Assertion failure in commands.cpp when running with L0 on PVC

[igchor]

Describe the bug

When running syclcompat/memory/memory_management_test3.cpp with L0 v2 adapter (with copy offload disabled) on PVC, I got:

env env UR_LOADER_USE_LEVEL_ZERO_V2=1 ONEAPI_DEVICE_SELECTOR=level_zero:gpu  /__w/llvm/llvm/build-e2e/syclcompat/memory/Output/memory_management_test3.cpp.tmp.out
# executed command: env env UR_LOADER_USE_LEVEL_ZERO_V2=1 ONEAPI_DEVICE_SELECTOR=level_zero:gpu /__w/llvm/llvm/build-e2e/syclcompat/memory/Output/memory_management_test3.cpp.tmp.out
# .---command stdout------------
# | void test_free_memory()
# | void test_free_memory_q()
# | void test_wait_and_free_memory()
# | void test_wait_and_free_memory_q()
# | void test_memcpy_async()
# | void test_memcpy_async_q()
# | void test_memcpy_async_pitched()
# `-----------------------------
# .---command stderr------------
# | memory_management_test3.cpp.tmp.out: /__w/llvm/llvm/src/sycl/source/detail/scheduler/commands.cpp:942: virtual bool sycl::detail::Command::enqueue(EnqueueResultT &, BlockingT, std::vector<Command *> &): Assertion `!MMarkedForCleanup' failed.
# `-----------------------------

See https://github.com/intel/llvm/actions/runs/14388998049/job/40352151771?pr=17696

To reproduce

llvm-lit --param sycl_devices="level_zero_v2:gpu" --param extra_environment="UR_L0_V2_FORCE_DISABLE_COPY_OFFLOAD=1" ./sycl/test-e2e/syclcompat/memory/memory_management_test3.cpp

Environment

• CI docker env
• OS: ubuntu 24.04
• sycl-ls output:

[level_zero:gpu][level_zero:0] Intel(R) oneAPI Unified Runtime over Level-Zero, Intel(R) Data Center GPU Max 1100 12.60.7 [1.6.32961.700000]
[level_zero:gpu][level_zero:1] Intel(R) oneAPI Unified Runtime over Level-Zero, Intel(R) Data Center GPU Max 1100 12.60.7 [1.6.32961.700000]
[native_cpu:cpu][native_cpu:0] SYCL_NATIVE_CPU, SYCL Native CPU 0.1 [0.0.0]
[opencl:gpu][opencl:0] Intel(R) OpenCL Graphics, Intel(R) Data Center GPU Max 1100 OpenCL 3.0 NEO  [25.09.32961.7]
[opencl:gpu][opencl:1] Intel(R) OpenCL Graphics, Intel(R) Data Center GPU Max 1100 OpenCL 3.0 NEO  [25.09.32961.7]
[opencl:cpu][opencl:2] Intel(R) OpenCL, Intel(R) Xeon(R) Platinum 8

Additional context

FYI, @intel/llvm-reviewers-runtime

[uditagarwal97]

Another instance of this failure: https://github.com/intel/llvm/actions/runs/14406803376/job/40406771137?pr=17967#step:17:307

The failure is not only with V2 adapter, but with V1 as well.

[againull]

This is a regression after #17871. @cperkinsintel Could you please take a look.