PCI: Allow D3cold for PCIe hierarchies
Subsystems (4)
v2 cover PCI: Allow D3cold for PCIe hierarchies
Hi all,
This patch series aims to allow PCIe hierarchies enter D3cold both during system sleep and runtime, including PCIe hotplug ports.
The motivation of this series are recent systems such as Lenovo Thinkpad X1 Carbon 6th gen and upcoming Dell laptops where the Thunderbolt controller is always present in the system (pretty much like what Apple systems have been doing for years). Because it is always present it consumes energy so it is important to power it off when idle.
The PCIe root port hosting the Thunderbolt controller and all the connected external PCIe devices includes a standard ACPI power resource and related methods _PR3, _PR0, _PRW and _DSW that take care of the actual transition to D3cold and back to D0.
With current kernels entering system sleep states leaves the root port into D3hot which means the ACPI power resource is still on. This is bad when system enters suspend-to-idle (all these systems are s2idle only) because the BIOS is not involved and the devices are effectively left on, consuming battery unnecessarily.
Patches 1-4 add D3cold support for system sleep and the subsequent patches do the same for runtime PM.
In case someone wants to try out runtime PM, the xHCI controller that is part of the PCIe switch below the root port needs to have runtime PM "unblocked" manually (this will be automatic in future).
For Thinkpad:
# echo auto > /sys/bus/pci/devices/0000:3b:00.0/power/control
For Dell:
# echo auto > /sys/bus/pci/devices/0000:39:00.0/power/control
Note if the root port enters D3cold, running things like 'lspci' brings it back t0 D0 (because PCI config space is not accessible in D3cold) so if one wants to check out the power state without accidentally bringing the device back into D0 needs to do that indirectly. If the root port is 1d.0 power state can be checked:
# cat /sys/bus/pci/devices/0000:00:1d.0/power/runtime_status suspended # cat /sys/bus/pci/devices/0000:00:1d.0/firmware_node/real_power_state D3cold
Even if this again revolves around Thunderbolt I think these apply to any PCIe system supporting D3cold.
These patches apply on top of pci.git pci/hotplug. The previus version of the series can be viewed here:
https://www.spinics.net/lists/linux-acpi/msg83841.html
Changes from v1:
* Updated comment in patch [1/10] according what Rafael suggested
* Added empty line in patch [2/10]
* Check for !pciehp_poll_mode in [4/10]
* Use DPM_FLAG_NEVER_SKIP instead of DPM_FLAG_SMART_PREPARE which simplifies patch [5/10] signicantly
* Use const in acpi_data_get_property() and change int -> unsigned int
in patch [9/10]
* Added tags
Mika Westerberg (10):
PCI: Do not skip power managed bridges in pci_enable_wake()
PCI / ACPI: Enable wake automatically for power managed bridges
PCI: pciehp: Disable hotplug interrupt during suspend
PCI: pciehp: Do not handle events if interrupts are masked
PCI: portdrv: Resume upon exit from system suspend if left runtime suspended
PCI: portdrv: Add runtime PM hooks for port service drivers
PCI: pciehp: Implement runtime PM callbacks
PCI/PME: Implement runtime PM callbacks
ACPI / property: Allow multiple property compatible _DSD entries PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
drivers/acpi/property.c | 97 ++++++++++++++++++++++--------- drivers/acpi/x86/apple.c | 2 +- drivers/gpio/gpiolib-acpi.c | 2 +- drivers/pci/hotplug/pciehp.h | 2 + drivers/pci/hotplug/pciehp_core.c | 37 ++++++++++++ drivers/pci/hotplug/pciehp_hpc.c | 16 ++++- drivers/pci/pci-acpi.c | 57 +++++++++++++++++- drivers/pci/pci.c | 18 +++++- drivers/pci/pci.h | 3 + drivers/pci/pcie/pme.c | 27 +++++++++ drivers/pci/pcie/portdrv.h | 4 ++ drivers/pci/pcie/portdrv_core.c | 20 +++++++ drivers/pci/pcie/portdrv_pci.c | 14 ++--- include/acpi/acpi_bus.h | 8 ++- include/linux/acpi.h | 9 +++ 15 files changed, 273 insertions(+), 43 deletions(-)
Signature · 1 line
2.18.0
PATCH v2 2/10 PCI / ACPI: Enable wake automatically for power managed bridges
We enable power management automatically for bridges where pci_bridge_d3_possible() returns true. However, these bridges may have ACPI methods such as _DSW that need to be called before D3 entry. For example in Lenovo Thinkpad X1 Carbon 6th _DSW method is used to prepare D3cold for the PCIe root port hosting Thunderbolt chain. Because wake is not enabled _DSW method is never called and the port does not enter D3cold properly consuming more power than necessary.
