RFR: 8319589: Attach from root to a user java process not supported in Mac [v8]

[Sergey Chernyshev]

On Thu, 4 Dec 2025 00:31:37 GMT, Serguei Spitsyn <sspitsyn at openjdk.org> wrote:

>> Sergey Chernyshev has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   addressed review comments
>
> We had a private discussion with Larry on this PR and undesired dependency on the `sun.jvmstat` module.
> We got the following concerns:
>  - Do we really need 3x implementations of the MacOS `tempdir` finder function?
>  - Do we really need a modular dependency from `jdk.attach` to `jdk.internal.jvmstat`?
> New method `getTemporaryDirectories(int pid)` is placed into the `jdk.internal.jvmstat/macosx/classes/sun/jvmstat/PlatformSupportImpl.java`. It seems it was needed to inherit the `PlatformSupport` class which has the method `PlatformSupport.getTemporaryDirectory()`. But it returns the same as the removed function `Java_sun_tools_attach_VirtualMachineImpl_getTempDir()`. So, this implementation can be moved to the `jdk.attach` module.
> My understanding is that class `PlatformSupportImpl` and its method `getTemporaryDirectories(int pid)`  is not currently used for `jvmstat` implementation. Is it correct? Is it right that the `jvmstat` implementation still does not properly support root user on macOS?
> Anyway, some unification and refactoring is needed here. We could file a follow up enhancement to sort this out but also wanted to know your opinion on this.

Hi @sspitsyn ,

Thank you for looking at this patch!

> This is a nice fix in general. Thank you for this work! I hope to complete my review tomorrow with one more pass. Question: How was this update tested? Do you have a jtreg test or you've done it manually? Also, did you run mach5 tiers 1-6 to be safe and protected against regressions?

The patch was tested against the current jtreg test groups (tier1-3) on macOS both amd64 and arm64. I also run jcmd and jps tools manually under root, because the change only affects the behavior under root.

> * It seems it was needed to inherit the `PlatformSupport` class which has the method `PlatformSupport.getTemporaryDirectory()`. But it returns the same as the removed function `Java_sun_tools_attach_VirtualMachineImpl_getTempDir()`. So, this implementation can be moved to the `jdk.attach` module.

The removed function `Java_sun_tools_attach_VirtualMachineImpl_getTempDir()` would return always the same string cached inside the libsystem_coreservices.dylib, that would correspond the user specific temporary path as of how the JVM process has started, that would have nothing to do with the target process/pid temporary directory.

> * My understanding is that class `PlatformSupportImpl` and its method `getTemporaryDirectories(int pid)`  is not currently used for `jvmstat` implementation. Is it correct?

I think this is not exactly so. `getTemporaryDirectories(int pid)` is used currently in the LocalVmManager to get the list of active VMs, that is consumed by `jcmd`, `jps` and `jconsole` utilities. When 0 is passed as a parameter, it returns the list of temp directories for all processes, if they were different (for example if there are containerized JVM processes under Linux).

> * Is it right that the `jvmstat` implementation still does not properly support root user on macOS?

Yes and this is what this patch was intended for.

>     * Do we really need 3x implementations of the MacOS `tempdir` finder function?

It was already 2x implementation of HotSpot perfdata files lookup, HotSpot itself uses the `os::get_temp_directory()`, on the java side the LocalVmManager in `sun.jvmstat.perfdata.monitor` relies on `getTemporaryDirectories(int pid)` in the PlatformSupport. The 3rd was the macOS native `Java_sun_tools_attach_VirtualMachineImpl_getTempDir()`, now it's proposed for removal as it returns only the current user temp path, no matter what process is being attached to.

>     * Do we really need a modular dependency from `jdk.attach` to `jdk.internal.jvmstat`?

It was already a dependency of `jdk.attach` to `sun.jvmstat.monitor`. This change exports also `sun.jvmstat`, in addition to `sun.jvmstat.monitor` as it was before.

> src/hotspot/os/bsd/os_bsd.cpp line 956:
> 
>> 954: }
>> 955: #endif
>> 956: 
> 
> Nit: I'd suggest to get rid of empty lines which do not improve readability:
>  890, 903, 908, 917, 923, 930, 933, 941, 944
>  Also, the comment at line 931 is better to start with a capital letter: `// If the ..` .

Thanks @sspitsyn , done.

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

PR Comment: https://git.openjdk.org/jdk/pull/25824#issuecomment-3616052858
PR Review Comment: https://git.openjdk.org/jdk/pull/25824#discussion_r2592160722