Re: [PATCH] acpi: check the online state of all children in container

Re: [PATCH] acpi: check the online state of all children in container

[Rafael J. Wysocki]

On Wednesday, March 22, 2017 09:01:48 AM Lee, Chun-Yi wrote:
> Just checking the state of container is not enough to confirm that
> the whole container is offlined.

And why is that so?

> Kernel should checks all children's
> offline state as the logic in acpi_container_offline().
> 
> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
> Cc: Len Brown <lenb@kernel.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: Jiri Kosina <jkosina@suse.cz>
> Signed-off-by: "Lee, Chun-Yi" <jlee@suse.com>
> ---
>  drivers/acpi/scan.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 1926918..f08ca31 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -260,13 +260,15 @@ static int acpi_scan_try_to_offline(struct acpi_device *device)
>  static int acpi_scan_hot_remove(struct acpi_device *device)
>  {
>  	acpi_handle handle = device->handle;
> +	struct acpi_device *child;
>  	unsigned long long sta;
>  	acpi_status status;
>  
>  	if (device->handler && device->handler->hotplug.demand_offline
>  	    && !acpi_force_hot_remove) {
> -		if (!acpi_scan_is_offline(device, true))
> -			return -EBUSY;
> +		list_for_each_entry(child, &device->children, node)
> +			if (!acpi_scan_is_offline(child, false))
> +				return -EBUSY;
>  	} else {
>  		int error = acpi_scan_try_to_offline(device);
>  		if (error)
> 

[PATCH] acpi: check the online state of all children in container

[Lee, Chun-Yi]

Just checking the state of container is not enough to confirm that
the whole container is offlined. Kernel should checks all children's
offline state as the logic in acpi_container_offline().

Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Len Brown <lenb@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: "Lee, Chun-Yi" <jlee@suse.com>
---
 drivers/acpi/scan.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 1926918..f08ca31 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -260,13 +260,15 @@ static int acpi_scan_try_to_offline(struct acpi_device *device)
 static int acpi_scan_hot_remove(struct acpi_device *device)
 {
 	acpi_handle handle = device->handle;
+	struct acpi_device *child;
 	unsigned long long sta;
 	acpi_status status;
 
 	if (device->handler && device->handler->hotplug.demand_offline
 	    && !acpi_force_hot_remove) {
-		if (!acpi_scan_is_offline(device, true))
-			return -EBUSY;
+		list_for_each_entry(child, &device->children, node)
+			if (!acpi_scan_is_offline(child, false))
+				return -EBUSY;
 	} else {
 		int error = acpi_scan_try_to_offline(device);
 		if (error)
-- 
2.10.2

Re: [PATCH] acpi: check the online state of all children in container

[joeyli]

On Wed, Mar 22, 2017 at 01:58:30AM +0100, Rafael J. Wysocki wrote:
> On Wednesday, March 22, 2017 09:01:48 AM Lee, Chun-Yi wrote:
> > Just checking the state of container is not enough to confirm that
> > the whole container is offlined.
> 
> And why is that so?
>

Actually there does not have real kernel issue triggered by this code now.
I reviewed code and found the difference between acpi_container_offline().

Considering a container that it includes devices and sub-containers
like this:

Scope (_SB)
    Device (MODU)
	Name (_HID, "ACPI0004")		<===  main-container
        Device (PCIE)
		Name (_HID, EisaId ("PNP0A08"))
        Device (SUBM)
            	Name (_HID, "ACPI0004")		<=== sub-container
		Device (MEM0)
			Name (_HID, EisaId ("PNP0C80"))
	...

The original code checks the physical nodes on the main container but
doesn't check children's physical nodes. So, it may happen the sub-container
didn't offline but the offline checking of main container is pass. 

Please kindly direct me if I misunderstood or missed any detail in the codes
about physcial node and container offline.

Thank a lot!
Joey Lee