On Fri, 11 Mar 2022 10:44:25 GMT, Thomas Stuefe <stu...@openjdk.org> wrote:
> > On 3/11/22 10:12, Thomas Stuefe wrote: We could also redefine SafeFetch on > > MacOS/AArch64 to not need WX. We could do this by statically generating > > SafeFetch on that platform, and it wouldn't be in the JIT region at all. > > Why not just do that? Do you mean using inline assembly? > > I'd use out-of-line assembly, as I do for atomic compare-and-swap on linux: > > https://github.com/openjdk/jdk/blob/master/src/hotspot/os_cpu/linux_aarch64/atomic_linux_aarch64.S > > But I guess inline would work. > > Oh, this is neat. It would work on all platforms too, or on all we care to > implement it for. And it would nicely solve the initialization window problem > since it would work before stub routines are generated. We could throw > `CanUseSafeFetch` away. > > It seems we already do static assembly on bsd aarch. So there is already a > path to follow. > > But this could also be done as a follow up enhancement. I still like the OS > TLS variable idea. I spent some time doing a static implementation of SafeFetch on Linux x64, and its not super trivial. The problem is that we need to know addresses of instructions inside that function. I can set labels in assembly, and I can export them, but so far I have been unable to use them as addresses in C++ code. I will research some more. ------------- PR: https://git.openjdk.java.net/jdk/pull/7727
