Repository navigation
test(amdsmi): cover loader resilience under a restrictive import guard - #9879
Merged
marifamd merged 3 commits intoAug 31, 2026
Merged
Conversation
dmitrii-galantsev
force-pushed
the
users/dgalants/amdsmi-packaging-scriptlet-scope
branch
from
August 10, 2026 16:50
f1ece37 to
997ec8a
Compare
dmitrii-galantsev
force-pushed
the
users/marifamd/amdsmi-loader-import-guard-test
branch
from
August 10, 2026 16:50
ee64282 to
3d72487
Compare
✅ All Policy Checks Passed
📖 Need help? See the Policy FAQ for details on every check and how to fix failures. |
dmitrii-galantsev
force-pushed
the
users/dgalants/amdsmi-packaging-scriptlet-scope
branch
from
August 11, 2026 22:01
997ec8a to
d055df1
Compare
Base automatically changed from
users/dgalants/amdsmi-packaging-scriptlet-scope
to
develop
August 12, 2026 21:21
github-actions
Bot
added
the
github actions
dmitrii-galantsev
force-pushed
the
users/marifamd/amdsmi-loader-import-guard-test
branch
from
August 28, 2026 16:26
092ad8a to
c2dcef8
Compare
dmitrii-galantsev
approved these changes
Aug 28, 2026
dmitrii-galantsev
left a comment
dmitrii-galantsev
left a comment
Contributor
There was a problem hiding this comment.
claude:
Rebased onto current develop and now a clean single-file change to test_abi_compat.py, with the already-merged #9695 content, the SPDX-header reverts, and the configure-breaking CMakeLists.txt changes all dropped. The mutation test was re-run after the rebase and still fails without the guarded branch, so the coverage is real — LGTM.
The wrapper loader must not abort import when a downstream consumer installs a restrictive import guard. rocprofiler-compute runs amd-smi under a profile-mode guard that raises a bare ImportError (the parent of ModuleNotFoundError) for any non-whitelisted top-level package. A prior loader change imported a third-party package in its search path and caught only (ModuleNotFoundError, FileNotFoundError), so the guard's ImportError escaped and broke import amdsmi in downstream CI. Add RestrictiveImportGuardTest, which reimports the wrapper under a minimal clone of that guard with the offending package left off the whitelist, and asserts the loader still resolves the override path. Any unguarded third-party import that returns to the load path now fails in amd-smi CI instead of downstream. Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
AMDSMI_LIB_OVERRIDE short-circuits _load_library() at step 1, so the guard test never entered the relocatable-ROCm-tree branch -- the one place a rocm_sdk import would plausibly live, and the exact regression this test was added for. A mutation injected there passed unnoticed. Add a case with no override so the loader walks steps 2-4, and assert on the resolved path so the test reports more than "the import did not raise". Drop the 3.10-only sys.stdlib_module_names dependency: it was incidental but skipped 4 of the 9 CI matrix legs, while pyproject declares >=3.6. Short-circuit the guard on already-imported modules; the guard window is process-wide, and failing a cached re-import is hostile to any in-process tracer or plugin rather than a property worth asserting.
marifamd
force-pushed
the
users/marifamd/amdsmi-loader-import-guard-test
branch
from
August 29, 2026 20:32
c2dcef8 to
88d7a94
Compare
gabrpham
approved these changes
Aug 31, 2026
Contributor
TheRock Submodule Bump ActivityNewest first
|
1 task done
dmitrii-galantsev
added a commit
that referenced
this pull request
Nov 6, 2026
…#14510) ## Motivation TheRock's `rocm-sdk-core` wheel carries the `amdsmi` Python module only at `_rocm_sdk_core/share/amd_smi` (since #9340). Copied into site-packages, `import amdsmi; amdsmi.amdsmi_init()` fails with `libamd_smi.so.27: cannot open shared object file` unless `torch` was imported first, because only PyTorch's preload makes the library findable. ## Technical Details `_load_library()` in `amdsmi_wrapper.py` (mirrored in `tools/generator.py`): - **`_rocm_sdk_core` lookup.** When no system library is found or the one found is refused, the loader takes `lib/libamd_smi.so.27` from the `_rocm_sdk_core` package (`sys.modules`, else `sys.path`). It checks paths only and imports nothing, so import guards like rocprofiler-compute's cannot break it (#9879). - **`_AMDSMI_ROCM_SDK_CORE` flag.** TheRock sets it in the copy it ships in `rocm-sdk-core` (ROCm/TheRock#9578). That copy then uses the relocatable path and `_rocm_sdk_core` only, never a system `libamd_smi.so` from another release. The rpm/deb copy keeps its order and only gains the fallback. - **Reuse before load.