* Re: [PATCH] PCI:PM: Support platforms that do not implement ACPI
[not found] <20230609023038.61388-1-zhiren.chen@mediatek.com>
@ 2023-06-09 11:49 ` Bjorn Helgaas
2023-06-09 17:58 ` Rafael J. Wysocki
0 siblings, 1 reply; 5+ messages in thread
From: Bjorn Helgaas @ 2023-06-09 11:49 UTC (permalink / raw)
To: Zhiren Chen
Cc: Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
Rafael J. Wysocki, linux-pci, linux-pm
[+cc Rafael, linux-pm]
On Fri, Jun 09, 2023 at 10:30:38AM +0800, Zhiren Chen wrote:
> From: Zhiren Chen <Zhiren.Chen@mediatek.com>
>
> The platform_pci_choose_state function and other low-level platform
> interfaces used by PCI power management processing did not take into
> account non-ACPI-supported platforms. This shortcoming can result in
> limitations and issues.
>
> For example, in embedded systems like smartphones, a PCI device can be
> shared by multiple processors for different purposes. The PCI device and
> some of the processors are controlled by Linux, while the rest of the
> processors runs its own operating system.
> When Linux initiates system-level sleep, if it does not consider the
> working state of the shared PCI device and forcefully sets the PCI device
> state to D3, it will affect the functionality of other processors that
> are currently using the PCI device.
>
> To address this problem, an interface should be created for PCI devices
> that don't support ACPI to enable accurate reporting of the power state
> during the PCI PM handling process.
>
> Signed-off-by: Zhiren Chen <Zhiren.Chen@mediatek.com>
> ---
> drivers/pci/pci.c | 24 ++++++++++++++++++++++++
> drivers/pci/pci.h | 40 ++++++++++++++++++++++++++++++++++++++++
> include/linux/pci.h | 1 +
> 3 files changed, 65 insertions(+)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 5ede93222bc1..9f03406f3081 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -1014,6 +1014,9 @@ static void pci_restore_bars(struct pci_dev *dev)
>
> static inline bool platform_pci_power_manageable(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->is_manageable)
> + return dev->platform_pm_ops->is_manageable(dev);
> +
> if (pci_use_mid_pm())
> return true;
>
> @@ -1023,6 +1026,9 @@ static inline bool platform_pci_power_manageable(struct pci_dev *dev)
> static inline int platform_pci_set_power_state(struct pci_dev *dev,
> pci_power_t t)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->set_state)
> + return dev->platform_pm_ops->set_state(dev, t);
> +
> if (pci_use_mid_pm())
> return mid_pci_set_power_state(dev, t);
>
> @@ -1031,6 +1037,9 @@ static inline int platform_pci_set_power_state(struct pci_dev *dev,
>
> static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->get_state)
> + return dev->platform_pm_ops->get_state(dev);
> +
> if (pci_use_mid_pm())
> return mid_pci_get_power_state(dev);
>
> @@ -1039,12 +1048,18 @@ static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
>
> static inline void platform_pci_refresh_power_state(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->refresh_state)
> + dev->platform_pm_ops->refresh_state(dev);
> +
> if (!pci_use_mid_pm())
> acpi_pci_refresh_power_state(dev);
> }
>
> static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->choose_state)
> + return dev->platform_pm_ops->choose_state(dev);
> +
> if (pci_use_mid_pm())
> return PCI_POWER_ERROR;
>
> @@ -1053,6 +1068,9 @@ static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
>
> static inline int platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->set_wakeup)
> + return dev->platform_pm_ops->set_wakeup(dev, enable);
> +
> if (pci_use_mid_pm())
> return PCI_POWER_ERROR;
>
> @@ -1061,6 +1079,9 @@ static inline int platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
>
> static inline bool platform_pci_need_resume(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->need_resume)
> + return dev->platform_pm_ops->need_resume(dev);
> +
> if (pci_use_mid_pm())
> return false;
>
> @@ -1069,6 +1090,9 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)
>
> static inline bool platform_pci_bridge_d3(struct pci_dev *dev)
> {
> + if (dev->platform_pm_ops && dev->platform_pm_ops->bridge_d3)
> + return dev->platform_pm_ops->bridge_d3(dev);
> +
> if (pci_use_mid_pm())
> return false;
>
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> index 2475098f6518..85154470c083 100644
> --- a/drivers/pci/pci.h
> +++ b/drivers/pci/pci.h
> @@ -71,6 +71,42 @@ struct pci_cap_saved_state *pci_find_saved_ext_cap(struct pci_dev *dev,
> */
> #define PCI_RESET_WAIT 1000 /* msec */
>
> +/**
> + * struct pci_platform_pm_ops - Firmware PM callbacks
> + *
> + * @is_manageable: returns 'true' if given device is power manageable by the
> + * platform firmware
> + *
> + * @set_state: invokes the platform firmware to set the device's power state
> + *
> + * @get_state: queries the platform firmware for a device's current power state
> + *
> + * @choose_state: returns PCI power state of given device preferred by the
> + * platform; to be used during system-wide transitions from a
> + * sleeping state to the working state and vice versa
> + *
> + * @set_wakeup: enables/disables wakeup capability for the device
> + *
> + * @need_resume: returns 'true' if the given device (which is currently
> + * suspended) needs to be resumed to be configured for system
> + * wakeup.
