[AI Generated] BugFix: skip known upstream xdp-tools promiscuous test failures#4451
Conversation
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
This PR relaxes the xdp-tools test runner behavior by treating two known upstream promiscuous-mode failures as non-fatal, so XDP validation doesn’t fail on Ubuntu 24.04+ due to an upstream issue.
Changes:
- Add a logger and a allowlist of known upstream-failing tests.
- Split abnormal xdp-tools test results into known vs unexpected failures.
- Log known failures as warnings while still failing on unexpected ones.
| _KNOWN_UPSTREAM_FAILURES: Set[str] = { | ||
| "test_promiscuous_selfload", | ||
| "test_promiscuous_preload", | ||
| } |
There was a problem hiding this comment.
The PR description says these failures are specific to Ubuntu 24.04+ (kernel 6.8+), but the code unconditionally ignores them on all distros/kernels. This risks masking real regressions on other platforms. Consider gating the allowlist by OS/version and/or kernel version (e.g., only apply when running on Ubuntu >= 24.04 or kernel >= 6.8), and otherwise treat these as unexpected failures.
| if unexpected: | ||
| raise LisaException(f"found failed tests: {unexpected}") | ||
| if not known: | ||
| result.assert_exit_code( | ||
| 0, | ||
| "unknown error on xdp tests, please check log for more details.", | ||
| ) |
There was a problem hiding this comment.
When known is non-empty, the code skips checking result's exit code entirely. That can allow non-test-related failures (e.g., runner crash, bad invocation, missing deps) to pass as long as the parsed test list only contains known failures. A safer approach is to still validate the exit code in the known case—e.g., allow only the specific failure exit code(s) that correspond to test failures, and fail on any other non-zero exit code.
| return True | ||
|
|
||
|
|
||
| _log = logging.getLogger(__name__) |
There was a problem hiding this comment.
no need this, use node.log
| # These are not LISA or driver regressions; they are upstream test issues. | ||
| # test_promiscuous_selfload / test_promiscuous_preload: fail on Ubuntu 24.04+ | ||
| # (kernel 6.8+) because the upstream test's promiscuous-mode detection logic | ||
| # is incompatible with newer kernel behaviour. |
There was a problem hiding this comment.
provide the doc link here
| } | ||
|
|
||
| if known: | ||
| _log.warning("ignoring known upstream test failures: %s", known) |
There was a problem hiding this comment.
don't use %s use f-format
Gnandeep99
force-pushed
the
bugfix/xdp-skip-promiscuous-upstream-failures_290426_155629
branch
from
c4e01a0 to
aa79ac6
Compare
…eam failures - Gate allowlist by Ubuntu + kernel >= 6.8 (don't skip on other distros) - Always validate exit code ([0, 2]); reject crashes/OOM/missing deps - Use self.node.log.warning with f-string instead of module-level _log - Add upstream issue link placeholder and removal guidance in comment - Remove unused 'import logging'
| result.assert_exit_code( | ||
| 0, "unknown error on xdp tests, please check log for more details." | ||
| [0, _MAKE_TEST_FAILED_EXIT_CODE], | ||
| "unknown error on xdp tests, please check log for more details.", | ||
| ) | ||
|
|
| try: | ||
| kernel_version = self.node.os.get_kernel_information().version | ||
| except Exception as ex: # noqa: BLE001 - defensive: kernel info missing | ||
| self.node.log.debug( | ||
| f"could not detect kernel version, " | ||
| f"disabling xdp-tools allowlist: {ex}" | ||
| ) | ||
| return False |
Summary
Skip known upstream xdp-tools promiscuous mode test failures (test_promiscuous_selfload, test_promiscuous_preload) that consistently fail on Ubuntu 24.04+ due to an upstream xdp-tools issue unrelated to Azure/LISA. These failures are now logged as warnings instead of causing the entire test to fail.
Validation Results