Users can work this around by writing "enabled" to "wakeup" sysfs file under the device in question but that is not something an ordinary user is expected to do.
Since we already automatically enable power management for PCIe ports with ->bridge_d3 set extend that to enable wake for them as well, assuming the port has any ACPI wakeup related objects implemented in the namespace (adev->wakeup.flags.valid is true). This ensures the necessary ACPI methods get called at appropriate times and allows the root port in Thinkpad X1 Carbon 6th to go into D3cold.
--- drivers/pci/pci-acpi.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-)
drivers/pci/pci-acpi.c +15 -1
@@ -762,19 +762,33 @@ static void pci_acpi_setup(struct device *dev)return ;device_set_wakeup_capable (dev, true); +/* +* For bridges that can do D3 we enable wake automatically (as +* we do for the power management itself in that case). The +* reason is that the bridge may have additional methods such as +* _DSW that need to be called. +*/ +if (pci_dev->bridge_d3) +device_wakeup_enable (dev); +acpi_pci_wakeup (pci_dev, false); }static void pci_acpi_cleanup (struct device *dev) {struct acpi_device *adev =ACPI_COMPANION (dev); +struct pci_dev *pci_dev =to_pci_dev (dev);if (!adev)return ;pci_acpi_remove_pm_notifier (adev); -if (adev->wakeup.flags.valid) +if (adev->wakeup.flags.valid) { +if (pci_dev->bridge_d3) +device_wakeup_disable (dev); +device_set_wakeup_capable (dev, false); + } }static bool pci_acpi_bus_match (struct device *dev)
Signature · 1 line
2.18.0
PATCH v2 3/10 PCI: pciehp: Disable hotplug interrupt during suspend
In reply to Mika WesterbergWhen PCIe hotplug port is transitioned into D3hot link to the downstream component will go down. If hotplug interrupt generation is enabled when that happens it will trigger immediately waking up the system and bringing the link back up.
To prevent this disable hotplug interrupt generation when system suspend is entered. This does not prevent wakeup from low power states according to PCIe 4.0 spec section 6.7.3.4:
Software enables a hot-plug event to generate a wakeup event by enabling software notification of the event as described in Section 6.7.3.1. Note that in order for software to disable interrupt generation while keeping wakeup generation enabled, the Hot-Plug Interrupt Enable bit must be cleared.
So as long as we have set the slot event mask accordingly wakeup should work even if slot interrupt is disabled. The port should trigger wake and then send PME to the root port when the PCIe hierarchy is brought back up.
Limit this to systems using native PME mechanism to make sure older Apple systems depending on commit e3354628c376 ("PCI: pciehp: Support interrupts sent from D3hot") still continue working.
--- drivers/pci/hotplug/pciehp.h | 2 ++ drivers/pci/hotplug/pciehp_core.c | 18 ++++++++++++++++++ drivers/pci/hotplug/pciehp_hpc.c | 10 ++++++++++ 3 files changed, 30 insertions(+)
drivers/pci/hotplug/pciehp.h +2 -0
@@ -184,6 +184,8 @@ struct controller *pcie_init(struct pcie_device *dev);int pcie_init_notification (struct controller *ctrl);void pcie_shutdown_notification (struct controller *ctrl);void pcie_clear_hotplug_events (struct controller *ctrl); +void pcie_enable_interrupt (struct controller *ctrl); +void pcie_disable_interrupt (struct controller *ctrl);int pciehp_power_on_slot (struct slot *slot);void pciehp_power_off_slot (struct slot *slot);void pciehp_get_power_status (struct slot *slot,u8 *status);
drivers/pci/hotplug/pciehp_core.c +18 -0
@@ -265,8 +265,23 @@ static void pciehp_remove(struct pcie_device *dev) }#ifdef CONFIG_PM +static bool pme_is_native (struct pcie_device *dev) +{ +const struct pci_host_bridge *host; + + host =pci_find_host_bridge (dev->port->bus); +return pcie_ports_native || host->native_pme; +} +static int pciehp_suspend (struct pcie_device *dev) { +/* +* Disable hotplug interrupt so that it does not trigger +* immediately when the downstream link goes down. +*/ +if (pme_is_native (dev)) +pcie_disable_interrupt (get_service_data (dev)); +return 0 ; }
@@ -290,6 +305,9 @@ static int pciehp_resume(struct pcie_device *dev) {struct controller *ctrl =get_service_data (dev); +if (pme_is_native (dev)) +pcie_enable_interrupt (ctrl); +pciehp_check_presence (ctrl);return 0 ;
drivers/pci/hotplug/pciehp_hpc.c +10 -0
@@ -748,6 +748,16 @@ void pcie_clear_hotplug_events(struct controller *ctrl) PCI_EXP_SLTSTA_PDC | PCI_EXP_SLTSTA_DLLSC); } +void pcie_enable_interrupt (struct controller *ctrl) +{ +pcie_write_cmd (ctrl,PCI_EXP_SLTCTL_HPIE ,PCI_EXP_SLTCTL_HPIE ); +} + +void pcie_disable_interrupt (struct controller *ctrl) +{ +pcie_write_cmd (ctrl,0 ,PCI_EXP_SLTCTL_HPIE ); +} +/* * pciehp has a 1:1 bus:slot relationship so we ultimately want a secondary * bus reset of the bridge, but at the same time we want to ensure that it is
Signature · 1 line
2.18.0
PATCH v2 4/10 PCI: pciehp: Do not handle events if interrupts are masked
In reply to Mika WesterbergPCIe native hotplug shares MSI vector with native PME so the interrupt handler might get called even the hotplug interrupt is masked. In that case we should not handle any events because the interrupt was not meant for us. Modify the PCIe hotplug interrupt handler to check this accordingly and bail out if it finds out that the interrupt was not about hotplug.