> + *
> + * @bridge_d3: return 'true' if given device supoorts D3 when it is a bridge
> + *
> + * @refresh_state: refresh the given device's power state
> + *
> + */
> +struct pci_platform_pm_ops {
> + bool (*is_manageable)(struct pci_dev *dev);
> + int (*set_state)(struct pci_dev *dev, pci_power_t state);
> + pci_power_t (*get_state)(struct pci_dev *dev);
> + pci_power_t (*choose_state)(struct pci_dev *dev);
> + int (*set_wakeup)(struct pci_dev *dev, bool enable);
> + bool (*need_resume)(struct pci_dev *dev);
> + bool (*bridge_d3)(struct pci_dev *dev);
> + void (*refresh_state)(struct pci_dev *dev);
> +};
> +
> void pci_update_current_state(struct pci_dev *dev, pci_power_t state);
> void pci_refresh_power_state(struct pci_dev *dev);
> int pci_power_up(struct pci_dev *dev);
> @@ -96,6 +132,10 @@ void pci_bridge_d3_update(struct pci_dev *dev);
> void pci_bridge_reconfigure_ltr(struct pci_dev *dev);
> int pci_bridge_wait_for_secondary_bus(struct pci_dev *dev, char *reset_type);
>
> +static inline void pci_set_platform_pm(struct pci_dev *dev, struct pci_platform_pm_ops *ops)
> +{
> + dev->platform_pm_ops = ops;
> +}
> static inline void pci_wakeup_event(struct pci_dev *dev)
> {
> /* Wait 100 ms before the system can be put into a sleep state. */
> diff --git a/include/linux/pci.h b/include/linux/pci.h
> index 60b8772b5bd4..a0171f1abf2f 100644
> --- a/include/linux/pci.h
> +++ b/include/linux/pci.h
> @@ -327,6 +327,7 @@ struct pci_dev {
> void *sysdata; /* Hook for sys-specific extension */
> struct proc_dir_entry *procent; /* Device entry in /proc/bus/pci */
> struct pci_slot *slot; /* Physical slot this device is in */
> + struct pci_platform_pm_ops *platform_pm_ops;
>
> unsigned int devfn; /* Encoded device & function index */
> unsigned short vendor;
> --
> 2.17.0
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] PCI:PM: Support platforms that do not implement ACPI
2023-06-09 11:49 ` [PATCH] PCI:PM: Support platforms that do not implement ACPI Bjorn Helgaas
@ 2023-06-09 17:58 ` Rafael J. Wysocki
2023-06-13 7:57 ` Zhiren Chen (陈志仁)
0 siblings, 1 reply; 5+ messages in thread
From: Rafael J. Wysocki @ 2023-06-09 17:58 UTC (permalink / raw)
To: Bjorn Helgaas, Zhiren Chen
Cc: Bjorn Helgaas, Matthias Brugger, AngeloGioacchino Del Regno,
Rafael J. Wysocki, linux-pci, linux-pm
On Fri, Jun 9, 2023 at 1:49 PM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> [+cc Rafael, linux-pm]
>
> On Fri, Jun 09, 2023 at 10:30:38AM +0800, Zhiren Chen wrote:
> > From: Zhiren Chen <Zhiren.Chen@mediatek.com>
> >
> > The platform_pci_choose_state function and other low-level platform
> > interfaces used by PCI power management processing did not take into
> > account non-ACPI-supported platforms. This shortcoming can result in
> > limitations and issues.
