Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

8265753: Remove manual JavaThread transitions to blocked #3875

Closed
wants to merge 10 commits into from

Conversation

robehn
Copy link
Contributor

@robehn robehn commented May 5, 2021

Please consider this change which removes the manual transitions to blocked.
This adds a preprocess template/functor which is executed in the destructor of 'ThreadBlockInVM' if we are going to do any processing.
This gives us a way to backout of the object/raw monitor before suspend or other processing, such as a safepoint.

The object monitor code could be straight forward changed to use this instead of manual transitions.

Raw monitors on the other hand are a bit more complicated due to 'implicit' rules (consequences of the specs).
Added a comment in hpp trying to explain it; we cannot simply transition with a raw monitor held.
This caused the change in the destructor ~ThreadInVMfromNative() (this specific change have also been tested in unrelated exploration of transition), now this RAII does the same as we do when going to native from Java, just setting the state.
Since we are going from an unsafe state, in VM, to a safe state, in native, we do not need to check the poll.
That made it possible to careful use ThreadInVMfromNative in raw monitors.

I also remove the early CAS in raw_enter.
We lock a lock to do a CAS, in the uncontended case means CAS on lock then CAS raw monitor.
Now we instead do a transitions, in the uncontended case means fence, CAS raw monitor, fence.
(multiple fence (CAS is also a fence) very close to each other have little additional performance impact on contemporary hardware)

Passes t1-t7 and manual stressing relevant test groups.


Progress

  • Change must not contain extraneous whitespace
  • Commit message must refer to an issue
  • Change must be properly reviewed

Issue

  • JDK-8265753: Remove manual JavaThread transitions to blocked

Reviewers

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.java.net/jdk pull/3875/head:pull/3875
$ git checkout pull/3875

Update a local copy of the PR:
$ git checkout pull/3875
$ git pull https://git.openjdk.java.net/jdk pull/3875/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 3875

View PR using the GUI difftool:
$ git pr show -t 3875

Using diff file

Download this PR as a diff file:
https://git.openjdk.java.net/jdk/pull/3875.diff

@bridgekeeper
Copy link

bridgekeeper bot commented May 5, 2021

👋 Welcome back rehn! 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
Copy link

openjdk bot commented May 5, 2021

@robehn The following labels will be automatically applied to this pull request:

  • hotspot
  • serviceability

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

@openjdk openjdk bot added serviceability serviceability-dev@openjdk.org hotspot hotspot-dev@openjdk.org labels May 5, 2021
@robehn
Copy link
Contributor Author

robehn commented May 5, 2021

/label remove hotspot
/label add hotspot-runtime

@openjdk openjdk bot removed the hotspot hotspot-dev@openjdk.org label May 5, 2021
@openjdk
Copy link

openjdk bot commented May 5, 2021

@robehn
The hotspot label was successfully removed.

@openjdk openjdk bot added the hotspot-runtime hotspot-runtime-dev@openjdk.org label May 5, 2021
@openjdk
Copy link

openjdk bot commented May 5, 2021

@robehn
The hotspot-runtime label was successfully added.

@robehn robehn marked this pull request as ready for review May 5, 2021 11:49
@openjdk openjdk bot added the rfr Pull request is ready for review label May 5, 2021
@mlbridge
Copy link

mlbridge bot commented May 5, 2021

Copy link
Member

@dcubed-ojdk dcubed-ojdk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thumbs up on the over all logic. I only have minor nits and suggestions.

src/hotspot/share/prims/jvmtiRawMonitor.cpp Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.cpp Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/interfaceSupport.inline.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.cpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.cpp Show resolved Hide resolved
@openjdk
Copy link

openjdk bot commented May 11, 2021

@robehn 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:

8265753: Remove manual JavaThread transitions to blocked

Reviewed-by: dcubed, rrich, dholmes, pchilanomate

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 28 new commits pushed to the master branch:

