8340491: Thread stack-base assertion should report which thread has the un-set stack #21102

dholmes-ora · 2024-09-20T02:25:19Z

Please review this simple enhancement to an assertion so we can identify which thread had the problem.

I had to move the function to the cpp file due to include file issues.

We are limited in what we can print due to this being used very early in the thread initialization process - in particular no ResourceMarks are possible so we can't print the name.

Testing:

tier 5 (used to debug JDK-8340401)
tiers 1-3 sanity

Thanks

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8340491: Thread stack-base assertion should report which thread has the un-set stack (Enhancement - P3)

Reviewers

Stefan Karlsson (@stefank - Reviewer) Review applies to e6358a57
Aleksey Shipilev (@shipilev - Reviewer)
Kevin Walls (@kevinjwalls - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/21102/head:pull/21102
$ git checkout pull/21102

Update a local copy of the PR:
$ git checkout pull/21102
$ git pull https://git.openjdk.org/jdk.git pull/21102/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 21102

View PR using the GUI difftool:
$ git pr show -t 21102

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/21102.diff

Webrev

Link to Webrev Comment

…he un-set stack

bridgekeeper · 2024-09-20T02:26:00Z

👋 Welcome back dholmes! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2024-09-20T02:27:11Z

@dholmes-ora This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8340491: Thread stack-base assertion should report which thread has the un-set stack

Reviewed-by: shade, kevinw, stefank

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 129 new commits pushed to the master branch:

988a531: 8340181: Shenandoah: Cleanup ShenandoahRuntime stubs
822a773: 8340605: Open source several AWT PopupMenu tests
6514aef: 8340419: ZGC: Create an UseLargePages adaptation of TestAllocateHeapAt.java
ae4d2f1: 8340621: Open source several AWT List tests
dd56990: 8340639: Open source few more AWT List tests
ade17ec: 8340560: Open Source several AWT/2D font and rendering tests
73ebb84: 8340721: Clarify special case handling of unboxedType and getWildcardType
ed140f5: 8341101: [ARM32] Error: ShouldNotReachHere() in TemplateInterpreterGenerator::generate_math_entry after 8338694
082125d: 8340404: CharsetProvider specification updates
a7bfced: 8337679: Memset warning in src/hotspot/share/adlc/adlArena.cpp
... and 119 more: https://git.openjdk.org/jdk/compare/fdc16a373459cb2311316448c765b1bee5c73694...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

openjdk · 2024-09-20T02:27:47Z

@dholmes-ora The following label will be automatically applied to this pull request:

hotspot-runtime

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2024-09-20T02:31:08Z

Webrevs

stefank

Looks fine, but could you explain when this is needed?

If it is the current thread that is crashes then we already print the thread info.

If it is crashing when called on another thread, it might be easier to add informative asserts in that sub-system instead. You could for example add a non-asserting getter and perform all relevant checks on the caller side.

Another alternative could be to leverage the VMErrorCallback functionality I introduced a little while ago. With that you can register a callback to be called by the hs_err printer if the current thread crashes. With that you can have much more context about what the calling code was doing and what information that thread has available. That would look something like this (~~I've not compiled this, so there might be errors~~ Edit: updated the code):

diff --git a/src/hotspot/share/nmt/memMapPrinter.cpp b/src/hotspot/share/nmt/memMapPrinter.cpp
index 5f920b102a9..a2aae39b0ca 100644
--- a/src/hotspot/share/nmt/memMapPrinter.cpp
+++ b/src/hotspot/share/nmt/memMapPrinter.cpp
@@ -44,6 +44,7 @@
 #include "runtime/vmThread.hpp"
 #include "utilities/globalDefinitions.hpp"
 #include "utilities/ostream.hpp"
+#include "utilities/vmError.hpp"

 // Note: throughout this code we will use the term "VMA" for OS system level memory mapping