> >
> > For example, in embedded systems like smartphones, a PCI device can be
> > shared by multiple processors for different purposes. The PCI device and
> > some of the processors are controlled by Linux, while the rest of the
> > processors runs its own operating system.
> > When Linux initiates system-level sleep, if it does not consider the
> > working state of the shared PCI device and forcefully sets the PCI device
> > state to D3, it will affect the functionality of other processors that
> > are currently using the PCI device.
> >
> > To address this problem, an interface should be created for PCI devices
> > that don't support ACPI to enable accurate reporting of the power state
> > during the PCI PM handling process.
> >
> > Signed-off-by: Zhiren Chen <Zhiren.Chen@mediatek.com>
Something like the pci_platform_pm_ops introduced here had been there
for several years and the only users of it known to me were ACPI and
Intel MID, which is why it was dropped.
I would like to see the platform code using these new callbacks in the
first place.
> > ---
> > drivers/pci/pci.c | 24 ++++++++++++++++++++++++
> > drivers/pci/pci.h | 40 ++++++++++++++++++++++++++++++++++++++++
> > include/linux/pci.h | 1 +
> > 3 files changed, 65 insertions(+)
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index 5ede93222bc1..9f03406f3081 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -1014,6 +1014,9 @@ static void pci_restore_bars(struct pci_dev *dev)
> >
> > static inline bool platform_pci_power_manageable(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->is_manageable)
> > + return dev->platform_pm_ops->is_manageable(dev);
> > +
> > if (pci_use_mid_pm())
> > return true;
> >
> > @@ -1023,6 +1026,9 @@ static inline bool platform_pci_power_manageable(struct pci_dev *dev)
> > static inline int platform_pci_set_power_state(struct pci_dev *dev,
> > pci_power_t t)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->set_state)
> > + return dev->platform_pm_ops->set_state(dev, t);
> > +
> > if (pci_use_mid_pm())
> > return mid_pci_set_power_state(dev, t);
> >
> > @@ -1031,6 +1037,9 @@ static inline int platform_pci_set_power_state(struct pci_dev *dev,
> >
> > static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->get_state)
> > + return dev->platform_pm_ops->get_state(dev);
> > +
> > if (pci_use_mid_pm())
> > return mid_pci_get_power_state(dev);
> >
> > @@ -1039,12 +1048,18 @@ static inline pci_power_t platform_pci_get_power_state(struct pci_dev *dev)
> >
> > static inline void platform_pci_refresh_power_state(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->refresh_state)
> > + dev->platform_pm_ops->refresh_state(dev);
> > +
> > if (!pci_use_mid_pm())
> > acpi_pci_refresh_power_state(dev);
> > }
> >
> > static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->choose_state)
> > + return dev->platform_pm_ops->choose_state(dev);
> > +
> > if (pci_use_mid_pm())
> > return PCI_POWER_ERROR;
> >
> > @@ -1053,6 +1068,9 @@ static inline pci_power_t platform_pci_choose_state(struct pci_dev *dev)
> >
> > static inline int platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->set_wakeup)
> > + return dev->platform_pm_ops->set_wakeup(dev, enable);
> > +
> > if (pci_use_mid_pm())
> > return PCI_POWER_ERROR;
> >
> > @@ -1061,6 +1079,9 @@ static inline int platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
> >
> > static inline bool platform_pci_need_resume(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->need_resume)
> > + return dev->platform_pm_ops->need_resume(dev);
> > +
> > if (pci_use_mid_pm())
> > return false;
> >
> > @@ -1069,6 +1090,9 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)
> >
> > static inline bool platform_pci_bridge_d3(struct pci_dev *dev)
> > {
> > + if (dev->platform_pm_ops && dev->platform_pm_ops->bridge_d3)
> > + return dev->platform_pm_ops->bridge_d3(dev);
> > +
> > if (pci_use_mid_pm())
> > return false;
> >
> > diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> > index 2475098f6518..85154470c083 100644
> > --- a/drivers/pci/pci.h
> > +++ b/drivers/pci/pci.h
> > @@ -71,6 +71,42 @@ struct pci_cap_saved_state *pci_find_saved_ext_cap(struct pci_dev *dev,
> > */
> > #define PCI_RESET_WAIT 1000 /* msec */
> >
> > +/**
> > + * struct pci_platform_pm_ops - Firmware PM callbacks
> > + *
> > + * @is_manageable: returns 'true' if given device is power manageable by the
> > + * platform firmware
> > + *
> > + * @set_state: invokes the platform firmware to set the device's power state
> > + *
> > + * @get_state: queries the platform firmware for a device's current power state
> > + *
> > + * @choose_state: returns PCI power state of given device preferred by the
> > + * platform; to be used during system-wide transitions from a
> > + * sleeping state to the working state and vice versa
> > + *
> > + * @set_wakeup: enables/disables wakeup capability for the device
> > + *
> > + * @need_resume: returns 'true' if the given device (which is currently
> > + * suspended) needs to be resumed to be configured for system
> > + * wakeup.
