fix(rocr): Dup dmabuf fd on VMemoryImportShareableHandle for deferred import

[dayatsin-amd]

Summary

Re-land the dmabuf fd duplication fix from #8304, which was reverted in #8436 to unblock TheRock R-S bump.

After #6471 deferred per-GPU dmabuf import to hsa_amd_vmem_set_access time, VMemoryImportShareableHandle must own a dup of the caller's fd so deferred import still works after the caller closes its fd (normal IPC ownership). This restores correct behavior for multi-rank VMM/IPC peer-access flows (e.g. RCCL CE collectives).

Also removes LogPrint from DmaBufDup on Linux to avoid rocdxg WSL link issues (same pattern as #8436 follow-up).

Adds a rocrtst regression test that imports a shareable handle, closes the caller fd before set_access, and verifies GPU access succeeds.

JIRA ID

JIRA ID : ROCM-27777

Test plan

• Run ImportedShareableHandleSetAccessAfterFdClose rocrtst subtest on Linux with VMM support
• Verify HIP/RCCL multi-rank VMM peer-access reproducer from ROCM-27777 (hip-allgather-batch-cumem1.cpp)
• Confirm WSL/rocdxg build succeeds without LogPrint link errors in DmaBufDup

Made with Cursor

[therock-pr-bot]

❌ PR Check — Action Required

Check	Status	Details
🌿 Branch Name	✅ Pass	—
📝 PR Title/Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	❌ Fail	Error: Source/code files changed without an accompanying unit test. Expected: add at least one test file named like test_<name>.py / test_<name>.cpp (or <name>_test.*). Current: code file(s) changed: projects/rocr-runtime/rocrtst/suites/functional/virtual_memory.cc, projects/rocr-runtime/rocrtst/suites/functional/virtual_memory.h, projects/rocr-runtime/rocrtst/suites/test_common/main.cc, projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp, projects/rocr-runtime/runtime/hsa-runtime/core/util/lnx/os_linux.cpp (+2 more); no test file found
🔎 pre-commit	⏳ Pending	⏳ Still running…
🚫 Draft PR	🔜 To Be Enabled	—
🚩 Feature Flag	🔜 To Be Enabled	—
📊 Code Coverage	🔜 To Be Enabled	—

⚠️ 1 policy check(s) failed. Please address the issues above before this PR can be Reviewed.

🚫 Please fix the failed policies

• ❌ Unit Test

The Not ready to Review label was added to this PR. Once all policies pass, the label is removed automatically.

📖 Need help? See the Policy FAQ for details on every check and how to fix failures.

[therock-pr-bot]

🚫 Please fix the failed policies before requesting reviews.

The following policy checks failed:

• ❌ Unit Test

The Not ready to Review label has been added to this PR.
Once all policies pass, the label will be removed automatically.

[shwetagkhatri]

LGTM.

[amd-justchen]

Errors:
https://github.com/ROCm/rocm-systems/actions/runs/29136101237/job/86515367494?pr=8440#step:10:1716

1644 - Unit_Buffered_Printf_Flags (Failed) level_2 printf

https://github.com/ROCm/rocm-systems/actions/runs/29136101237/job/86515367484?pr=8440#step:11:1800

761 - Unit_hipMemcpyBatchAsync_H2D_Pageable_Stream_SourceAccess (Failed) level_2 memory stream transfer
1525 - Unit_hipOccupancyMaxActiveBlocksPerMultiprocessor_Positive_RangeValidation (Failed) level_2 occupancy
1533 - Unit_hipOccupancyMaxPotBlkSizeVariableSMemWithFlags_Functor (Failed) level_2 occupancy
1645 - Unit_Buffered_Printf_Specifier (Failed) level_2 printf

https://github.com/ROCm/rocm-systems/actions/runs/29136101237/job/86515367476?pr=8440#step:11:1878

6 - Unit_hiprtc_devicemalloc (SEGFAULT) integration level_2 rtc
762 - Unit_hipMemcpyBatchAsync_D2H_Functional (Failed) level_2 memory
1526 - Unit_hipOccupancyMaxActiveBlocksPerMultiprocessor_Positive_TemplateInvocation (Failed) level_2 occupancy
1534 - Unit_hipOccupancyMaxPotBlkSizeVariableSMemWithFlags_Lambda (Failed) level_2 occupancy
1642 - Unit_Printf_specifier_Sanity_Positive (Failed) level_2 printf
1646 - Unit_Host_Printf (SEGFAULT) level_2 printf
2538 - Unit_deviceAllocation_SingKernels_MulThreads (SEGFAULT) deviceLib level_2 memory

https://github.com/ROCm/rocm-systems/actions/runs/29136101237/job/86515367468?pr=8440#step:11:1890

759 - Unit_hipMemcpyBatchAsync_H2D_Functional (Failed) level_2 memory
1507 - Unit_hipEventIpc_DestroyNonBlocking (Failed) event ipc level_2 lifecycle
1531 - Unit_hipOccupancyMaxPotBlkSizeVariableSMemWithFlags_chkRange (Failed) level_2 occupancy
1539 - Unit_hipOccupancyAvailableDynamicSMemPerBlock_Positive (Failed) level_2 occupancy
2539 - Unit_deviceAllocationFollowedByDeviceReset (SEGFAULT) deviceLib level_2 memory
2847 - Unit_AccessMallocAcrossKernels (SEGFAULT) level_2 sanityTests

[dayatsin-amd]

These errors are not caused by this change:

Other PRs with identical Windows ROCR failures

The same 4/4 shard ROCR failure pattern appears on unrelated open PRs — same platform (gfx1151), same backend (ROCR):

PR	Title
#8293	fix(rocr): Return hipErrorNotReady when fabric export not ready
#8445	Bump PAL_CLIENT_INTERFACE_MAJOR_VERSION 954 --> 995
#8439	fix(clr): remove unused x86 SSE headers
#8435	feat(rocr): PC sampling trap handler refactor and lock-free MPSC queue
#8424	Fix: [hip-tests] add --offload-device-only to copyKernel.s build target

Failure themes are printf, occupancy, memcpy batch, event IPC, RTC, and device allocation — none related to dmabuf fd duplication or deferred VMM import.

Conclusion: Windows gfx1151 ROCR hip-test failures look like a pre-existing platform baseline issue, not caused by this PR.