--- drivers/pci/hotplug/pciehp_hpc.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/pci/hotplug/pciehp_hpc.c +4 -2
@@ -533,9 +533,11 @@ static irqreturn_t pciehp_isr(int irq, void *dev_id)u16 status, events;/* -* Interrupts only occur in D3hot or shallower (PCIe r4.0, sec 6.7.3.4). +* Interrupts only occur in D3hot or shallower and only if enabled +* in the Slot Control register (PCIe r4.0, sec 6.7.3.4). */ -if (pdev->current_state == PCI_D3cold) +if (pdev->current_state == PCI_D3cold || + (!(ctrl->slot_ctrl &PCI_EXP_SLTCTL_HPIE ) && !pciehp_poll_mode))return IRQ_NONE ;/*
Signature · 1 line
2.18.0
PATCH v2 5/10 PCI: portdrv: Resume upon exit from system suspend if left runtime suspended
In reply to Mika WesterbergCurrently we try to keep PCIe ports runtime suspended over system suspend if possible. This mostly happens when entering suspend-to-idle because there is no need to re-configure wake settings.
This causes problems if the parent port goes into D3cold and it gets resumed upon exit from system suspend. This may happen for example if the port is part of PCIe switch and the same switch is connected to a PCIe endpoint that needs to be resumed. The way exit from D3cold works according PCIe 4.0 spec 5.3.1.4.2 is that power is restored and cold reset is signaled. After this the device is in D0unitialized state keeping PME context if it supports wake from D3cold.
The problem occurs when a PCIe hotplug port is left suspended and the parent port goes into D3cold and back to D0, the port keeps its PME context but since everything else is reset back to defaults (D0unitialized) it is not set to detect hotplug events anymore.
For this reason change the PCIe portdrv power management logic so that it is fine to keep the port runtime suspended over system suspend but it needs to be resumed upon exit to make sure it gets properly re-initialized.
--- drivers/pci/pcie/portdrv_pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/pci/pcie/portdrv_pci.c +2 -2
@@ -109,8 +109,8 @@ static int pcie_portdrv_probe(struct pci_dev *dev,pci_save_state (dev); -dev_pm_set_driver_flags (&dev->dev,DPM_FLAG_SMART_SUSPEND | -DPM_FLAG_LEAVE_SUSPENDED ); +dev_pm_set_driver_flags (&dev->dev,DPM_FLAG_NEVER_SKIP | +DPM_FLAG_SMART_SUSPEND );if (pci_bridge_d3_possible (dev)) {/*
Signature · 1 line
2.18.0
Re: [PATCH v2 05/10] PCI: portdrv: Resume upon exit from system suspend if left runtime suspended
Quoted 32 lines · CEST Mika Westerberg
Currently we try to keep PCIe ports runtime suspended over system suspend if possible. This mostly happens when entering suspend-to-idle because there is no need to re-configure wake settings.
This causes problems if the parent port goes into D3cold and it gets resumed upon exit from system suspend. This may happen for example if the port is part of PCIe switch and the same switch is connected to a PCIe endpoint that needs to be resumed. The way exit from D3cold works according PCIe 4.0 spec 5.3.1.4.2 is that power is restored and cold reset is signaled. After this the device is in D0unitialized state keeping PME context if it supports wake from D3cold.
The problem occurs when a PCIe hotplug port is left suspended and the parent port goes into D3cold and back to D0, the port keeps its PME context but since everything else is reset back to defaults (D0unitialized) it is not set to detect hotplug events anymore.
For this reason change the PCIe portdrv power management logic so that it is fine to keep the port runtime suspended over system suspend but it needs to be resumed upon exit to make sure it gets properly re-initialized.
