When the system is going to S5 resuming devices is unnecessary at
best and may cause problems for power consumption at S5 at worst.
Check the target `system_state` in the pci driver shutdown() callback
and skip the runtime resume step if the system is being powered off.
Cc: [email protected]
Signed-off-by: Mario Limonciello <[email protected]>
---
drivers/pci/pci-driver.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
index 51ec9e7e784f..5a6f7a786952 100644
--- a/drivers/pci/pci-driver.c
+++ b/drivers/pci/pci-driver.c
@@ -507,7 +507,9 @@ static void pci_device_shutdown(struct device *dev)
struct pci_dev *pci_dev = to_pci_dev(dev);
struct pci_driver *drv = pci_dev->driver;
- pm_runtime_resume(dev);
+ /* If system will go to S5, don't wake up PCI devices for shutdown */
+ if (system_state != SYSTEM_POWER_OFF)
+ pm_runtime_resume(dev);
if (drv && drv->shutdown)
drv->shutdown(pci_dev);
--
2.34.1
On Wed, Dec 13, 2023 at 7:27 PM Mario Limonciello
<[email protected]> wrote:
>
> When the system is going to S5 resuming devices is unnecessary at
> best and may cause problems for power consumption at S5 at worst.
>
> Check the target `system_state` in the pci driver shutdown() callback
> and skip the runtime resume step if the system is being powered off.
>
> Cc: [email protected]
> Signed-off-by: Mario Limonciello <[email protected]>
> ---
> drivers/pci/pci-driver.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
> index 51ec9e7e784f..5a6f7a786952 100644
> --- a/drivers/pci/pci-driver.c
> +++ b/drivers/pci/pci-driver.c
> @@ -507,7 +507,9 @@ static void pci_device_shutdown(struct device *dev)
> struct pci_dev *pci_dev = to_pci_dev(dev);
> struct pci_driver *drv = pci_dev->driver;
>
> - pm_runtime_resume(dev);
> + /* If system will go to S5, don't wake up PCI devices for shutdown */
> + if (system_state != SYSTEM_POWER_OFF)
> + pm_runtime_resume(dev);
What's the guarantee that the driver's shutdown callback will not
access MMIO regions of a runtime-suspended device?
> if (drv && drv->shutdown)
> drv->shutdown(pci_dev);
> --
On 12/13/2023 12:31, Rafael J. Wysocki wrote:
> On Wed, Dec 13, 2023 at 7:27 PM Mario Limonciello
> <[email protected]> wrote:
>>
>> When the system is going to S5 resuming devices is unnecessary at
>> best and may cause problems for power consumption at S5 at worst.
>>
>> Check the target `system_state` in the pci driver shutdown() callback
>> and skip the runtime resume step if the system is being powered off.
>>
>> Cc: [email protected]
>> Signed-off-by: Mario Limonciello <[email protected]>
>> ---
>> drivers/pci/pci-driver.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
>> index 51ec9e7e784f..5a6f7a786952 100644
>> --- a/drivers/pci/pci-driver.c
>> +++ b/drivers/pci/pci-driver.c
>> @@ -507,7 +507,9 @@ static void pci_device_shutdown(struct device *dev)
>> struct pci_dev *pci_dev = to_pci_dev(dev);
>> struct pci_driver *drv = pci_dev->driver;
>>
>> - pm_runtime_resume(dev);
>> + /* If system will go to S5, don't wake up PCI devices for shutdown */
>> + if (system_state != SYSTEM_POWER_OFF)
>> + pm_runtime_resume(dev);
>
> What's the guarantee that the driver's shutdown callback will not
> access MMIO regions of a runtime-suspended device?
This is something that may need to be audited.
I have an expectation that drivers will runtime resume a device before
they try to access it.
>
>> if (drv && drv->shutdown)
>> drv->shutdown(pci_dev);
>> --
On Wed, Dec 13, 2023 at 7:33 PM Mario Limonciello
<[email protected]> wrote:
>
> On 12/13/2023 12:31, Rafael J. Wysocki wrote:
> > On Wed, Dec 13, 2023 at 7:27 PM Mario Limonciello
> > <[email protected]> wrote:
> >>
> >> When the system is going to S5 resuming devices is unnecessary at
> >> best and may cause problems for power consumption at S5 at worst.
> >>
> >> Check the target `system_state` in the pci driver shutdown() callback
> >> and skip the runtime resume step if the system is being powered off.
> >>
> >> Cc: [email protected]
> >> Signed-off-by: Mario Limonciello <[email protected]>
> >> ---
> >> drivers/pci/pci-driver.c | 4 +++-
> >> 1 file changed, 3 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
> >> index 51ec9e7e784f..5a6f7a786952 100644
> >> --- a/drivers/pci/pci-driver.c
> >> +++ b/drivers/pci/pci-driver.c
> >> @@ -507,7 +507,9 @@ static void pci_device_shutdown(struct device *dev)
> >> struct pci_dev *pci_dev = to_pci_dev(dev);
> >> struct pci_driver *drv = pci_dev->driver;
> >>
> >> - pm_runtime_resume(dev);
> >> + /* If system will go to S5, don't wake up PCI devices for shutdown */
> >> + if (system_state != SYSTEM_POWER_OFF)
> >> + pm_runtime_resume(dev);
> >
> > What's the guarantee that the driver's shutdown callback will not
> > access MMIO regions of a runtime-suspended device?
>
> This is something that may need to be audited.
>
> I have an expectation that drivers will runtime resume a device before
> they try to access it.
But they don't need to do it, because the core does it right now,
which is my point.