RFR: 8253495: CDS generates non-deterministic output

[Ioi Lam]

On Fri, 11 Mar 2022 05:59:00 GMT, David Holmes <david.holmes at oracle.com> wrote:

> > I ended up changing `os::malloc()` to zero the buffer when running with -Xshare:dump. Hopefully one extra check of `if (DumpSharedSpaces)` doesn't matter too much for regular VM executions because `os::malloc()` already has a high overhead.
> 
> This is raising red flags for me sorry. Every user of the JDK is now paying a penalty because of something only needed when dumping the shared archive. It might not be much but it is the old "death by a thousand cuts". Is there any way to tell the OS to pre-zero all memory provided to the current process, such that we could set that when dumping and not have to check on each allocation?

I don't know how to tell the OS (or C library) to zero out the buffer returned by malloc. However, in the current code path, we already have a test for an uncommon condition when `os::malloc()` calls `MemTracker::record_malloc()` which calls `MallocTracker::record_malloc()`


void* MallocTracker::record_malloc(void* malloc_base, size_t size, MEMFLAGS flags,
  const NativeCallStack& stack)
{
  if (MemTracker::tracking_level() == NMT_detail) {
    MallocSiteTable::allocation_at(stack, size, &mst_marker, flags);
  }


I can combine the tests for `MemTracker::tracking_level()` and `DumpSharedSpaces` into a single test and do more work only when the uncommon path is taken. This would require some refactoring of the MemTracker/MallocTracker code. I'd rather do that in a separate RFE.

In fact, `MemTracker::_tracking_level` is tested twice in the current implementation. We can change it to do a single test in the most common case (NMT_summary) if we really want to cut down the number of tests. But honestly I don't think this makes any difference.

> And I have to wonder how easy it would be to re-introduce non-deterministic values in these data structures that are being dumped. Does malloc itself even guarantee to return the same set of addresses for the same sequence of requests in different executions of a program?

The malloc'ed objects are copied into the CDS archive at deterministic addresses. Any pointers inside such objects will be relocated.

-------------

PR: https://git.openjdk.java.net/jdk/pull/7748