Signed-off-by Mika Westerberg <[email protected]>--- drivers/pci/pcie/portdrv_pci.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/pci/pcie/portdrv_pci.c +0
-0
drivers/pci/pcie/portdrv_pci.c static int pcie_portdrv_probe(struct pci_dev *dev, · line 116
Quoted 10 lines
drivers/pci/pcie/portdrv_pci.c +2
-2@@ -109,8 +109,8 @@ static int pcie_portdrv_probe(struct pci_dev *dev,pci_save_state (dev); -dev_pm_set_driver_flags (&dev->dev,DPM_FLAG_SMART_SUSPEND | -DPM_FLAG_LEAVE_SUSPENDED ); +dev_pm_set_driver_flags (&dev->dev,DPM_FLAG_NEVER_SKIP | +DPM_FLAG_SMART_SUSPEND );if (pci_bridge_d3_possible (dev)) {/*
PATCH v2 6/10 PCI: portdrv: Add runtime PM hooks for port service drivers
In reply to Mika WesterbergWhen PCIe port is runtime suspended/resumed some extra steps might be needed to be executed from the port service driver side. For instance we may need to disable PCIe hotplug interrupt to prevent it from triggering immediately when PCIe link to the downstream component goes down.
To make the above possible add optional ->runtime_suspend() and ->runtime_resume() callbacks to struct pcie_port_service_driver and call them for each port service in runtime suspend/resume callbacks of portdrv.
--- drivers/pci/pcie/portdrv.h | 4 ++++ drivers/pci/pcie/portdrv_core.c | 20 ++++++++++++++++++++ drivers/pci/pcie/portdrv_pci.c | 10 ++++------ 3 files changed, 28 insertions(+), 6 deletions(-)
drivers/pci/pcie/portdrv.h +4 -0
@@ -52,6 +52,8 @@ struct pcie_port_service_driver {int (*suspend) (struct pcie_device *dev);int (*resume_noirq) (struct pcie_device *dev);int (*resume) (struct pcie_device *dev); +int (*runtime_suspend) (struct pcie_device *dev); +int (*runtime_resume) (struct pcie_device *dev);/* Device driver may resume normal operations */ void (*error_resume)(struct pci_dev *dev);
@@ -85,6 +87,8 @@ int pcie_port_device_register(struct pci_dev *dev);int pcie_port_device_suspend (struct device *dev);int pcie_port_device_resume_noirq (struct device *dev);int pcie_port_device_resume (struct device *dev); +int pcie_port_device_runtime_suspend (struct device *dev); +int pcie_port_device_runtime_resume (struct device *dev);#endif void pcie_port_device_remove (struct pci_dev *dev);int __must_check pcie_port_bus_register (void );
drivers/pci/pcie/portdrv_core.c +20 -0
@@ -395,6 +395,26 @@ int pcie_port_device_resume(struct device *dev)size_t off = offsetof(struct pcie_port_service_driver , resume); return device_for_each_child(dev, &off, pm_iter); } + +/** +* pcie_port_device_runtime_suspend - runtime suspend port services +* @dev: PCI Express port to handle +*/ +int pcie_port_device_runtime_suspend (struct device *dev) +{ +size_t off = offsetof(struct pcie_port_service_driver , runtime_suspend); +return device_for_each_child (dev, &off, pm_iter); +} + +/** +* pcie_port_device_runtime_resume - runtime resume port services +* @dev: PCI Express port to handle +*/ +int pcie_port_device_runtime_resume (struct device *dev) +{ +size_t off = offsetof(struct pcie_port_service_driver , runtime_resume); +return device_for_each_child (dev, &off, pm_iter); +}#endif /* PM */ static int remove_iter (struct device *dev,void *data)
drivers/pci/pcie/portdrv_pci.c +4 -6
@@ -45,12 +45,10 @@ __setup("pcie_ports=", pcie_port_setup);#ifdef CONFIG_PM static int pcie_port_runtime_suspend (struct device *dev) { -return to_pci_dev (dev)->bridge_d3 ?0 : -EBUSY ; -} +if (!to_pci_dev (dev)->bridge_d3) +return -EBUSY ; -static int pcie_port_runtime_resume (struct device *dev) -{ -return 0 ; +return pcie_port_device_runtime_suspend (dev); }static int pcie_port_runtime_idle (struct device *dev)
@@ -73,7 +71,7 @@ static const struct dev_pm_ops pcie_portdrv_pm_ops = { .restore_noirq = pcie_port_device_resume_noirq, .restore = pcie_port_device_resume, .runtime_suspend = pcie_port_runtime_suspend, - .runtime_resume = pcie_port_runtime_resume, + .runtime_resume = pcie_port_device_runtime_resume, .runtime_idle = pcie_port_runtime_idle, };
Signature · 1 line
2.18.0
PATCH v2 7/10 PCI: pciehp: Implement runtime PM callbacks
In reply to Mika WesterbergBasically we need to do the same thing when runtime suspending than with system sleep so re-use those operations here. This makes sure hotplug interrupt does not trigger immediately when the link goes down.