> > + *
> > + * @bridge_d3: return 'true' if given device supoorts D3 when it is a bridge
> > + *
> > + * @refresh_state: refresh the given device's power state
> > + *
> > + */
> > +struct pci_platform_pm_ops {
> > + bool (*is_manageable)(struct pci_dev *dev);
> > + int (*set_state)(struct pci_dev *dev, pci_power_t state);
> > + pci_power_t (*get_state)(struct pci_dev *dev);
> > + pci_power_t (*choose_state)(struct pci_dev *dev);
> > + int (*set_wakeup)(struct pci_dev *dev, bool enable);
> > + bool (*need_resume)(struct pci_dev *dev);
> > + bool (*bridge_d3)(struct pci_dev *dev);
> > + void (*refresh_state)(struct pci_dev *dev);
> > +};
> > +
> > void pci_update_current_state(struct pci_dev *dev, pci_power_t state);
> > void pci_refresh_power_state(struct pci_dev *dev);
> > int pci_power_up(struct pci_dev *dev);
> > @@ -96,6 +132,10 @@ void pci_bridge_d3_update(struct pci_dev *dev);
> > void pci_bridge_reconfigure_ltr(struct pci_dev *dev);
> > int pci_bridge_wait_for_secondary_bus(struct pci_dev *dev, char *reset_type);
> >
> > +static inline void pci_set_platform_pm(struct pci_dev *dev, struct pci_platform_pm_ops *ops)
> > +{
> > + dev->platform_pm_ops = ops;
> > +}
> > static inline void pci_wakeup_event(struct pci_dev *dev)
> > {
> > /* Wait 100 ms before the system can be put into a sleep state. */
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index 60b8772b5bd4..a0171f1abf2f 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -327,6 +327,7 @@ struct pci_dev {
> > void *sysdata; /* Hook for sys-specific extension */
> > struct proc_dir_entry *procent; /* Device entry in /proc/bus/pci */
> > struct pci_slot *slot; /* Physical slot this device is in */
> > + struct pci_platform_pm_ops *platform_pm_ops;
> >
> > unsigned int devfn; /* Encoded device & function index */
> > unsigned short vendor;
> > --
> > 2.17.0
> >
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] PCI:PM: Support platforms that do not implement ACPI
2023-06-09 17:58 ` Rafael J. Wysocki
@ 2023-06-13 7:57 ` Zhiren Chen (陈志仁)
2023-06-13 14:30 ` Rafael J. Wysocki
0 siblings, 1 reply; 5+ messages in thread
From: Zhiren Chen (陈志仁) @ 2023-06-13 7:57 UTC (permalink / raw)
To: rafael@kernel.org, helgaas@kernel.org
Cc: linux-pm@vger.kernel.org, Haijun Liu (刘海军),
rjw@rjwysocki.net, Mingchuang Qiao (乔明闯),
Lambert Wang (王伟), matthias.bgg@gmail.com,
linux-pci@vger.kernel.org, bhelgaas@google.com,
angelogioacchino.delregno@collabora.com
On Fri, 2023-06-09 at 19:58 +0200, Rafael J. Wysocki wrote:
>
> On Fri, Jun 9, 2023 at 1:49 PM Bjorn Helgaas <helgaas@kernel.org>
> wrote:
> >
> > [+cc Rafael, linux-pm]
> >
> > On Fri, Jun 09, 2023 at 10:30:38AM +0800, Zhiren Chen wrote:
> > > From: Zhiren Chen <Zhiren.Chen@mediatek.com>
> > >
> > > The platform_pci_choose_state function and other low-level
> platform
> > > interfaces used by PCI power management processing did not take
> into
> > > account non-ACPI-supported platforms. This shortcoming can result
> in
> > > limitations and issues.