  • de91643: 8267611: Print more info when pointer_delta assert fails
  • a4c46e1: 8263202: Update Hebrew/Indonesian/Yiddish ISO 639 language codes to current
  • 9c346a1: 8266963: Remove safepoint poll introduced in 8262443 due to reentrance issue
  • 45e0597: 8264302: Create implementation for Accessibility native peer for Splitpane java role
  • 4343997: 8267708: Remove references to com.sun.tools.javadoc.**
  • f632254: 8267221: jshell feedback is incorrect when creating method with array varargs parameter
  • bf8d4a8: 8267583: jmod fails on symlink to class file
  • 083416d: 8267130: Memory Overflow in Disassembler::load_library
  • 9d305b9: 8252372: Check if cloning is required to move loads out of loops in PhaseIdealLoop::split_if_with_blocks_post()
  • 0394416: 8267468: Rename refill waster counters in ThreadLocalAllocBuffer
  • ... and 18 more: https://git.openjdk.java.net/jdk/compare/31d0f0d895ef4039d2e96a8fb6e990e93eed4d41...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 openjdk bot added the ready Pull request is ready to be integrated label May 11, 2021
@robehn
Copy link
Contributor Author

robehn commented May 12, 2021

Thumbs up on the over all logic. I only have minor nits and suggestions.

Thank you!

Copy link
Member

@reinrich reinrich left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi Robbin,

I haven't found the time for a proper review yet but I've experimented a little bit with lambdas. I could not make it work because g++ created references to ::new which isn't allowed.

Thanks, Richard.

@robehn
Copy link
Contributor Author

robehn commented May 12, 2021

Hi Robbin,

I haven't found the time for a proper review yet but I've experimented a little bit with lambdas. I could not make it work because g++ created references to ::new which isn't allowed.

Thanks, Richard.

Hi Richard,
I tested lamdba, which is just a fancy way to write a crazy typed functor, we need to capture the lamdba so we can run it in the destructor. AFAICT the way to do that is using std::function.

Regarding ThreadClosure, we could use it, maybe that is preferable?!

@reinrich
Copy link
Member

Hi Robbin,
I haven't found the time for a proper review yet but I've experimented a little bit with lambdas. I could not make it work because g++ created references to ::new which isn't allowed.
Thanks, Richard.

Hi Richard,
I tested lamdba, which is just a fancy way to write a crazy typed functor, we need to capture the lamdba so we can run it in the destructor. AFAICT the way to do that is using std::function.

Yes, I'd think so too.

Regarding ThreadClosure, we could use it, maybe that is preferable?!

Personally I'd prefer it but I'm fine with the current version too.

Copy link
Member

@dholmes-ora dholmes-ora left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi Robbin,

Sorry for the delay in getting through this.

Overall approach looks good. I have a few queries below and some requested naming changes to make things clearer.

Thanks,
David

src/hotspot/share/prims/jvmtiRawMonitor.hpp Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.cpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.hpp Outdated Show resolved Hide resolved
Copy link
Member

@reinrich reinrich left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi Robbin,

there seem to be issues in the jvmtiRawMonitor part of the change. Besides that it looks good.

Cheers, Richard.

src/hotspot/share/prims/jvmtiRawMonitor.cpp Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.cpp Outdated Show resolved Hide resolved
src/hotspot/share/prims/jvmtiRawMonitor.hpp Outdated Show resolved Hide resolved
Comment on lines 383 to 386
_recursions = 0;
_waiters++;
ret = simple_wait(self, millis);
_waiters--;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We don't own the monitor yet so you cannot modify _waiters here.
Any reason you moved and duplicated this block?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good catch! Even if the block wasn't moved we still no longer own the monitor after simple_wait.

This is going to be tricky to fix in a clear/clean way as the waiters count has to be maintained properly.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually the field is unused.
The _entry_list is the list of waiters, which is the only thing we really care about.
If there was a reader of _waiter reading it without lock can only be used as some kind of hint.
If you have lock you can just check _entry_list.

I duplicated it because otherwise I would need two branchse on if (self->is_Java_thread()) which creates more complicated code.