--- drivers/pci/hotplug/pciehp_core.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
drivers/pci/hotplug/pciehp_core.c +19 -0
@@ -312,6 +312,23 @@ static int pciehp_resume(struct pcie_device *dev)return 0 ; } + +static int pciehp_runtime_resume (struct pcie_device *dev) +{ +struct controller *ctrl =get_service_data (dev); +struct slot *slot = ctrl->slot; + +/* pci_restore_state() just wrote to the Slot Control register */ + ctrl->cmd_started = jiffies; + ctrl->cmd_busy = true; + +/* clear spurious events from rediscovery of inserted card */ +if ((slot->state ==ON_STATE || slot->state ==BLINKINGOFF_STATE ) && +pme_is_native (dev)) +pcie_clear_hotplug_events (ctrl); + +return pciehp_resume (dev); +}#endif /* PM */ static struct pcie_port_service_driver hpdriver_portdrv = {
@@ -326,6 +343,8 @@ static struct pcie_port_service_driver hpdriver_portdrv = { .suspend = pciehp_suspend, .resume_noirq = pciehp_resume_noirq, .resume = pciehp_resume, + .runtime_suspend = pciehp_suspend, + .runtime_resume = pciehp_runtime_resume,#endif /* PM */ };
Signature · 1 line
2.18.0
PATCH v2 8/10 PCI/PME: Implement runtime PM callbacks
In reply to Mika WesterbergBasically we need to do the same steps than what we do when system sleep is entered and disable PME interrupt when the root port is runtime suspended. This prevents spurious wakeups immediately when the port is transitioned into D3cold.
--- drivers/pci/pcie/pme.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
drivers/pci/pcie/pme.c +27 -0
@@ -432,6 +432,31 @@ static void pcie_pme_remove(struct pcie_device *srv)kfree (get_service_data (srv)); } +static int pcie_pme_runtime_suspend (struct pcie_device *srv) +{ +struct pcie_pme_service_data *data =get_service_data (srv); + +spin_lock_irq (&data->lock); +pcie_pme_interrupt_enable (srv->port, false); +pcie_clear_root_pme_status (srv->port); + data->noirq = true; +spin_unlock_irq (&data->lock); + +return 0 ; +} + +static int pcie_pme_runtime_resume (struct pcie_device *srv) +{ +struct pcie_pme_service_data *data =get_service_data (srv); + +spin_lock_irq (&data->lock); +pcie_pme_interrupt_enable (srv->port, true); + data->noirq = false; +spin_unlock_irq (&data->lock); + +return 0 ; +} +static struct pcie_port_service_driver pcie_pme_driver = { .name ="pcie_pme" , .port_type =PCI_EXP_TYPE_ROOT_PORT ,
@@ -439,6 +464,8 @@ static struct pcie_port_service_driver pcie_pme_driver = { .probe = pcie_pme_probe, .suspend = pcie_pme_suspend, + .runtime_suspend = pcie_pme_runtime_suspend, + .runtime_resume = pcie_pme_runtime_resume, .resume = pcie_pme_resume, .remove = pcie_pme_remove, };
Signature · 1 line
2.18.0
PATCH v2 9/10 ACPI / property: Allow multiple property compatible _DSD entries
In reply to Mika WesterbergIt is possible to have _DSD entries where the data is compatible with device properties format but are using different GUID for various reasons. In addition to that there can be many such _DSD entries for a single device such as for PCIe root port used to host a Thunderbolt hierarchy:
Scope (\_SB.PCI0.RP21)
{
Name (_DSD, Package () {
ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"),
Package () {
Package () {"HotPlugSupportInD3", 1}
},
ToUUID ("efcc06cc-73ac-4bc3-bff0-76143807c389"),
Package () {
Package () {"ExternalFacingPort", 1},
Package () {"UID", 0 }
}
})
}For more information about these new _DSD entries can be found in:
https://docs.microsoft.com/en-us/windows-hardware/drivers/pci/dsd-for-pcie-root-ports
In order to make these available for drivers via unified device property APIs modify ACPI property core so that it supports multiple _DSD entries organized in a linked list. We also store GUID of each _DSD entry in struct acpi_device_properties in case there is need to differerentiate between entries. The supported GUIDs are then listed in prp_guids array.