> > >
> > > For example, in embedded systems like smartphones, a PCI device
> can be
> > > shared by multiple processors for different purposes. The PCI
> device and
> > > some of the processors are controlled by Linux, while the rest of
> the
> > > processors runs its own operating system.
> > > When Linux initiates system-level sleep, if it does not consider
> the
> > > working state of the shared PCI device and forcefully sets the
> PCI device
> > > state to D3, it will affect the functionality of other processors
> that
> > > are currently using the PCI device.
> > >
> > > To address this problem, an interface should be created for PCI
> devices
> > > that don't support ACPI to enable accurate reporting of the power
> state
> > > during the PCI PM handling process.
> > >
> > > Signed-off-by: Zhiren Chen <Zhiren.Chen@mediatek.com>
>
> Something like the pci_platform_pm_ops introduced here had been there
> for several years and the only users of it known to me were ACPI and
> Intel MID, which is why it was dropped.
>
> I would like to see the platform code using these new callbacks in
> the
> first place.
>
I think that more and more embedded products will use PCI devices to
achieve higher performance for data transfer, and these products may
not necessarily support ACPI.
When developing the Mediatek T8xx modem chip driver, I found that there
was no good way for T8xx to skip D3 setting in certain PM scenarios.
Best Regards,
Zhiren
> > > ---
> > > drivers/pci/pci.c | 24 ++++++++++++++++++++++++
> > > drivers/pci/pci.h | 40
> ++++++++++++++++++++++++++++++++++++++++
> > > include/linux/pci.h | 1 +
> > > 3 files changed, 65 insertions(+)
> > >
> > > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > > index 5ede93222bc1..9f03406f3081 100644
> > > --- a/drivers/pci/pci.c
> > > +++ b/drivers/pci/pci.c
> > > @@ -1014,6 +1014,9 @@ static void pci_restore_bars(struct pci_dev
> *dev)
> > >
> > > static inline bool platform_pci_power_manageable(struct pci_dev
> *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >is_manageable)
> > > + return dev->platform_pm_ops->is_manageable(dev);
> > > +
> > > if (pci_use_mid_pm())
> > > return true;
> > >
> > > @@ -1023,6 +1026,9 @@ static inline bool
> platform_pci_power_manageable(struct pci_dev *dev)
> > > static inline int platform_pci_set_power_state(struct pci_dev
> *dev,
> > > pci_power_t t)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >set_state)
> > > + return dev->platform_pm_ops->set_state(dev, t);
> > > +
> > > if (pci_use_mid_pm())
> > > return mid_pci_set_power_state(dev, t);
> > >
> > > @@ -1031,6 +1037,9 @@ static inline int
> platform_pci_set_power_state(struct pci_dev *dev,
> > >
> > > static inline pci_power_t platform_pci_get_power_state(struct
> pci_dev *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >get_state)
> > > + return dev->platform_pm_ops->get_state(dev);
> > > +
> > > if (pci_use_mid_pm())
> > > return mid_pci_get_power_state(dev);
> > >
> > > @@ -1039,12 +1048,18 @@ static inline pci_power_t
> platform_pci_get_power_state(struct pci_dev *dev)
> > >
> > > static inline void platform_pci_refresh_power_state(struct
> pci_dev *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >refresh_state)
> > > + dev->platform_pm_ops->refresh_state(dev);
> > > +
> > > if (!pci_use_mid_pm())
> > > acpi_pci_refresh_power_state(dev);
> > > }
> > >
> > > static inline pci_power_t platform_pci_choose_state(struct
> pci_dev *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >choose_state)
> > > + return dev->platform_pm_ops->choose_state(dev);
> > > +
> > > if (pci_use_mid_pm())
> > > return PCI_POWER_ERROR;