@@ -164,8 +165,29 @@ class CachedNMTInformation : public VirtualMemoryWalker {

 /////// Thread information //////////////////////////

+class VMAInspectionErrorCallback : public VMErrorCallback {
+  const void* const   _from;
+  const void* const   _to;
+  const Thread* const _thread;
+
+public:
+  VMAInspectionErrorCallback(const void* from, const void* to, const Thread* thread) :
+    _from(from), _to(to), _thread(thread) {}
+
+  void call(outputStream* st) override {
+    st->print_cr("Crashing while printing VMA details for " PTR_FORMAT " " PTR_FORMAT " for thread %s with id: %d",
+                 p2i(_from),
+                 p2i(_to),
+                 _thread->name(),
+                 _thread->osthread() != nullptr ? _thread->osthread()->thread_id() : 0);
+  }
+};
+
 // Given a VMA [from, to) and a thread, check if vma intersects with thread stack
 static bool vma_touches_thread_stack(const void* from, const void* to, const Thread* t) {
+  VMAInspectionErrorCallback on_error(from, to, t);
+  VMErrorCallbackMark mark(&on_error);
+
   // Java thread stacks (and sometimes also other threads) have guard pages. Therefore they typically occupy
   // at least two distinct neighboring VMAs. Therefore we typically have a 1:n relationshipt between thread
   // stack and vma.

Note that in this context we do have a ResourceMark, so it should be fine to print the name.

dholmes-ora · 2024-09-20T06:00:15Z

Thanks for the review @stefank !

If it is crashing when called on another thread, it might be easier to add informative asserts in that sub-system instead. You could for example add a non-asserting getter and perform all relevant checks on the caller side.

The assert is firing when stack_base() is called on another thread. In that case we had nothing to tell us which thread it was, hence this enhancement to provide that information. We want this asserting getter to be used, and for the assert to fire .

src/hotspot/share/runtime/thread.cpp

shipilev · 2024-09-20T06:37:31Z

src/hotspot/share/runtime/thread.hpp

@@ -532,7 +532,7 @@ class Thread: public ThreadShadow {

 public:
  // Stack overflow support
-  address stack_base() const           { assert(_stack_base != nullptr,"Sanity check"); return _stack_base; }
+  address stack_base() const DEBUG_ONLY(;) NOT_DEBUG({ return _stack_base; })


I think it is a bit cleaner to "just" outline the checking method like:

thread.hpp: address stack_base() const { assert_stack_base(); return _stack_base; } void assert_stack_base() const NOT_DEBUG_RETURN; thread.cpp: #ifdef ASSERT void Thread::assert_stack_base() const { // Note: can't report Thread::name() here as that can require a ResourceMark which we // can't use because this gets called too early in the thread initialization. assert(_stack_base != nullptr, "Stack base not yet set for thread id:%d (0 if not set)", osthread() != nullptr ? osthread()->thread_id() : 0); } #endif

This is the way we do it in existing code, e.g.

jdk/src/hotspot/share/runtime/stackWatermark.hpp

Line 105 in 46b02f4

void assert_is_frame_safe(const frame& f) NOT_DEBUG_RETURN;

So many ways to achieve the same thing ...

I would not want assert_stack_base in the public API.

I did need to fix the ifdef though.

DEBUG_ONLY(;) just triggers me, sorry. We have NOT_DEBUG_RETURN exactly for this reason: mark the methods that are only defined in debug builds, and shortcut otherwise, without inventing new things :)

shipilev

Only have a nitpicking comment, see above ^

stefank · 2024-09-20T09:53:37Z

Thanks for the review @stefank !

If it is crashing when called on another thread, it might be easier to add informative asserts in that sub-system instead. You could for example add a non-asserting getter and perform all relevant checks on the caller side.

The assert is firing when stack_base() is called on another thread. In that case we had nothing to tell us which thread it was, hence this enhancement to provide that information.

That was fully understood and I think you are missing my point. I was trying to convey that it might be better to rewrite the calling code to perform its own checks / logging instead of relying on a "low-level" assert in Thread. 2c.

dholmes-ora · 2024-09-23T04:36:32Z

I was trying to convey that it might be better to rewrite the calling code to perform its own checks / logging instead of relying on a "low-level" assert in Thread.

