[PATCH] ACPI / hotplug / PCI: Take runtime PM autosuspend into account

[PATCH] ACPI / hotplug / PCI: Take runtime PM autosuspend into account

[Mika Westerberg]

PCIe ports (the only ones we do runtime PM) are using runtime PM
autosuspend to keep the port powered on for a while after it becomes
idle. However, ACPI hotplug does not take this into account so if we get
multiple hotplug events in a short period of time we may be powering
ports on and off and then back on unnecessarily.

For this reason call pm_runtime_put_autosuspend() for them (with the
accompanying pm_runtime_mark_last_busy()).

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
---
 drivers/pci/hotplug/acpiphp_glue.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/pci/hotplug/acpiphp_glue.c b/drivers/pci/hotplug/acpiphp_glue.c
index b3869951c0eb..c8264af40643 100644
--- a/drivers/pci/hotplug/acpiphp_glue.c
+++ b/drivers/pci/hotplug/acpiphp_glue.c
@@ -672,7 +672,8 @@ static void trim_stale_devices(struct pci_dev *dev)
 		list_for_each_entry_safe_reverse(child, tmp, &bus->devices, bus_list)
 			trim_stale_devices(child);
 
-		pm_runtime_put(&dev->dev);
+		pm_runtime_mark_last_busy(&dev->dev);
+		pm_runtime_put_autosuspend(&dev->dev);
 	}
 }
 
@@ -714,8 +715,10 @@ static void acpiphp_check_bridge(struct acpiphp_bridge *bridge)
 		}
 	}
 
-	if (bridge->pci_dev)
-		pm_runtime_put(&bridge->pci_dev->dev);
+	if (bridge->pci_dev) {
+		pm_runtime_mark_last_busy(&bridge->pci_dev->dev);
+		pm_runtime_put_autosuspend(&bridge->pci_dev->dev);
+	}
 }
 
 /*
-- 
2.24.0

Re: [PATCH] ACPI / hotplug / PCI: Take runtime PM autosuspend into account

[Rafael J. Wysocki]

On Wed, Nov 13, 2019 at 4:41 PM Mika Westerberg
<mika.westerberg@linux.intel.com> wrote:
>
> PCIe ports (the only ones we do runtime PM) are using runtime PM
> autosuspend to keep the port powered on for a while after it becomes
> idle. However, ACPI hotplug does not take this into account so if we get
> multiple hotplug events in a short period of time we may be powering
> ports on and off and then back on unnecessarily.
>
> For this reason call pm_runtime_put_autosuspend() for them (with the
> accompanying pm_runtime_mark_last_busy()).
>
> Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>

Makes sense:

Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

> ---
>  drivers/pci/hotplug/acpiphp_glue.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/pci/hotplug/acpiphp_glue.c b/drivers/pci/hotplug/acpiphp_glue.c
> index b3869951c0eb..c8264af40643 100644
> --- a/drivers/pci/hotplug/acpiphp_glue.c
> +++ b/drivers/pci/hotplug/acpiphp_glue.c
> @@ -672,7 +672,8 @@ static void trim_stale_devices(struct pci_dev *dev)
>                 list_for_each_entry_safe_reverse(child, tmp, &bus->devices, bus_list)
>                         trim_stale_devices(child);
>
> -               pm_runtime_put(&dev->dev);
> +               pm_runtime_mark_last_busy(&dev->dev);
> +               pm_runtime_put_autosuspend(&dev->dev);
>         }
>  }
>
> @@ -714,8 +715,10 @@ static void acpiphp_check_bridge(struct acpiphp_bridge *bridge)
>                 }
>         }
>
> -       if (bridge->pci_dev)
> -               pm_runtime_put(&bridge->pci_dev->dev);
> +       if (bridge->pci_dev) {
> +               pm_runtime_mark_last_busy(&bridge->pci_dev->dev);
> +               pm_runtime_put_autosuspend(&bridge->pci_dev->dev);
> +       }
>  }
>
>  /*
> --