> > >
> > > @@ -1053,6 +1068,9 @@ static inline pci_power_t
> platform_pci_choose_state(struct pci_dev *dev)
> > >
> > > static inline int platform_pci_set_wakeup(struct pci_dev *dev,
> bool enable)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >set_wakeup)
> > > + return dev->platform_pm_ops->set_wakeup(dev,
> enable);
> > > +
> > > if (pci_use_mid_pm())
> > > return PCI_POWER_ERROR;
> > >
> > > @@ -1061,6 +1079,9 @@ static inline int
> platform_pci_set_wakeup(struct pci_dev *dev, bool enable)
> > >
> > > static inline bool platform_pci_need_resume(struct pci_dev *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >need_resume)
> > > + return dev->platform_pm_ops->need_resume(dev);
> > > +
> > > if (pci_use_mid_pm())
> > > return false;
> > >
> > > @@ -1069,6 +1090,9 @@ static inline bool
> platform_pci_need_resume(struct pci_dev *dev)
> > >
> > > static inline bool platform_pci_bridge_d3(struct pci_dev *dev)
> > > {
> > > + if (dev->platform_pm_ops && dev->platform_pm_ops-
> >bridge_d3)
> > > + return dev->platform_pm_ops->bridge_d3(dev);
> > > +
> > > if (pci_use_mid_pm())
> > > return false;
> > >
> > > diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> > > index 2475098f6518..85154470c083 100644
> > > --- a/drivers/pci/pci.h
> > > +++ b/drivers/pci/pci.h
> > > @@ -71,6 +71,42 @@ struct pci_cap_saved_state
> *pci_find_saved_ext_cap(struct pci_dev *dev,
> > > */
> > > #define PCI_RESET_WAIT 1000 /* msec */
> > >
> > > +/**
> > > + * struct pci_platform_pm_ops - Firmware PM callbacks
> > > + *
> > > + * @is_manageable: returns 'true' if given device is power
> manageable by the
> > > + * platform firmware
> > > + *
> > > + * @set_state: invokes the platform firmware to set the device's
> power state
> > > + *
> > > + * @get_state: queries the platform firmware for a device's
> current power state
> > > + *
> > > + * @choose_state: returns PCI power state of given device
> preferred by the
> > > + * platform; to be used during system-wide
> transitions from a
> > > + * sleeping state to the working state and vice
> versa
> > > + *
> > > + * @set_wakeup: enables/disables wakeup capability for the
> device
> > > + *
> > > + * @need_resume: returns 'true' if the given device (which is
> currently
> > > + * suspended) needs to be resumed to be configured for
> system
> > > + * wakeup.
> > > + *
> > > + * @bridge_d3: return 'true' if given device supoorts D3 when it
> is a bridge
> > > + *
> > > + * @refresh_state: refresh the given device's power state
> > > + *
> > > + */
> > > +struct pci_platform_pm_ops {
> > > + bool (*is_manageable)(struct pci_dev *dev);
> > > + int (*set_state)(struct pci_dev *dev, pci_power_t state);
> > > + pci_power_t (*get_state)(struct pci_dev *dev);
> > > + pci_power_t (*choose_state)(struct pci_dev *dev);
> > > + int (*set_wakeup)(struct pci_dev *dev, bool enable);
> > > + bool (*need_resume)(struct pci_dev *dev);
> > > + bool (*bridge_d3)(struct pci_dev *dev);
> > > + void (*refresh_state)(struct pci_dev *dev);
> > > +};
> > > +
> > > void pci_update_current_state(struct pci_dev *dev, pci_power_t
> state);
> > > void pci_refresh_power_state(struct pci_dev *dev);
> > > int pci_power_up(struct pci_dev *dev);
> > > @@ -96,6 +132,10 @@ void pci_bridge_d3_update(struct pci_dev
> *dev);
> > > void pci_bridge_reconfigure_ltr(struct pci_dev *dev);
> > > int pci_bridge_wait_for_secondary_bus(struct pci_dev *dev, char
> *reset_type);
> > >
> > > +static inline void pci_set_platform_pm(struct pci_dev *dev,
> struct pci_platform_pm_ops *ops)
> > > +{
> > > + dev->platform_pm_ops = ops;
> > > +}
> > > static inline void pci_wakeup_event(struct pci_dev *dev)