** Each candidate is first opened with `RTLD_NOLOAD`. PyTorch 2.14's `_amdsmi_cdll_hook` (in the ROCm 10.1.0 wheels; narrowed upstream in pytorch/pytorch#197482) redirects any `libamd_smi.so*` load to `$ROCM_PATH/lib/libamd_smi.so` first. With `ROCM_PATH` set and a system ROCm installed, the module loaded that second copy and both `torch.cuda.device_count()` and amdsmi reported 0 GPUs. Asked with `RTLD_NOLOAD`, the hook can only return a copy already in the process. - **Library version check.** SONAME 27 changed struct sizes between 27.1 and 27.2: `amdsmi_fabric_info_t` grew from 320 to 552 bytes when its union gained `v2`, and 10.2 nightlies built before #13926 have a 1792-byte `amdsmi_driver_info_t` that `amdsmi_get_gpu_driver_info()` zeroes in full, against the 768 bytes a 27.1 module allocates. A library newer than the module's `_version.py` is closed and skipped, so it cannot write past the module's buffers. An older one warns. `AMDSMI_LIB_OVERRIDE` still loads anything. - **Version stays in sync.** `get_version_from_file` makes CMake reconfigure when `amdsmi.h` changes, so `_version.py` cannot lag the library in an incremental build. - **Docs.** `docs/packaging.md`, `docs/install/install.md` and `py-interface/README.md` describe the new loader order, the flag, `RTLD_NOLOAD` reuse and the version check. This does not make `share/amd_smi` pip-installable again. A module installed from there would not depend on `rocm-sdk-core` and would go stale on the next ROCm upgrade. ## Issue Tracking JIRA ID: ROCM-33336 Related: ROCM-31613, ROCm/TheRock#9226 (with #14409), ROCm/TheRock#3054 ## Test Plan - Python loader tests (`tests/python/test_abi_compat.py`): flag on and off, the fallback, `RTLD_NOLOAD` reuse, newer and older libraries, closing a refused library, unusable `sys.path` entries. Each new check was disabled once to confirm its test fails. - The 10.1.0 `rocm-sdk-core` module patched with this loader, on the shipped wheels (`rocm-sdk-core` 10.1.0, `torch 2.14.0+rocm10.1.0`), 2× MI210, host with a 10.2 ROCm in `ld.so.cache` and `ROCM_PATH` set: - copied into site-packages and via `PYTHONPATH`; - imported alone, before torch and after torch; - also with `/opt` and `ld.so.cache` hidden; - also with the 10.2 library loaded first. - The `amd-smi` CLI from an install tree and from `rocm-sdk-core`'s entry point. - A version bump in `amdsmi.h` followed by `ninja`: reconfigures and updates `_version.py`. - The `DESTDIR` install: the site-packages copy matches `share/amd_smi`, `_version.py` included. ## Test Result - Python tests: 55 passed. - Module, every case: one `libamd_smi` mapped, torch and amdsmi each report 2 GPUs. - With a newer 10.2 library already loaded before PyTorch 2.14's hook imports the module, the module refuses it with the version error rather than binding it. - `amd-smi version` and `amd-smi list`: work from both locations. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/rocm-systems/blob/develop/CONTRIBUTING.md. --------- Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
The Python wrapper loader must survive a hostile import environment set up by a downstream consumer. rocprofiler-compute runs amd-smi under a profile-mode import guard that raises a bare
ImportError(the parent ofModuleNotFoundError) for any non-whitelisted top-level package. A past loader change imported a third-party package in its search path but caught only(ModuleNotFoundError, FileNotFoundError), so the guard'sImportErrorescaped, abortedimport amdsmi, and broke 21 tests in TheRock CI, forcing a revert. Nothing in amd-smi's own suite catches that failure class today.Technical Details
Stacked on #9695. Adds one test to
projects/amdsmi/tests/python/test_abi_compat.py:_ProfileModeLikeGuard: a minimal clone of the downstream guard. Allows stdlib plus a small ROCm whitelist and raisesImportErrorfor every other top-level package, from both ameta_pathfinder and abuiltins.__import__hook.rocm_sdkis deliberately left off the whitelist.RestrictiveImportGuardTest.test_loader_survives_restrictive_import_guard: reimports the wrapper under that guard, reusing theAMDSMI_LIB_OVERRIDEescape hatch and the existing_Patch/STABLE_SYMBOLSfake-CDLL harness, then asserts the loader still resolves the override path.sys.stdlib_module_names.If any unguarded third-party import creeps back into the load path, this fails here in amd-smi CI instead of downstream in TheRock.
Issue Tracking
JIRA ID: ROCM-3941
Test Plan
cd projects/amdsmi/tests/python && python3 -m unittest test_abi_compat -vpre-commit run --config projects/amdsmi/.pre-commit-config.yaml --files projects/amdsmi/tests/python/test_abi_compat.pyTest Result
Ran 11 tests in 0.166s/OK(10 existing + the newRestrictiveImportGuardTest).Submission Checklist