| Message ID | 20160716075920.GA4547@Karyakshetra |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On Sat, Jul 16, 2016 at 01:29:20PM +0530, Bhaktipriya Shridhar wrote: > The workqueue health->wq was used as per device private health thread. > This was done so that system error handling could be processed > concurrently. Not exactly, AFAIK it was intended to perform delayed work and not relevant to concurrency. > The workqueue has a single workitem(&health->work) and > hence doesn't require ordering. It is involved in handling the health of > the deviceand is not being used on a memory reclaim path. > Hence, the singlethreaded workqueue has been replaced with the use of > system_wq. Yes > > System workqueues have been able to handle high level of concurrency > for a long time now and hence it's not required to have a singlethreaded > workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue > created with create_singlethread_workqueue(), system_wq allows multiple > work items to overlap executions even on the same CPU; however, a > per-cpu workqueue doesn't have any CPU locality or global ordering > guarantee unless the target CPU is explicitly specified and thus the > increase of local concurrency shouldn't make any difference. Not relevant. > > Work item has been flushed in mlx5_health_cleanup() to ensure that > there are no pending tasks while disconnecting the driver. > > Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com> > --- > drivers/net/ethernet/mellanox/mlx5/core/health.c | 7 ++----- > 1 file changed, 2 insertions(+), 5 deletions(-) > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/health.c b/drivers/net/ethernet/mellanox/mlx5/core/health.c > index 42d16b9..9acbccf 100644 > --- a/drivers/net/ethernet/mellanox/mlx5/core/health.c > +++ b/drivers/net/ethernet/mellanox/mlx5/core/health.c > @@ -267,7 +267,7 @@ static void poll_health(unsigned long data) > if (in_fatal(dev) && !health->sick) { > health->sick = true; > print_health_info(dev); > - queue_work(health->wq, &health->work); > + schedule_work(&health->work); > } > } > > @@ -296,7 +296,7 @@ void mlx5_health_cleanup(struct mlx5_core_dev *dev) > { > struct mlx5_core_health *health = &dev->priv.health; > > - destroy_workqueue(health->wq); > + flush_work(&health->work); > } > > int mlx5_health_init(struct mlx5_core_dev *dev) > @@ -311,10 +311,7 @@ int mlx5_health_init(struct mlx5_core_dev *dev) > > strcpy(name, "mlx5_health"); > strcat(name, dev_name(&dev->pdev->dev)); > - health->wq = create_singlethread_workqueue(name); > kfree(name); You need to remove "name" initialization/usage too. It is not needed. > - if (!health->wq) > - return -ENOMEM; > > INIT_WORK(&health->work, health_care); > > -- > 2.1.4 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
Sure. Will make the changes in v2. Thanks, Bhaktipriya On Sun, Jul 17, 2016 at 10:43 AM, Leon Romanovsky <leonro@mellanox.com> wrote: > On Sat, Jul 16, 2016 at 01:29:20PM +0530, Bhaktipriya Shridhar wrote: >> The workqueue health->wq was used as per device private health thread. >> This was done so that system error handling could be processed >> concurrently. > > Not exactly, AFAIK it was intended to perform delayed work and not > relevant to concurrency. > >> The workqueue has a single workitem(&health->work) and >> hence doesn't require ordering. It is involved in handling the health of >> the deviceand is not being used on a memory reclaim path. >> Hence, the singlethreaded workqueue has been replaced with the use of >> system_wq. > > Yes > >> >> System workqueues have been able to handle high level of concurrency >> for a long time now and hence it's not required to have a singlethreaded >> workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue >> created with create_singlethread_workqueue(), system_wq allows multiple >> work items to overlap executions even on the same CPU; however, a >> per-cpu workqueue doesn't have any CPU locality or global