> > > {
> > > /* Wait 100 ms before the system can be put into a sleep
> state. */
> > > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > > index 60b8772b5bd4..a0171f1abf2f 100644
> > > --- a/include/linux/pci.h
> > > +++ b/include/linux/pci.h
> > > @@ -327,6 +327,7 @@ struct pci_dev {
> > > void *sysdata; /* Hook for sys-specific
> extension */
> > > struct proc_dir_entry *procent; /* Device entry in
> /proc/bus/pci */
> > > struct pci_slot *slot; /* Physical slot this
> device is in */
> > > + struct pci_platform_pm_ops *platform_pm_ops;
> > >
> > > unsigned int devfn; /* Encoded device &
> function index */
> > > unsigned short vendor;
> > > --
> > > 2.17.0
> > >
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] PCI:PM: Support platforms that do not implement ACPI
2023-06-13 7:57 ` Zhiren Chen (陈志仁)
@ 2023-06-13 14:30 ` Rafael J. Wysocki
2023-06-14 2:41 ` Zhiren Chen (陈志仁)
0 siblings, 1 reply; 5+ messages in thread
From: Rafael J. Wysocki @ 2023-06-13 14:30 UTC (permalink / raw)
To: Zhiren Chen (陈志仁)
Cc: rafael@kernel.org, helgaas@kernel.org, linux-pm@vger.kernel.org,
Haijun Liu (刘海军), rjw@rjwysocki.net,
Mingchuang Qiao (乔明闯),
Lambert Wang (王伟), matthias.bgg@gmail.com,
linux-pci@vger.kernel.org, bhelgaas@google.com,
angelogioacchino.delregno@collabora.com
On Tue, Jun 13, 2023 at 9:57 AM Zhiren Chen (陈志仁)
<Zhiren.Chen@mediatek.com> wrote:
>
> On Fri, 2023-06-09 at 19:58 +0200, Rafael J. Wysocki wrote:
> >
> > On Fri, Jun 9, 2023 at 1:49 PM Bjorn Helgaas <helgaas@kernel.org>
> > wrote:
> > >
> > > [+cc Rafael, linux-pm]
> > >
> > > On Fri, Jun 09, 2023 at 10:30:38AM +0800, Zhiren Chen wrote:
> > > > From: Zhiren Chen <Zhiren.Chen@mediatek.com>
> > > >
> > > > The platform_pci_choose_state function and other low-level
> > platform
> > > > interfaces used by PCI power management processing did not take
> > into
> > > > account non-ACPI-supported platforms. This shortcoming can result
> > in
> > > > limitations and issues.
> > > >
> > > > For example, in embedded systems like smartphones, a PCI device
> > can be
> > > > shared by multiple processors for different purposes. The PCI
> > device and
> > > > some of the processors are controlled by Linux, while the rest of
> > the
> > > > processors runs its own operating system.
> > > > When Linux initiates system-level sleep, if it does not consider
> > the
> > > > working state of the shared PCI device and forcefully sets the
> > PCI device
> > > > state to D3, it will affect the functionality of other processors
> > that
> > > > are currently using the PCI device.
> > > >
> > > > To address this problem, an interface should be created for PCI
> > devices
> > > > that don't support ACPI to enable accurate reporting of the power
> > state
> > > > during the PCI PM handling process.
> > > >
> > > > Signed-off-by: Zhiren Chen <Zhiren.Chen@mediatek.com>
> >
> > Something like the pci_platform_pm_ops introduced here had been there
> > for several years and the only users of it known to me were ACPI and
> > Intel MID, which is why it was dropped.
> >
> > I would like to see the platform code using these new callbacks in
> > the
> > first place.
> >
> I think that more and more embedded products will use PCI devices to
> achieve higher performance for data transfer, and these products may
> not necessarily support ACPI.
>
> When developing the Mediatek T8xx modem chip driver, I found that there
> was no good way for T8xx to skip D3 setting in certain PM scenarios.
Well, is there any code that you are planning to add to the mainline
Linux kernel that is going to use the proposed interface?
