From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
Previously, the rescuer scanned for all matching work items at once and
processed them within a single rescuer thread, which could cause one
blocking work item to stall all others.
Make the rescuer process work items one-by-one instead of slurping all
matches in a single pass.
Break the rescuer loop after finding and processing the first matching
work item, then restart the search to pick up the next. This gives
normal worker threads a chance to process other items which gives them
the opportinity to be processed instead of waiting on the rescuer's
queue and prevents a blocking work item from stalling the rest once
memory pressure is relieved.
Introduce a dummy cursor work item to avoid potentially O(N^2)
rescans of the work list. The marker records the resume position for
the next scan, eliminating redundant traversals.
Change from v2:
Renames.
Change WARN_ON_ONCE() conditions in assign_work().
Remove the check whether the cursor is on the list in send_mayday()
which has the benefit of enabling the ability to only process a
limit number of work items per run and still allow the cursor in
worklist between runs.
Also remove the insertion of the cursor in send_mayday() and contain
the code of handling the cursor in assign[_rescuer]_work().
Split patch.
V2: https://lore.kernel.org/lkml/20251118093832.81920-1-jiangshanlai@gmail.com/
Changed from v1:
Insert the marker and use it purely as iteration marker as Tejun
request.
It is hard to maintain the proper state of it, expecially maintaining
it only in pool->worklist which also requires an unlikely path in the
worker_thread() path.
Add an unlikely path in the worker_thread() path(in assign_work()).
Add other code in other place to make it not be treated like a work
item. I'm sure all the paths have been covered now, but I still feel
a bit nevous about it for future changes.
Extra the code to assign work in the rescuer_thread() out as
assign_rescue_work().
V1: https://lore.kernel.org/lkml/20251113163426.2950-1-jiangshanlai@gmail.com/
Lai Jiangshan (7):
workqueue: Factor out assign_rescuer_work()
workqueue: Only assign rescuer work when really needed
workqueue: Don't rely on wq->rescuer to stop rescuer
workqueue: Loop over in rescuer until all its work is done
workqueue: Process rescuer work items one-by-one using a cursor
workqueue: Limit number of processed works in rescuer per turn
workqueue: Process extra works in rescuer when there are no more to
rescue
kernel/workqueue.c | 146 ++++++++++++++++++++++++++++++++-------------
1 file changed, 103 insertions(+), 43 deletions(-)
--
2.19.1.6.gb485710b