Skip to content

Fix VMBus channel count validation for netvsc devices #4513

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
knelsonmeister wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix VMBus channel count validation for netvsc devices #4513

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
649001d
Update LsVmBus test logic to use logical CPU count for expected chann…
kevinnelson-affirmed Jun 2, 2026
592f113
Fix VMBus channel count validation for netvsc devices
kevinnelson-affirmed Jun 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
56 changes: 42 additions & 14 deletions lisa/microsoft/testsuites/core/lsvmbus.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,13 +118,19 @@ def verify_vmbus_devices_channels_bsd(self, node: Node) -> None:
=> Expected channel count = min(num of vCPUs, 8).
- New logic (after Linux commit 646f071d315b75e87583de290d333478d42ccde1): # noqa: E501
2.1.3 If vCPU count <= 16, expected channel count = num of vCPUs.
2.1.4 If vCPU count > 16, expected channel count =
min(64, max(16, physical core count / 2)).
2.1.4 If vCPU count > 16, two valid formulas exist:
- Pre-646f071d: min(64, max(16, core_count // 2))
- Post-646f071d: min(64, max(16, thread_count // 2))
where thread_count = num_present_cpus (logical CPUs)
and core_count = physical cores.
Reference:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=646f071d315b75e87583de290d333478d42ccde1
- Test logic:
The code will first validate against the legacy rule.
If actual channel count does not match, it will then apply the new rule.
If actual channel count does not match, it checks against both
the pre-646f071d formula (core_count // 2) and the post-646f071d
formula (thread_count // 2), accepting either as valid since
different kernel versions coexist in production.
2.2 Check expected channel count of each storvsc SCSI device is min (num of
vcpu/4, 64).
2.2.1 Calculate channel count of each storvsc SCSI device.
Expand All @@ -147,6 +153,7 @@ def verify_vmbus_devices_channels(self, node: Node, log: Logger) -> None:
# vmbus channels created and associated.
lscpu_tool = node.tools[Lscpu]
thread_count = lscpu_tool.get_thread_count()
core_count = lscpu_tool.get_core_count()
# Each netvsc device should have "the_number_of_vCPUs" channel(s)
# with a cap value of 8.
expected_network_channel_count = min(thread_count, 8)
Expand All @@ -164,23 +171,44 @@ def verify_vmbus_devices_channels(self, node: Node, log: Logger) -> None:
)
# Note: mismatch may occur due to kernel change (commit 646f071d315b).
# In that case, validate again using the new logic.
# Two valid formulas exist depending on kernel version:
# - Pre-646f071d (Azure 6.8 etc): min(64, max(16, core_count // 2))
# - Post-646f071d: min(64, max(16, thread_count // 2))
# where thread_count = num_present_cpus (logical CPUs).
if actual_channels != expected_network_channel_count:
if thread_count <= 16:
expected_network_channel_count = thread_count
log.info(
"Applying new logic: expected channels = core_count "
f"({thread_count})"
"Applying new logic: expected channels = "
f"thread_count ({thread_count})"
)
else:
core_count = lscpu_tool.get_core_count()
expected_network_channel_count = min(
64, max(16, core_count // 2)
)
log.info(
"Applying new logic: expected channels = min(64, "
"max(16, physical_core_count // 2)) "
f"= {expected_network_channel_count}"
)
new_expected = min(64, max(16, thread_count // 2))
old_expected = min(64, max(16, core_count // 2))
if actual_channels == old_expected:
expected_network_channel_count = old_expected

@LiliDeng LiliDeng Jun 3, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice compatibility fix. I think this would be easier to read if we computed the allowed channel counts directly instead of updating expected_network_channel_count through fallback branches.

For example:

allowed = {min(thread_count, 8)}
if thread_count <= 16:
    allowed.add(thread_count)
else:
    allowed.add(min(64, max(16, core_count // 2)))
    allowed.add(min(64, max(16, thread_count // 2)))

assert_that(actual_channels).is_in(allowed)

That makes the old/new kernel behavior explicit and keeps the core_count vs thread_count logic easier to follow.

log.info(
"Matches pre-646f071d formula: min(64, "
"max(16, core_count // 2)) "
f"= {old_expected}"
)
elif actual_channels == new_expected:
expected_network_channel_count = new_expected
log.info(
"Matches post-646f071d formula: min(64, "
"max(16, thread_count // 2)) "
f"= {new_expected}"
)
else:
# Neither formula matches; fail with
# the new expected value for diagnostics.
expected_network_channel_count = new_expected
log.info(
f"Actual channels ({actual_channels}) "
"matches neither pre-646f071d "
f"({old_expected}) nor post-646f071d "
f"({new_expected}) formula."
)
assert_that(vmbus_device.channel_vp_map).is_length(
expected_network_channel_count
)
Expand Down
Loading