The calling code is of course being fixed (or at least the cause dealt with one way or another). The point here was to simply enhance the existing assertion to be more informative for the next time it gets triggered.

stefank · 2024-09-23T08:11:20Z

I was trying to convey that it might be better to rewrite the calling code to perform its own checks / logging instead of relying on a "low-level" assert in Thread.

The calling code is of course being fixed (or at least the cause dealt with one way or another). The point here was to simply enhance the existing assertion to be more informative for the next time it gets triggered.

Right. And that was what I was curious to know more about because I think there would be better ways to get better diagnostics by changing code there. Anyways, I'll leave you to deal with that in a separate issue.

dholmes-ora · 2024-09-29T20:47:30Z

Anyone want to add the second review or shall I just abort this?

kevinjwalls

Looks good, and useful!

shipilev

Well, I am still nitpicking about DEBUG_ONLY(;), but I can live with it.

stefank · 2024-09-30T08:45:44Z

This PR is much more complicated than it needs to be.

We would probably be fine with always having the implementation in the .cpp file. I don't think that we need to optimize away the call for this getter. I don't see hot code using this function.
If we really want to optimize it away, we can just stick the assert in the header just like we already do with related functions:

  bool is_in_stack_range(address adr, address limit, bool inclusive) const {
    assert(stack_base() > limit && limit >= stack_end(), "limit is outside of stack");
    return stack_base() > adr && (inclusive ? adr >= limit : adr > limit);
  }

There's no point in trying to separate the debug/release builds into .cpp/.hpp files if the related functions already put the assert in the header.

dholmes-ora · 2024-09-30T09:06:03Z

we can just stick the assert in the header just like we already do with related functions:

@stefank as mentioned in the description I ran into include file issues referring to osthread()->thread_id() in thread.hpp, as we only have a forward decl of OSThread.

dholmes-ora · 2024-09-30T09:06:40Z

Thanks for the review @kevinjwalls !

stefank · 2024-09-30T11:17:09Z

we can just stick the assert in the header just like we already do with related functions:

@stefank as mentioned in the description I ran into include file issues referring to osthread()->thread_id() in thread.hpp, as we only have a forward decl of OSThread.

OK. Did you consider my bullet (1) at all?

stefank · 2024-09-30T11:25:16Z

we can just stick the assert in the header just like we already do with related functions:

@stefank as mentioned in the description I ran into include file issues referring to osthread()->thread_id() in thread.hpp, as we only have a forward decl of OSThread.

OK. Did you consider my bullet (1) at all?

I re-read the PR and see that this is what you started with and that Thomas requested this to be inlined.

stefank · 2024-09-30T11:27:57Z

FTR, I'm not blocking this, I'm just a little annoyed that we unnecessarily (IMHO) uglify the code.

dholmes-ora · 2024-09-30T22:05:18Z

Okay thanks for the reviews.

/integrate

openjdk · 2024-09-30T22:06:24Z

Going to push as commit 31858fc.
Since your change was applied there have been 142 commits pushed to the master branch:

bfdeb33: 8340332: Open source mixed AWT tests - Set3
a6b3188: 8337632: AES-GCM Algorithm optimization for x86_64
5586f83: 8341064: Define anchor point and index term for "wrapper classes"
4168faf: 8341100: Add index entries for terms used in java.lang.Class
f1bf469: 8341199: Use ClassFile's new API loadConstant(int)
860d49d: 8211400: nsk.share.gc.Memory::getArrayLength returns wrong value
cff420d: 8339711: ZipFile.Source.initCEN needlessly reads END header
180affc: 8320318: ObjectMonitor Responsible thread
e19c7d8: 8340874: Open source some of the AWT Geometry/Button tests
58b6fc5: 8341197: [BACKOUT] 8322770: Implement C2 VectorizedHashCode on AArch64
... and 132 more: https://git.openjdk.org/jdk/compare/fdc16a373459cb2311316448c765b1bee5c73694...master

Your commit was automatically rebased without conflicts.

openjdk · 2024-09-30T22:06:32Z

@dholmes-ora Pushed as commit 31858fc.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

