[aarch64-port-dev ] JEP: 342: Limit Speculative Execution - on aarch64?

[Derek White]

Re-ping on this topic:

Does anyone have a definitive answer if any specific GCC compilation options are needed to provide mitigation for speculative execution on aarch64?

I've created issue JDK-8223042 "Introduce alternate jvm variant with speculative execution disabled for AARCH64" to track this issue (even if the answer is "there's no issue").

JEP 342 may be providing an x86 fix for JDK 13, and it would be good to understand if we have a problem or not.

Thanks, 
 - Derek

> -----Original Message-----
> From: aarch64-port-dev <aarch64-port-dev-bounces at openjdk.java.net> On
> Behalf Of White, Derek
> Sent: Tuesday, November 20, 2018 5:13 PM
> To: aarch64-port-dev at openjdk.java.net
> Subject: [aarch64-port-dev ] JEP: 342: Limit Speculative Execution - on
> aarch64?
> 
> External Email
> 
> Hi porters,
> 
> Re - JEP 342: Limit Speculative Execution
> https://bugs.openjdk.java.net/browse/JDK-8207206
> 
> JEP 342 proposes that we simply build an alternative "non-speculative" JVM
> by compiling with gcc options:
>     -mindirect-branch=thunk -mfunction-return=thunk -mindirect-branch-
> register
> 
> These options are x86 only. I've asked Jesper to update the JEP to mention
> this.
> 
> I don't have a firm mapping in my mind between each Spectre variant and
> each mitigation, and the equivalency (or not) between the x86 and aarch64
> mitigations, and what needs to be handled at user-level vs. in the kernel.
> 
> But it seems that JEP 342 is trying to mitigate Spectre variant 2, but aarch64
> handles Spectre Variant 2 purely in kernel/firmware(?).
> 
> Are we safe then? Or do want to do anything special for building a non-
> speculative JVM for aarch64? Is the "-mtrack-speculation" option
> appropriate? Sufficient?
> 
> Thanks,
> 
>  - Derek
> 
> BTW - The JEP also mentions that the JIT compilers will need to be updated
> similarly, but no details are given. This will also be cpu-specific (and
> potentially more work).
> 
> -----Original Message-----
> From: White, Derek
> Sent: Tuesday, November 20, 2018 4:43 PM
> To: jdk-dev at openjdk.java.net; jesper.wilhelmsson at oracle.com
> Subject: RE: JEP: 342: Limit Speculative Execution
> 
> Hi Jesper,
> 
> [Sorry for the "timely" response.]
> 
> Although the mechanisms for managing and identifying an alternative JVM
> are not platform-specific, the mitigation strategy proposed (e.g. the specific
> gcc options) are x86-specific.
> 
> I suggest adding a note in the JEP that the specific mitigations proposed are
> x86 specific, and that other ports will need to identify what mitigations are
> appropriate (if any).
> 
> I'll start up a discussion in the aarch64 list on what we should do about
> limiting speculative execution, and we may link aarch64-specific
> enhancements to this JEP.  This needs to be looked at for both C++ options
> and JIT implementation on aarch64.
> 
> Thanks,
> 
>  - Derek
> 
> > -----Original Message-----
> > From: jdk-dev <jdk-dev-bounces at openjdk.java.net> On Behalf Of
> > mark.reinhold at oracle.com
> > Sent: Friday, September 07, 2018 1:53 PM
> > To: jesper.wilhelmsson at oracle.com
> > Cc: jdk-dev at openjdk.java.net
> > Subject: New candidate JEP: 342: Limit Speculative Execution
> >
> > External Email
> >
> > http://openjdk.java.net/jeps/342
> >
> > - Mark