--- drivers/acpi/property.c | 94 +++++++++++++++++++++++++++---------- drivers/acpi/x86/apple.c | 2 +- drivers/gpio/gpiolib-acpi.c | 2 +- include/acpi/acpi_bus.h | 8 +++- include/linux/acpi.h | 9 ++++ 5 files changed, 86 insertions(+), 29 deletions(-)
drivers/acpi/property.c +68 -26
@@ -24,11 +24,12 @@ static int acpi_data_get_property_array(const struct acpi_device_data *data,acpi_object_type type, const union acpi_object **obj); -/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ -static const guid_t prp_guid = +static const guid_t prp_guids[] = { +/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c , -0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ); -/* ACPI _DSD data subnodes GUID: dbb8e3e6-5886-4ba6-8795-1319f52a966b */ +0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +}; +static const guid_t ads_guid =GUID_INIT (0xdbb8e3e6 ,0x5886 ,0x4ba6 ,0x87 ,0x95 ,0x13 ,0x19 ,0xf5 ,0x2a ,0x96 ,0x6b );
@@ -56,6 +57,7 @@ static bool acpi_nondev_subnode_extract(const union acpi_object *desc, dn->name = link->package.elements[0 ].string.pointer; dn->fwnode.ops = &acpi_data_fwnode_ops; dn->parent = parent; +INIT_LIST_HEAD (&dn->data.properties);INIT_LIST_HEAD (&dn->data.subnodes); result =acpi_extract_properties (desc, &dn->data);
@@ -288,6 +290,35 @@ static void acpi_init_of_compatible(struct acpi_device *adev) adev->flags.of_compatible_ok =1 ; } +static bool acpi_is_property_guid (const guid_t *guid) +{ +int i; + +for (i =0 ; i <ARRAY_SIZE (prp_guids); i++) { +if (guid_equal (guid, &prp_guids[i])) +return true; + } + +return false; +} + +struct acpi_device_properties * +acpi_data_add_props (struct acpi_device_data *data,const guid_t *guid, +const union acpi_object *properties) +{ +struct acpi_device_properties *props; + + props =kzalloc (sizeof (*props),GFP_KERNEL ); +if (props) { +INIT_LIST_HEAD (&props->list); + props->guid = guid; + props->properties = properties; +list_add_tail (&props->list, &data->properties); + } + +return props; +} +static bool acpi_extract_properties (const union acpi_object *desc,struct acpi_device_data *data) {
@@ -312,7 +343,7 @@ static bool acpi_extract_properties(const union acpi_object *desc, properties->type != ACPI_TYPE_PACKAGE)break ; -if (!guid_equal ((guid_t *)guid->buffer.pointer, &prp_guid)) +if (!acpi_is_property_guid ((guid_t *)guid->buffer.pointer))continue ;/*
@@ -320,13 +351,13 @@ static bool acpi_extract_properties(const union acpi_object *desc,* package immediately following it. */ if (!acpi_properties_format_valid (properties)) -break ; +continue ; - data->properties = properties; -return true; +acpi_data_add_props (data, (const guid_t *)guid->buffer.pointer, + properties); } -return false; +return !list_empty (&data->properties); }void acpi_init_properties (struct acpi_device *adev)
@@ -336,6 +367,7 @@ void acpi_init_properties(struct acpi_device *adev)acpi_status status;bool acpi_of = false; +INIT_LIST_HEAD (&adev->data.properties);INIT_LIST_HEAD (&adev->data.subnodes);if (!adev->handle)
@@ -398,11 +430,16 @@ static void acpi_destroy_nondev_subnodes(struct list_head *list)void acpi_free_properties (struct acpi_device *adev) { +struct acpi_device_properties *props, *tmp; +acpi_destroy_nondev_subnodes (&adev->data.subnodes);ACPI_FREE ((void *)adev->data.pointer); adev->data.of_compatible =NULL ; adev->data.pointer =NULL ; - adev->data.properties =NULL ; +list_for_each_entry_safe (props, tmp, &adev->data.properties, list) { +list_del (&props->list); +kfree (props); + } }/**
@@ -427,32 +464,37 @@ static int acpi_data_get_property(const struct acpi_device_data *data, const char *name, acpi_object_type type, const union acpi_object **obj) { - const union acpi_object *properties; -int i; + const struct acpi_device_properties *props;if (!data || !name)return -EINVAL ; -if (!data->pointer || !data->properties) +if (!data->pointer ||list_empty (&data->properties))return -EINVAL ; - properties = data->properties; -for (i =0 ; i < properties->package.count; i++) { -const union acpi_object *propname, *propvalue; -const union acpi_object *property; +list_for_each_entry (props, &data->properties, list) { +const union acpi_object *properties; +unsigned int i; - property = &properties->package.elements[i]; + properties = props->properties; +for (i =0 ; i < properties->package.count; i++) { +const union acpi_object *propname, *propvalue; +const union acpi_object *property; - propname = &property->package.elements[0 ]; - propvalue = &property->package.elements[1 ]; + property = &properties->package.elements[i]; -if (!strcmp (name, propname->string.pointer)) { -if (type !=ACPI_TYPE_ANY && propvalue->type != type) -return -EPROTO ; -if (obj) - *obj = propvalue; + propname = &property->package.elements[0 ]; + propvalue = &property->package.elements[1 ]; -return 0 ; +if (!strcmp (name, propname->string.pointer)) { +if (type !=ACPI_TYPE_ANY && + propvalue->type != type) +return -EPROTO ; +if (obj) + *obj = propvalue; + +return 0 ; + } } }return -EINVAL ;