@mlbridge
Copy link

mlbridge bot commented May 14, 2021

Mailing list message from David Holmes on hotspot-runtime-dev:

On 12/05/2021 8:56 pm, Robbin Ehn wrote:

On Wed, 12 May 2021 08:27:33 GMT, Richard Reingruber <rrich at openjdk.org> wrote:

Hi Robbin,

I haven't found the time for a proper review yet but I've experimented a little bit with lambdas. I could not make it work because g++ created references to ::new which isn't allowed.

Thanks, Richard.

Hi Richard,
I tested lamdba, which is just a fancy way to write a crazy typed functor, we need to capture the lamdba so we can run it in the destructor. AFAICT the way to do that is using std::function.

Regarding ThreadClosure, we could use it, maybe that is preferable?!

Isn't a ThreadClosure for applying an operation to a set of threads?
That is not what we are doing here.

David

@reinrich
Copy link
Member

Mailing list message from David Holmes on hotspot-runtime-dev:

On 12/05/2021 8:56 pm, Robbin Ehn wrote:

On Wed, 12 May 2021 08:27:33 GMT, Richard Reingruber wrote:

Hi Robbin,
I haven't found the time for a proper review yet but I've experimented a little bit with lambdas. I could not make it work because g++ created references to ::new which isn't allowed.
Thanks, Richard.

Hi Richard,
I tested lamdba, which is just a fancy way to write a crazy typed functor, we need to capture the lamdba so we can run it in the destructor. AFAICT the way to do that is using std::function.
Regarding ThreadClosure, we could use it, maybe that is preferable?!

Isn't a ThreadClosure for applying an operation to a set of threads?
That is not what we are doing here.

No it isn't. A closure is just a set of variable bindings and a function that can be executed. And yes we're doing just that.

A ThreadClosure is just an instance of this general concept. E.g. an AsyncHandshakeClosure (subclass of ThreadClosure) is a function (plus variable bindings) that is passed by a requesting thread to a target thread to be executed by it.

Thanks,
Richard.

@mlbridge
Copy link

mlbridge bot commented May 17, 2021

Mailing list message from David Holmes on hotspot-runtime-dev:

On 17/05/2021 5:05 pm, Robbin Ehn wrote:

On Thu, 13 May 2021 05:22:51 GMT, David Holmes <dholmes at openjdk.org> wrote:

Robbin Ehn has updated the pull request incrementally with one additional commit since the last revision:

Fixes for Dan

src/hotspot/share/prims/jvmtiRawMonitor.hpp line 48:

46: // The rules are:
47: // - We must never safepoint poll if raw monitor is owned.
48: // - We may safepoint poll before it is owned and after it has been released.

I'm not sure exactly what this is trying to say because user code can acquire a RawMonitor, then call into Java while holding the RawMonitor. That external use of RawMonitors should never cause any deadlock with the VMThread of course.

This comment applies to the RawMonitor code, where the typical use-case that otherwise can deadlock is:
JavaThread:
-lock RM
LOOP {
-wait RM
-do stufff with data from VM thread
}
-unlock RM

The user do not call into the VM/Java.

VM Thread:
-safepoint
-lock RM
-notify RM
-unlock RM

If we in this case safepoint between the lock and the unlock in wait() we deadlock with VM thread.

If the user would call into the VM/Java while holding the RM he obviously could deadlock with VM thread.

Only if the VMThread executes code that uses the same RM - which should
be a rare occurrence.

David
-----

Copy link
Member

@reinrich reinrich left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just one more, rather unimportant comment...

Either way: LGTM!

Thanks, Richard.

src/hotspot/share/prims/jvmtiRawMonitor.cpp Outdated Show resolved Hide resolved
Copy link
Member

@dholmes-ora dholmes-ora left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updates seem fine. Lucky _waiters was unused :)

Will take another full pass over the changes.

