8234262: Unmask SIGQUIT in a child process

[RogerRiggs]

Clear the signal mask of the child when launching with posix_spawn.

SIGQUIT signals are handled on non-Java Threads by the VM.
For Java threads the signal mask blocks SIGQUIT.
The ProcessBuilder uses posix_spawn on all platforms to create new processes.
Without a specific request, the child process inherits the signal masks from the parent.

---

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-8234262: Unmask SIGQUIT in a child process

Reviewers

• Brian Burkhalter (@bplb - Reviewer) ⚠️ Review applies to e0e854bd
• David Holmes (@dholmes-ora - Reviewer) ⚠️ Review applies to dc56e0d8
• Vyom Tewari (@vyommani - Committer) ⚠️ Review applies to dc56e0d8
• Thomas Stuefe (@tstuefe - Reviewer) ⚠️ Review applies to dc56e0d8
• Alan Bateman (@AlanBateman - Reviewer) ⚠️ Review applies to dc56e0d8

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk pull/10379/head:pull/10379
$ git checkout pull/10379

Update a local copy of the PR:
$ git checkout pull/10379
$ git pull https://git.openjdk.org/jdk pull/10379/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 10379

View PR using the GUI difftool:
$ git pr show -t 10379

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/10379.diff

[bridgekeeper]

👋 Welcome back rriggs! 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]

@RogerRiggs The following label will be automatically applied to this pull request:

• core-libs

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]

Webrevs

• 03: Full - Incremental (727b0b2a)
• 02: Full - Incremental (dc56e0d8)
• 01: Full - Incremental (e0e854bd)
• 00: Full (c3b8d950)

[bplb]

Looks correct.

[openjdk]

@RogerRiggs 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:

8234262: Unmask SIGQUIT in a child process

Reviewed-by: bpb, dholmes, vtewari, stuefe, alanb

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 41 new commits pushed to the master branch:

• 968af74: 8293567: AbstractSplittableWithBrineGenerator: salt has digits that duplicate the marker
• 36b61c5: 8293872: Make runtime/Thread/ThreadCountLimit.java more robust
• 2be3158: 4797982: Setting negative size of JSplitPane divider leads to unexpected results.
• 050eebf: 8294245: Make Compile::print_inlining_stream stack allocated
• 91a23d7: 8294142: make test should report only on executed tests
• 169a5d4: 8294193: Files.createDirectories throws FileAlreadyExistsException for a symbolic link whose target is an existing directory
• 3675f4c: 8293252: Shenandoah: ThreadMXBean synchronizer tests crash with aggressive heuristics
• 543851d: 8289607: Change hotspot/jtreg tests to not use Thread.suspend/resume
• e2f8251: 8293618: x86: Wrong code generation in class Assembler
• 6ecd081: 8294270: make test passes awkward -status:-status:error,fail to jtreg
• ... and 31 more: https://git.openjdk.org/jdk/compare/9f90eb0524aba7bc40c574a067ce37ccfd8753b1...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.

[bplb]

Update looks good.

[dholmes-ora]

Hi Roger,
Using the spawn attributes seems more far reaching than simply temporarily changing the signal mask of the calling thread. I'd be concerned this has some unintended side-effects.

[RogerRiggs]

Hi Roger, Using the spawn attributes seems more far reaching than simply temporarily changing the signal mask of the calling thread. I'd be concerned this has some unintended side-effects.

@dholmes-ora The signal masks for threads are currently initialized and managed by the VM. I was concerned about application actions and Java code changing and then restoring the mask in the face of possible async behaviors. It seemed cleaner for the changes to affect only the child.

Posix_spawn is used on all platforms and is very specific about its actions.
I considered just resetting the mask for SIGQUIT, copying the rest of the signal mask from the parent, but then it might come up that some other signal was masked and should not be.

The launching of a new executable is a two+ step process. A small executable (jspawnhelper) is spawned and it does the exec of the requested executable.
An alternate implementation is to have jspawnhelper reset its own signal mask before exec'ing the target application.

[AlanBateman]

The launching of a new executable is a two+ step process. A small executable (jspawnhelper) is spawned and it does the exec of the requested executable. An alternate implementation is to have jspawnhelper reset its own signal mask before exec'ing the target application.

Doing this in jspawnhelper would be good to try.

[tstuefe]

Hi Roger, Using the spawn attributes seems more far reaching than simply temporarily changing the signal mask of the calling thread. I'd be concerned this has some unintended side-effects.

I don't think that is a good idea. If the thread receives a SIGQUIT in the time it takes to fire up posix_spawn, will it be prepared to handle the signal? Using a pthread spawn attribute is much less intrusive, since it only affects the child process.

The launching of a new executable is a two+ step process. A small executable (jspawnhelper) is spawned and it does the exec of the requested executable. An alternate implementation is to have jspawnhelper reset its own signal mask before exec'ing the target application.

Doing this in jspawnhelper would be good to try.

I think using posix_spawn attributes is fine. It does what you would do manually in jspawnhelper: calling sigprocmask in the child process: https://github.com/lattera/glibc/blob/master/sysdeps/posix/spawni.c#L218 .

[dholmes-ora]

Not sure it makes any difference whether this happens in the initial spawn or in jspawnhelper. I agree it is nicer to not affect the calling thread/process at all. My concern is that unblocking all signals may have some unintended side-effect, versus just dealing with SIGQUIT i.e. take a minimalist approach.

[RogerRiggs]

Putting the fix in jspawnhelper will address the problem in the other launch modes fork and vfork, so it not just posix_spawn fix.
SigQuit is only a problem because its one that the VM sets up to manage within the Java process.
Normal launching by the shell doesn't set any signals and I would not think a launched program would expect any masking.

[dholmes-ora]

Okay. Thanks for fixing this.

[vyommani]

Looks OK to me.

[tstuefe]

+1

[RogerRiggs]

/integrate

[openjdk]

Going to push as commit 5ae6bc2.
Since your change was applied there have been 41 commits pushed to the master branch:

• 968af74: 8293567: AbstractSplittableWithBrineGenerator: salt has digits that duplicate the marker
• 36b61c5: 8293872: Make runtime/Thread/ThreadCountLimit.java more robust
• 2be3158: 4797982: Setting negative size of JSplitPane divider leads to unexpected results.
• 050eebf: 8294245: Make Compile::print_inlining_stream stack allocated
• 91a23d7: 8294142: make test should report only on executed tests
• 169a5d4: 8294193: Files.createDirectories throws FileAlreadyExistsException for a symbolic link whose target is an existing directory
• 3675f4c: 8293252: Shenandoah: ThreadMXBean synchronizer tests crash with aggressive heuristics
• 543851d: 8289607: Change hotspot/jtreg tests to not use Thread.suspend/resume
• e2f8251: 8293618: x86: Wrong code generation in class Assembler
• 6ecd081: 8294270: make test passes awkward -status:-status:error,fail to jtreg
• ... and 31 more: https://git.openjdk.org/jdk/compare/9f90eb0524aba7bc40c574a067ce37ccfd8753b1...master

Your commit was automatically rebased without conflicts.

[openjdk]

@RogerRiggs Pushed as commit 5ae6bc2.

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