drivers/acpi/x86/apple.c +1 -1
@@ -132,8 +132,8 @@ void acpi_extract_apple_properties(struct acpi_device *adev) }WARN_ON (free_space != (void *)newprops + newsize); - adev->data.properties = newprops; adev->data.pointer = newprops; +acpi_data_add_props (&adev->data, &apple_prp_guid, newprops); out_free:ACPI_FREE (props);
drivers/gpio/gpiolib-acpi.c +1 -1
@@ -1192,7 +1192,7 @@ int acpi_gpio_count(struct device *dev, const char *con_id)bool acpi_can_fallback_to_crs (struct acpi_device *adev,const char *con_id) {/* Never allow fallback if the device has properties */ -if (adev->data.properties || adev->driver_gpios) +if (acpi_dev_has_props (adev) || adev->driver_gpios)return false;return con_id ==NULL ;
include/acpi/acpi_bus.h +7 -1
@@ -346,10 +346,16 @@ struct acpi_device_physical_node { bool put_online:1; }; +struct acpi_device_properties { +const guid_t *guid; +const union acpi_object *properties; +struct list_head list; +}; +/* ACPI Device Specific Data (_DSD) */ struct acpi_device_data {const union acpi_object *pointer; -const union acpi_object *properties; +struct list_head properties;const union acpi_object *of_compatible;struct list_head subnodes; };
include/linux/acpi.h +9 -0
@@ -1074,6 +1074,15 @@ static inline int acpi_node_get_property_reference( NR_FWNODE_REFERENCE_ARGS, args); } +static inline bool acpi_dev_has_props (const struct acpi_device *adev) +{ +return !list_empty (&adev->data.properties); +} + +struct acpi_device_properties * +acpi_data_add_props (struct acpi_device_data *data,const guid_t *guid, +const union acpi_object *properties); +int acpi_node_prop_get (const struct fwnode_handle *fwnode,const char *propname,void **valptr); int acpi_dev_prop_read_single(struct acpi_device *adev,
Signature · 1 line
2.18.0
A message that is not in the archive (sent off-list or not yet received): [email protected]
Rafael J. Wysocki
Reviewed-byRe: [PATCH v2 01/10] PCI: Do not skip power managed bridges in pci_enable_wake()
Quoted 30 lines · CEST Mika Westerberg
Commit
baecc470d5fd("PCI / PM: Skip bridges inpci_enable_wake()") changedpci_enable_wake()so that all bridges are skipped when wakeup is enabled (or disabled) with the reasoning that bridges can only signal wakeup on behalf of their subordinate devices.However, there are bridges that can signal wakeup itself. For example PCIe downstream and root ports supporting hotplug may signal wakeup upon hotplug event.
For this reason change
pci_enable_wake()so that it skips all bridges except those that we power manage (->bridge_d3 is set). Those are the ones that can go into low power states and may need to signal wakeup.Signed-off-by Mika Westerberg <[email protected]>--- drivers/pci/pci.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)drivers/pci/pci.c +1
-2@@ -2133,10 +2133,13 @@ static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enableint ret =0 ;/* -* Bridges can only signal wakeup on behalf of subordinate devices, -* but that is set up elsewhere, so skip them. +* Bridges that are not power-manageable directly only signal
drivers/pci/pci.c static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enable · line 2145
Quoted 10 lines
drivers/pci/pci.c +5
-1+* wakeup on behalf of subordinate devices which is set up +* elsewhere, so skip them. However, bridges that are +* power-manageable may signal wakeup for themselves (for example, +* on a hotplug event) and they need to be covered here. */ -if (pci_has_subordinate (dev)) +if (!pci_power_manageable (dev))return 0 ;/* Don't do the same thing twice in a row for one device. */
Re: [PATCH v2 01/10] PCI: Do not skip power managed bridges in pci_enable_wake()
Quoted 39 lines · Rafael J. Wysocki
Quoted 38 lines · CEST Mika Westerberg
Commit
baecc470d5fd("PCI / PM: Skip bridges inpci_enable_wake()") changedpci_enable_wake()so that all bridges are skipped when wakeup is enabled (or disabled) with the reasoning that bridges can only signal wakeup on behalf of their subordinate devices.However, there are bridges that can signal wakeup itself. For example PCIe downstream and root ports supporting hotplug may signal wakeup upon hotplug event.
