, _symbol_mask = 3; }
> +
> private static synchronized void initialize(TypeDataBase db) throws
> WrongTypeException {
> Type type = db.lookupType("Symbol");
> length = type.getCIntegerField("_length");
> @@ -60,7 +62,8 @@
> if (addr == null) {
> return null;
> }
> - return new Symbol(addr);
> + return new Symbol(addr.andWithMask(~ TagBits._symbol_mask));
> }
>
> I'm not at all sure this is the right place to fix it, but it works.
>
> I'm just really surprised no-one noticed this before.
>
> Andrew.
>
--
Dmitry Samersoff
Oracle Java development team, Saint Petersburg, Russia
* I would love to change the world, but they won't give me the sources.
