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Outdated suggestions cannot be applied.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.
test_queue_supv_to_user() invokes a child thread which does some
work which must take place before the call to k_queue_cancel_wait()
is called by the parent.
However, with SMP enabled, the child thread will just run on another
CPU and we have a race between when child_thread_get() calls
k_queue_get(q, K_FOREVER) and the parent calls k_queue_cancel_wait().
If the parent thread gets there first, the whole test hangs as
the call to k_queue_get(q, K_FOREVER) sits forever.
The fix is to have test_queue_supv_to_user() be a 1cpu test, which
ensures that only one CPU is used.
It's not clear to me why this wasn't causing CI failures on other
SMP targets, but I am able to reproduce reliably on qemu_x86_64
with my user mode patches applied.
Signed-off-by: Andrew Boie andrew.p.boie@intel.com