Thanks,
David

src/hotspot/share/runtime/objectMonitor.hpp Outdated Show resolved Hide resolved
Copy link
Member

@dholmes-ora dholmes-ora left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi Robbin,

Overall this looks good to me, but there is one issue that needs fixing (partially pre-existing but now also affecting ObjectMonitor::enter).

Other minor comments below.

Thanks,
David

@robehn
Copy link
Contributor Author

robehn commented May 20, 2021

Just one more, rather unimportant comment...

Either way: LGTM!

Thanks, Richard.

Thanks!

@mlbridge
Copy link

mlbridge bot commented May 20, 2021

Mailing list message from David Holmes on hotspot-runtime-dev:

On 20/05/2021 5:10 pm, Robbin Ehn wrote:

On Wed, 19 May 2021 14:25:41 GMT, Richard Reingruber <rrich at openjdk.org> wrote:

Just one more, rather unimportant comment...

Either way: LGTM!

Thanks, Richard.

Thanks!

src/hotspot/share/prims/jvmtiRawMonitor.cpp line 382:

380:
381: _recursions = 0;
382: ret = simple_wait(self, millis);

IMHO the guarantee at L379 is redundant with the newly added identical guarantee in `JvmtiRawMonitor::simple_wait()` at L240.

In case you agree to remove the guarantee, I don't see why the following statements cannot be pulled out of the if-statement.

 \_recursions \= 0\;
 ret \= simple\_wait\(self\, millis\)\;
 \_recursions \= save\;

Yes, it can be re-written as something below, but is this more readable?

I'd say yes with two minor modifications:

_recursions = 0;
ret = simple_wait(self, millis);

+ // Now we need to re-enter the monitor. For JavaThread's
+ // we need to manage suspend requests.

if (self->is_Java_thread()) { // JavaThread re-enter
JavaThread* jt = self->as_Java_thread();
{

I think this extra block scope can also go.

Cheers,
David
-----

@robehn
Copy link
Contributor Author

robehn commented May 20, 2021

Hi Robbin,

Sorry for the delay in getting through this.

Overall approach looks good. I have a few queries below and some requested naming changes to make things clearer.

Thanks,
David

Thanks!

@robehn
Copy link
Contributor Author

robehn commented May 20, 2021

Mailing list message from David Holmes on hotspot-runtime-dev:

On 20/05/2021 5:10 pm, Robbin Ehn wrote:

On Wed, 19 May 2021 14:25:41 GMT, Richard Reingruber wrote:

Just one more, rather unimportant comment...
Either way: LGTM!
Thanks, Richard.

Thanks!

src/hotspot/share/prims/jvmtiRawMonitor.cpp line 382:

380:
381: _recursions = 0;
382: ret = simple_wait(self, millis);

IMHO the guarantee at L379 is redundant with the newly added identical guarantee in JvmtiRawMonitor::simple_wait() at L240.
In case you agree to remove the guarantee, I don't see why the following statements cannot be pulled out of the if-statement.

 \_recursions \= 0\;
 ret \= simple\_wait\(self\, millis\)\;
 \_recursions \= save\;

Yes, it can be re-written as something below, but is this more readable?

I'd say yes with two minor modifications:

_recursions = 0;
ret = simple_wait(self, millis);

  • // Now we need to re-enter the monitor. For JavaThread's
  • // we need to manage suspend requests.

if (self->is_Java_thread()) { // JavaThread re-enter
JavaThread* jt = self->as_Java_thread();
{

I think this extra block scope can also go.

Cheers,
David

Fixed, thanks, Robbin

Copy link
Member

@dholmes-ora dholmes-ora left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updates mostly good but still a storestore issue to resolve.

Thanks,
David

src/hotspot/share/prims/jvmtiRawMonitor.cpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/interfaceSupport.inline.hpp Outdated Show resolved Hide resolved
@robehn
Copy link
Contributor Author

robehn commented May 24, 2021

Are we good to go?

Thanks, Robbin

Copy link
Member

@dcubed-ojdk dcubed-ojdk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thumbs up.

