SchSeba / sriov-network-operator-1

63%

master: 64%

LAST BUILD ON BRANCH support_ib_netlink

branch: support_ib_netlink

CHANGE BRANCH

x

Reset

Sync Branches

• support_ib_netlink
• add_device_injector_validation
• add_hugepages_func_test
• add_logs
• add_mtu_func_test
• add_predicate_for_nodes
• add_rbac_to_nri
• add_rdma_documentation
• add_resync
• add_retry_on_delete
• allow_to_remove_default_pool
• block_resource_injector
• bug_remove_nic_config_from_host
• bump-dependencies-pr1027
• bump_cni_version
• bump_controller_runtime_0.20.2
• bump_ginkgo_version
• bump_k8s_1.30
• bump_k8s_version_ci
• bump_kube-rbac-proxy_v0.21.2
• bump_mock_package
• bump_ocp_version_4.18
• bumps_1.24
• change_vlan_max
• check_MCP_after_pod_restart
• check_lock_down_on_mlx_plugin
• ci_improvements
• clean_ci_images
• conditions/github-action-update
• conditions/network-crds
• conditions/nodestate
• conditions/operator-config
• conditions/operator-config-v2
• conditions/policy-poolconfig
• conditions/redeploy-script-fix
• create_desing_proposal_template
• create_platform_and_orchestrator_packages
• daemon_redesign
• design_conditions
• design_conditions_implementation
• design_externally_manage
• device_plugin_redesign
• devtask-d6444829-add-corerabbit-yaml-configuration-for-th
• documentation-re-org
• documentation_update
• drain-controller-test-fix
• drain_controller
• drain_daemon
• drain_improve
• enable_delve
• enable_feature_gate
• enable_mtu_test
• extend_time
• externally_manage
• externally_manage_tests
• fast-leader-election
• fix-webhook-image-user
• fix/daemon-test-flake
• fix_aws_test
• fix_device_plugin_label
• fix_drain_handler
• fix_drain_issue
• fix_drain_issue_v2
• fix_empty_nodeselector_in_policy
• fix_githubaction_artifact
• fix_ip_link_command_tests
• fix_k8s_ci
• fix_k8s_ci_multus
• fix_k8s_ci_virtual_cluster
• fix_k8s_ci_virtual_cluster_1
• fix_lock_race
• fix_log_test
• fix_matchcondition
• fix_mlx_plugin
• fix_mlx_plugin_1
• fix_mtu_disconnected_test
• fix_mtu_test_1
• fix_ocp_ci
• fix_ocp_ci_login
• fix_policy_order
• fix_race_condition
• fix_rdma_test
• fix_render
• fix_return_error
• fix_root_device_overlap
• fix_sriov_test_for_sno
• fix_status_patch
• fix_test
• fix_unittest
• fix_virtual_cluster_script
• fix_waiting_for_dp_when_no_policy
• fix_webhook
• func_e2e_fixes
• improve_drain_check
• improve_hypershift_check
• improve_logs
• improve_nic_selection
• install_kmod_for_mstfwreset
• interface_redesign
• master
• move_systemd_to_host_package
• multiple_bumps
• ocpbugs-64886
• parallel_drain
• parallel_drain_design
• platform_and_orchestrator_packages_design
• rdma-functest
• rdma-subsytem-mode
• readd_altname_docs
• reduce_rbac_resource_injector
• refs/pull/3/merge
• remove_DP_from_nodestate_spec
• remove_client
• remove_drain_on_none_vfs
• remove_leader_election
• remove_leader_election_test
• remove_timeout
• run_make_bundle
• security_improvements
• skip_drain_on_reboot
• skip_drain_on_reboot_single_node
• skip_virtual_nic_udev
• support_BF2_connectx
• support_PSA
• support_aws
• support_k8s_cluster_version
• support_post_quantum
• support_vfio_ib_in_webhook
• support_virtual_cluster
• sync-2025-02-03
• systemd
• temp_fix_ocp_ci
• test-separate-daemon-drain-fixes
• unit_test_no_mock
• unit_tests
• unit_tests_1
• update_bundle_to_stable
• update_ci_system
• update_github_actions
• update_ocp_ci_version
• use_context_to_annotate
• virtual_webhook

Build # 21545622043

Build Type

push

github

Committed by SchSeba

Commit Message

Fix "message too long" error for InfiniBand devices with many VFs

When configuring InfiniBand devices with many VFs, LinkByName() can fail
with "message too long" because the kernel's netlink response exceeds the
~16KB message limit. This happens because LinkByName requests all VF info
(including GUIDs) via RTEXT_FILTER_VF, which generates very large messages
for IB devices.

This commit adds a fallback mechanism:
- Add LinkByNameForSetVf() method that reads minimal link info (name and
  index) directly from sysfs instead of using netlink
- Update configSriovVFDevices() to catch "message too long" errors and
  fall back to LinkByNameForSetVf()
- Add unit test covering the fallback scenario

The fallback only triggers when the specific "message too long" error
occurs, preserving normal behavior for Ethernet devices and IB devices
with fewer VFs.

Signed-off-by: Sebastian Sch <sebassch@gmail.com>

Coverage Stats

17 of 50 new or added lines in 3 files covered. (34.0%)

198 existing lines in 2 files now uncovered.

9206 of 14528 relevant lines covered (63.37%)

0.71 hits per line