8340491: Thread stack-base assertion should report which thread has t…

e6358a5

…he un-set stack

openjdk bot added the rfr Pull request is ready for review label Sep 20, 2024

openjdk bot added the hotspot-runtime hotspot-runtime-dev@openjdk.org label Sep 20, 2024

stefank approved these changes Sep 20, 2024

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Sep 20, 2024

tstuefe reviewed Sep 20, 2024

View reviewed changes

src/hotspot/share/runtime/thread.cpp Show resolved Hide resolved

Keep inline version for release builds

c887fdf

openjdk bot removed the ready Pull request is ready to be integrated label Sep 20, 2024

shipilev reviewed Sep 20, 2024

View reviewed changes

Need ifdef ASSERT

7ebd2cc

kevinjwalls approved these changes Sep 30, 2024

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Sep 30, 2024

shipilev approved these changes Sep 30, 2024

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label Sep 30, 2024

openjdk bot closed this Sep 30, 2024

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Sep 30, 2024

dholmes-ora deleted the 8340491-stack-assert branch September 30, 2024 22:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

8340491: Thread stack-base assertion should report which thread has the un-set stack #21102

8340491: Thread stack-base assertion should report which thread has the un-set stack #21102

dholmes-ora commented Sep 20, 2024 •

edited by openjdk bot

Loading

bridgekeeper bot commented Sep 20, 2024

openjdk bot commented Sep 20, 2024 •

edited

Loading

openjdk bot commented Sep 20, 2024

mlbridge bot commented Sep 20, 2024 •

edited

Loading

stefank left a comment •

edited

Loading

dholmes-ora commented Sep 20, 2024

shipilev Sep 20, 2024

shipilev Sep 20, 2024

dholmes-ora Sep 20, 2024

shipilev Sep 20, 2024 •

edited

Loading

shipilev left a comment •

edited

Loading

stefank commented Sep 20, 2024

dholmes-ora commented Sep 23, 2024

stefank commented Sep 23, 2024

dholmes-ora commented Sep 29, 2024

kevinjwalls left a comment

shipilev left a comment

stefank commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

stefank commented Sep 30, 2024

stefank commented Sep 30, 2024

stefank commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

openjdk bot commented Sep 30, 2024

openjdk bot commented Sep 30, 2024

8340491: Thread stack-base assertion should report which thread has the un-set stack #21102

8340491: Thread stack-base assertion should report which thread has the un-set stack #21102

Conversation

dholmes-ora commented Sep 20, 2024 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Reviewing

Webrev

bridgekeeper bot commented Sep 20, 2024

openjdk bot commented Sep 20, 2024 • edited Loading

openjdk bot commented Sep 20, 2024

mlbridge bot commented Sep 20, 2024 • edited Loading

Webrevs

stefank left a comment • edited Loading

Choose a reason for hiding this comment

dholmes-ora commented Sep 20, 2024

shipilev Sep 20, 2024

Choose a reason for hiding this comment

shipilev Sep 20, 2024

Choose a reason for hiding this comment

dholmes-ora Sep 20, 2024

Choose a reason for hiding this comment

shipilev Sep 20, 2024 • edited Loading

Choose a reason for hiding this comment

shipilev left a comment • edited Loading

Choose a reason for hiding this comment

stefank commented Sep 20, 2024

dholmes-ora commented Sep 23, 2024

stefank commented Sep 23, 2024

dholmes-ora commented Sep 29, 2024

kevinjwalls left a comment

Choose a reason for hiding this comment

shipilev left a comment

Choose a reason for hiding this comment

stefank commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

stefank commented Sep 30, 2024

stefank commented Sep 30, 2024

stefank commented Sep 30, 2024

dholmes-ora commented Sep 30, 2024

openjdk bot commented Sep 30, 2024

openjdk bot commented Sep 30, 2024

dholmes-ora commented Sep 20, 2024 •

edited by openjdk bot

Loading

openjdk bot commented Sep 20, 2024 •

edited

Loading

mlbridge bot commented Sep 20, 2024 •

edited

Loading

stefank left a comment •

edited

Loading

shipilev Sep 20, 2024 •

edited

Loading

shipilev left a comment •

edited

Loading