Currently if a user enqueue a work item using schedule_delayed_work() the used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to schedule_work() that is using system_wq and queue_work(), that makes use again of WORK_CPU_UNBOUND. This lack of consistentcy cannot be addressed without refactoring the API.
alloc_workqueue() treats all queues as per-CPU by default, while unbound workqueues must opt-in via WQ_UNBOUND. This default is suboptimal: most workloads benefit from unbound queues, allowing the scheduler to place worker threads where they’re needed and reducing noise when CPUs are isolated. This change adds a new WQ_PERCPU flag to explicitly request alloc_workqueue() to be per-cpu when WQ_UNBOUND has not been specified. With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND), any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND must now use WQ_PERCPU. Once migration is complete, WQ_UNBOUND can be removed and unbound will become the implicit default. Suggested-by: Tejun Heo <[email protected]> Signed-off-by: Marco Crivellari <[email protected]> --- drivers/gpu/drm/i915/display/intel_display_driver.c | 4 ++-- drivers/gpu/drm/i915/i915_driver.c | 3 ++- drivers/gpu/drm/i915/selftests/i915_sw_fence.c | 2 +- drivers/gpu/drm/i915/selftests/mock_gem_device.c | 2 +- 4 files changed, 6 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/display/intel_display_driver.c b/drivers/gpu/drm/i915/display/intel_display_driver.c index cf1c14412abe..e12f9126b155 100644 --- a/drivers/gpu/drm/i915/display/intel_display_driver.c +++ b/drivers/gpu/drm/i915/display/intel_display_driver.c @@ -257,13 +257,13 @@ int intel_display_driver_probe_noirq(struct intel_display *display) goto cleanup_wq_modeset; } - display->wq.cleanup = alloc_workqueue("i915_cleanup", WQ_HIGHPRI, 0); + display->wq.cleanup = alloc_workqueue("i915_cleanup", WQ_HIGHPRI | WQ_PERCPU, 0); if (!display->wq.cleanup) { ret = -ENOMEM; goto cleanup_wq_flip; } - display->wq.unordered = alloc_workqueue("display_unordered", 0, 0); + display->wq.unordered = alloc_workqueue("display_unordered", WQ_PERCPU, 0); if (!display->wq.unordered) { ret = -ENOMEM; goto cleanup_wq_cleanup; diff --git a/drivers/gpu/drm/i915/i915_driver.c b/drivers/gpu/drm/i915/i915_driver.c index 0f33cdc11736..380cb20a47c6 100644 --- a/drivers/gpu/drm/i915/i915_driver.c +++ b/drivers/gpu/drm/i915/i915_driver.c @@ -143,7 +143,8 @@ static int i915_workqueues_init(struct drm_i915_private *dev_priv) * to be scheduled on the system_percpu_wq before moving to a driver * instance due deprecation of flush_scheduled_work(). */ - dev_priv->unordered_wq = alloc_workqueue("i915-unordered", 0, 0); + dev_priv->unordered_wq = alloc_workqueue("i915-unordered", WQ_PERCPU, + 0); if (dev_priv->unordered_wq == NULL) goto out_free_wq; diff --git a/drivers/gpu/drm/i915/selftests/i915_sw_fence.c b/drivers/gpu/drm/i915/selftests/i915_sw_fence.c index 8f5ce71fa453..b81d65c77458 100644 --- a/drivers/gpu/drm/i915/selftests/i915_sw_fence.c +++ b/drivers/gpu/drm/i915/selftests/i915_sw_fence.c @@ -526,7 +526,7 @@ static int test_ipc(void *arg) struct workqueue_struct *wq; int ret = 0; - wq = alloc_workqueue("i1915-selftest", 0, 0); + wq = alloc_workqueue("i1915-selftest", WQ_PERCPU, 0); if (wq == NULL) return -ENOMEM; diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index fb8751bd5df0..684e6ca0f960 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -221,7 +221,7 @@ struct drm_i915_private *mock_gem_device(void) if (!i915->wq) goto err_drv; - i915->unordered_wq = alloc_workqueue("mock-unordered", 0, 0); + i915->unordered_wq = alloc_workqueue("mock-unordered", WQ_PERCPU, 0); if (!i915->unordered_wq) goto err_wq; -- 2.51.0