If not, the interface itself will not be useful in the mainline Linux kernel.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] PCI:PM: Support platforms that do not implement ACPI
2023-06-13 14:30 ` Rafael J. Wysocki
@ 2023-06-14 2:41 ` Zhiren Chen (陈志仁)
0 siblings, 0 replies; 5+ messages in thread
From: Zhiren Chen (陈志仁) @ 2023-06-14 2:41 UTC (permalink / raw)
To: rafael@kernel.org
Cc: linux-pm@vger.kernel.org, Haijun Liu (刘海军),
helgaas@kernel.org, Mingchuang Qiao (乔明闯),
Lambert Wang (王伟), matthias.bgg@gmail.com,
linux-pci@vger.kernel.org, bhelgaas@google.com,
angelogioacchino.delregno@collabora.com
On Tue, 2023-06-13 at 16:30 +0200, Rafael J. Wysocki wrote:
>
> On Tue, Jun 13, 2023 at 9:57 AM Zhiren Chen (陈志仁)
> <Zhiren.Chen@mediatek.com> wrote:
> >
> > On Fri, 2023-06-09 at 19:58 +0200, Rafael J. Wysocki wrote:
> > >
> > > On Fri, Jun 9, 2023 at 1:49 PM Bjorn Helgaas <helgaas@kernel.org
> >
> > > wrote:
> > > >
> > > > [+cc Rafael, linux-pm]
> > > >
> > > > On Fri, Jun 09, 2023 at 10:30:38AM +0800, Zhiren Chen wrote:
> > > > > From: Zhiren Chen <Zhiren.Chen@mediatek.com>
> > > > >
> > > > > The platform_pci_choose_state function and other low-level
> > > platform
> > > > > interfaces used by PCI power management processing did not
> take
> > > into
> > > > > account non-ACPI-supported platforms. This shortcoming can
> result
> > > in
> > > > > limitations and issues.
> > > > >
> > > > > For example, in embedded systems like smartphones, a PCI
> device
> > > can be
> > > > > shared by multiple processors for different purposes. The PCI
> > > device and
> > > > > some of the processors are controlled by Linux, while the
> rest of
> > > the
> > > > > processors runs its own operating system.
> > > > > When Linux initiates system-level sleep, if it does not
> consider
> > > the
> > > > > working state of the shared PCI device and forcefully sets
> the
> > > PCI device
> > > > > state to D3, it will affect the functionality of other
> processors
> > > that
> > > > > are currently using the PCI device.
> > > > >
> > > > > To address this problem, an interface should be created for
> PCI
> > > devices
> > > > > that don't support ACPI to enable accurate reporting of the
> power
> > > state
> > > > > during the PCI PM handling process.
> > > > >
> > > > > Signed-off-by: Zhiren Chen <Zhiren.Chen@mediatek.com>
> > >
> > > Something like the pci_platform_pm_ops introduced here had been
> there
> > > for several years and the only users of it known to me were ACPI
> and
> > > Intel MID, which is why it was dropped.
> > >
> > > I would like to see the platform code using these new callbacks
> in
> > > the
> > > first place.
> > >
> > I think that more and more embedded products will use PCI devices
> to
> > achieve higher performance for data transfer, and these products
> may
> > not necessarily support ACPI.
> >
> > When developing the Mediatek T8xx modem chip driver, I found that
> there
> > was no good way for T8xx to skip D3 setting in certain PM
> scenarios.
>
> Well, is there any code that you are planning to add to the mainline
> Linux kernel that is going to use the proposed interface?
>
yes, Mediatek T8xx modem chip driver without PM code is under review
now.
refs:
https://patchwork.kernel.org/project/netdevbpf/cover/20230317080942.183514-1-yanchao.yang@mediatek.com/
> If not, the interface itself will not be useful in the mainline Linux
> kernel.
I will submit this patch based on T8xx after T8xx driver is added to
mainline. Thank you for your comment.
Best Regards,
Zhiren
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2023-06-14 2:41 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <20230609023038.61388-1-zhiren.chen@mediatek.com>
2023-06-09 11:49 ` [PATCH] PCI:PM: Support platforms that do not implement ACPI Bjorn Helgaas
2023-06-09 17:58 ` Rafael J. Wysocki
2023-06-13 7:57 ` Zhiren Chen (陈志仁)
2023-06-13 14:30 ` Rafael J. Wysocki
2023-06-14 2:41 ` Zhiren Chen (陈志仁)
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).