I just have some minor questions about a couple of the details.
Nothing blocking.

src/hotspot/share/prims/jvmtiRawMonitor.cpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/interfaceSupport.inline.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/interfaceSupport.inline.hpp Outdated Show resolved Hide resolved
src/hotspot/share/runtime/objectMonitor.cpp Show resolved Hide resolved
Comment on lines 449 to 451
// We cleared the pending monitor info since we've just gotten past
// the enter-check-for-suspend dance and we now own the monitor free
// and clear, i.e., it is no longer pending.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment explains why the set-pending-to-NULL code was here
(outside the loop). You moved the code, but now the comment is
misplaced and the code no longer works the same.

Update: I found this resolved comment below:

And I set current->set_current_pending_monitor(om); to OM again
in ExitOnSuspend if we exit the OM.

so that explains why we don't have a big change in behavior.

However, I think you still need to do something about this comment.

There is still an observable change in behavior in that the current
pending monitor can cycle between set->NULL->set->NULL when
we run into a suspend request. Previously, it did not cycle.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If we stop for a suspend request current pending monitor will always be set, so the cycling cannot be seen.
Updated comment(s).

@robehn
Copy link
Contributor Author

robehn commented May 25, 2021

Thumbs up.

I just have some minor questions about a couple of the details.
Nothing blocking.

Thanks Dan!

@reinrich
Copy link
Member

Still looks good to me.

Cheer's, Richard.

@robehn
Copy link
Contributor Author

robehn commented May 25, 2021

Still looks good to me.

Cheer's, Richard.

Thank you!

Copy link
Member

@dholmes-ora dholmes-ora left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nothing further from me.

Thanks,
David

@robehn
Copy link
Contributor Author

robehn commented May 26, 2021

Nothing further from me.

Thanks,
David

Thank you David!

Copy link
Member

@dcubed-ojdk dcubed-ojdk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thumbs up.

Reviewed the incremental between v05 -> v06. Looks good.

@robehn
Copy link
Contributor Author

robehn commented May 28, 2021

Thumbs up.

Reviewed the incremental between v05 -> v06. Looks good.

Thanks!

@robehn
Copy link
Contributor Author

robehn commented May 28, 2021

/integrate

@openjdk openjdk bot closed this May 28, 2021
@openjdk openjdk bot added integrated Pull request has been integrated and removed ready Pull request is ready to be integrated labels May 28, 2021
@openjdk
Copy link

openjdk bot commented May 28, 2021

@robehn Since your change was applied there have been 54 commits pushed to the master branch:

  • 6eb9114: 8266877: Missing local debug information when debugging JEP-330
  • 0c9daa7: 8265029: Preserve SIZED characteristics on slice operations (skip, limit)
  • 95b1fa7: 8267529: StringJoiner can create a String that breaks String::equals
  • 7f52c50: 8182043: Access to Windows Large Icons
  • 8a31c07: 8267886: ProblemList javax/management/remote/mandatory/connection/RMIConnector_NPETest.java
  • ae258f1: 8265418: Clean-up redundant null-checks of Class.getPackageName()
  • 41185d3: 8229517: Support for optional asynchronous/buffered logging
  • 7c85f35: 8267123: Remove RMI Activation
  • 0754266: 8267709: Investigate differences between HtmlStyle and stylesheet.css
  • 23189a1: 8191786: Thread-SMR hash table size should be dynamic
  • ... and 44 more: https://git.openjdk.java.net/jdk/compare/31d0f0d895ef4039d2e96a8fb6e990e93eed4d41...master

Your commit was automatically rebased without conflicts.

Pushed as commit 97ec5ad.

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

@openjdk openjdk bot removed the rfr Pull request is ready for review label May 28, 2021
@robehn robehn deleted the 8265753 branch May 28, 2021 07:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
hotspot-runtime hotspot-runtime-dev@openjdk.org integrated Pull request has been integrated serviceability serviceability-dev@openjdk.org
5 participants