RFR: 8367485: os::physical_memory is broken in 32-bit JVMs when running on 64-bit OSes [v2]

[Stefan Karlsson]

On Thu, 18 Sep 2025 12:51:39 GMT, Anton Artemov <duke at openjdk.org> wrote:

>> Hi, please consider the following changes:
>> 
>> In this PR we address the overflow issue in `os::physical_memory()` on Linux, which can occur when running a 32-bit JVM on a 64-bit machine, introduced by https://bugs.openjdk.org/browse/JDK-8357086. The problem is that the product of _SC_PHYS_PAGES and _SC_PAGESIZE can overflow according to the documentation. 
>> 
>> The issue is addressed by changing the output type of all related functions to `uint64_t`.
>> 
>> Tested in tiers 1 - 5.
>
> Anton Artemov has updated the pull request incrementally with one additional commit since the last revision:
> 
>   8367485: Refactoring with uint64_t

src/hotspot/os/linux/os_linux.cpp line 306:

> 304:     return false;
> 305:   }
> 306:   value = static_cast<uint64_t>(si.totalswap * si.mem_unit);

Could the `si.totalswap * si.mem_unit` calculation overflow on 32-bit JVMs? I wonder if it would be safer to write the code this way:

static_cast<uint64_t>(si.totalswap) * si.mem_unit;

The same comment goes for si.freeswap below.

src/hotspot/os/linux/os_linux.cpp line 2572:

> 2570:   struct sysinfo si;
> 2571:   sysinfo(&si);
> 2572:   uint64_t phys_mem = physical_memory();

You need to update the two `%zu` below.

src/hotspot/os/windows/os_windows.cpp line 3967:

> 3965: // Processor level is not available on non-NT systems, use vm_version instead
> 3966: int    os::win32::_processor_level           = 0;
> 3967: uint64_t os::win32::_physical_memory           = 0;

The indentation is misaligned now.

src/hotspot/os/windows/os_windows.hpp line 43:

> 41:   static int    _processor_type;
> 42:   static int    _processor_level;
> 43:   static uint64_t _physical_memory;

Indentation.

src/hotspot/share/gc/z/zLargePages.cpp line 34:

> 32:   pd_initialize();
> 33: 
> 34:   const uint64_t memory = os::physical_memory();

PROPERFMT below uses `%zu`. Given that ZGC is only compiled to 64-bits maybe you could make the memory variable size_t here. I don't know if you need an explicit cast or not.

src/hotspot/share/jfr/periodic/jfrPeriodic.cpp line 534:

> 532:   EventPhysicalMemory event;
> 533:   event.set_totalSize(totalPhysicalMemory);
> 534:   uint64_t avail_mem = 0;

I think you can remove all usages of u8 in this function.

src/hotspot/share/runtime/os.cpp line 1948:

> 1946:   /* Is this a server class machine? */
> 1947:   if ((os::active_processor_count() >= (int)server_processors) &&
> 1948:       (phys_mem >= static_cast<uint64_t>(server_memory - missing_memory))) {

You don't need the cast if you change the type of `server_memory` and `missing_memory` to uint64_t instead.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359179008
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359191538
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359195029
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359196770
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359215587
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359221762
PR Review Comment: https://git.openjdk.org/jdk/pull/27335#discussion_r2359258332