ordering >> guarantee unless the target CPU is explicitly specified and thus the >> increase of local concurrency shouldn't make any difference. > > Not relevant. > >> >> Work item has been flushed in mlx5_health_cleanup() to ensure that >> there are no pending tasks while disconnecting the driver. >> >> Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com> >> --- >> drivers/net/ethernet/mellanox/mlx5/core/health.c | 7 ++----- >> 1 file changed, 2 insertions(+), 5 deletions(-) >> >> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/health.c b/drivers/net/ethernet/mellanox/mlx5/core/health.c >> index 42d16b9..9acbccf 100644 >> --- a/drivers/net/ethernet/mellanox/mlx5/core/health.c >> +++ b/drivers/net/ethernet/mellanox/mlx5/core/health.c >> @@ -267,7 +267,7 @@ static void poll_health(unsigned long data) >> if (in_fatal(dev) && !health->sick) { >> health->sick = true; >> print_health_info(dev); >> - queue_work(health->wq, &health->work); >> + schedule_work(&health->work); >> } >> } >> >> @@ -296,7 +296,7 @@ void mlx5_health_cleanup(struct mlx5_core_dev *dev) >> { >> struct mlx5_core_health *health = &dev->priv.health; >> >> - destroy_workqueue(health->wq); >> + flush_work(&health->work); >> } >> >> int mlx5_health_init(struct mlx5_core_dev *dev) >> @@ -311,10 +311,7 @@ int mlx5_health_init(struct mlx5_core_dev *dev) >> >> strcpy(name, "mlx5_health"); >> strcat(name, dev_name(&dev->pdev->dev)); >> - health->wq = create_singlethread_workqueue(name); >> kfree(name); > > You need to remove "name" initialization/usage too. > It is not needed. > >> - if (!health->wq) >> - return -ENOMEM; >> >> INIT_WORK(&health->work, health_care); >> >> -- >> 2.1.4 >> >> -- >> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/health.c b/drivers/net/ethernet/mellanox/mlx5/core/health.c index 42d16b9..9acbccf 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/health.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/health.c @@ -267,7 +267,7 @@ static void poll_health(unsigned long data) if (in_fatal(dev) && !health->sick) { health->sick = true; print_health_info(dev); - queue_work(health->wq, &health->work); + schedule_work(&health->work); } } @@ -296,7 +296,7 @@ void mlx5_health_cleanup(struct mlx5_core_dev *dev) { struct mlx5_core_health *health = &dev->priv.health; - destroy_workqueue(health->wq); + flush_work(&health->work); } int mlx5_health_init(struct mlx5_core_dev *dev) @@ -311,10 +311,7 @@ int mlx5_health_init(struct mlx5_core_dev *dev) strcpy(name, "mlx5_health"); strcat(name, dev_name(&dev->pdev->dev)); - health->wq = create_singlethread_workqueue(name); kfree(name); - if (!health->wq) - return -ENOMEM; INIT_WORK(&health->work, health_care);
The workqueue health->wq was used as per device private health thread. This was done so that system error handling could be processed concurrently. The workqueue has a single workitem(&health->work) and hence doesn't require ordering. It is involved in handling the health of the deviceand is not being used on a memory reclaim path. Hence, the singlethreaded workqueue has been replaced with the use of system_wq. System workqueues have been able to handle high level of concurrency for a long time now and hence it's not required to have a singlethreaded workqueue just to gain concurrency. Unlike a dedicated per-cpu workqueue created with create_singlethread_workqueue(), system_wq allows multiple work items to overlap executions even on the same CPU; however, a per-cpu workqueue doesn't have any CPU locality or global ordering guarantee unless the target CPU is explicitly specified and thus the increase of local concurrency shouldn't make any difference. Work item has been flushed in mlx5_health_cleanup() to ensure that there are no pending tasks while disconnecting the driver. Signed-off-by: Bhaktipriya Shridhar <bhaktipriya96@gmail.com> --- drivers/net/ethernet/mellanox/mlx5/core/health.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) -- 2.1.4