For this reason change
pci_enable_wake()so that it skips all bridges except those that we power manage (->bridge_d3 is set). Those are the ones that can go into low power states and may need to signal wakeup.Signed-off-by Mika Westerberg <[email protected]>--- drivers/pci/pci.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)drivers/pci/pci.c +6
-3@@ -2133,10 +2133,13 @@ static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enableint ret =0 ;/* -* Bridges can only signal wakeup on behalf of subordinate devices, -* but that is set up elsewhere, so skip them. +* Bridges that are not power-manageable directly only signal +* wakeup on behalf of subordinate devices which is set up +* elsewhere, so skip them. However, bridges that are +* power-manageable may signal wakeup for themselves (for example, +* on a hotplug event) and they need to be covered here. */ -if (pci_has_subordinate (dev)) +if (!pci_power_manageable (dev))return 0 ;
Quoted 4 lines
drivers/pci/pci.c static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enable · line 2145
Quoted 2 lines
drivers/pci/pci.c +0
-0/* Don't do the same thing twice in a row for one device. */ Reviewed-by Rafael J. Wysocki <[email protected]>
Thanks!
For some reason this patch never hit the mailing lists (others did). Since you kept the whole context people should still be able to review it. I can resend this patch as well.
Rafael J. Wysocki
maintainerRe: [PATCH v2 01/10] PCI: Do not skip power managed bridges in pci_enable_wake()
Quoted 46 lines · Mika Westerberg
Quoted 39 lines · Rafael J. Wysocki
Quoted 38 lines · CEST Mika Westerberg
Commit
baecc470d5fd("PCI / PM: Skip bridges inpci_enable_wake()") changedpci_enable_wake()so that all bridges are skipped when wakeup is enabled (or disabled) with the reasoning that bridges can only signal wakeup on behalf of their subordinate devices.However, there are bridges that can signal wakeup itself. For example PCIe downstream and root ports supporting hotplug may signal wakeup upon hotplug event.
For this reason change
pci_enable_wake()so that it skips all bridges except those that we power manage (->bridge_d3 is set). Those are the ones that can go into low power states and may need to signal wakeup.Signed-off-by Mika Westerberg <[email protected]>--- drivers/pci/pci.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)drivers/pci/pci.c +6
-3@@ -2133,10 +2133,13 @@ static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enableint ret =0 ;/* -* Bridges can only signal wakeup on behalf of subordinate devices, -* but that is set up elsewhere, so skip them. +* Bridges that are not power-manageable directly only signal +* wakeup on behalf of subordinate devices which is set up +* elsewhere, so skip them. However, bridges that are +* power-manageable may signal wakeup for themselves (for example, +* on a hotplug event) and they need to be covered here. */ -if (pci_has_subordinate (dev)) +if (!pci_power_manageable (dev))return 0 ;Quoted 4 lines
drivers/pci/pci.c static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enable · line 2145
Quoted 2 lines
drivers/pci/pci.c +0
-0/* Don't do the same thing twice in a row for one device. */ Reviewed-by Rafael J. Wysocki <[email protected]>
Quoted 5 lines
Thanks!
For some reason this patch never hit the mailing lists (others did). Since you kept the whole context people should still be able to review it. I can resend this patch as well.
You may need to resend for Patchwork to pick it up, though. Depends on Bjorn I suppose.
Re: [PATCH v2 01/10] PCI: Do not skip power managed bridges in pci_enable_wake()
Quoted 2 lines · Rafael J. Wysocki
You may need to resend for Patchwork to pick it up, though. Depends on Bjorn I suppose.
OK, I'll resend it just in case :)
PATCH RESEND v2 1/10 PCI: Do not skip power managed bridges in pci_enable_wake()
In reply to Rafael J. WysockiCommit baecc470d5fd ("PCI / PM: Skip bridges in pci_enable_wake()") changed pci_enable_wake() so that all bridges are skipped when wakeup is enabled (or disabled) with the reasoning that bridges can only signal wakeup on behalf of their subordinate devices.
However, there are bridges that can signal wakeup itself. For example PCIe downstream and root ports supporting hotplug may signal wakeup upon hotplug event.
For this reason change pci_enable_wake() so that it skips all bridges except those that we power manage (->bridge_d3 is set). Those are the ones that can go into low power states and may need to signal wakeup.
---
Resending this because for some reason it did not hit the mailing lists. Added tag from Rafael as well.
drivers/pci/pci.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/pci/pci.c +6 -3
@@ -2133,10 +2133,13 @@ static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enableint ret =0 ;/* -* Bridges can only signal wakeup on behalf of subordinate devices, -* but that is set up elsewhere, so skip them. +* Bridges that are not power-manageable directly only signal +* wakeup on behalf of subordinate devices which is set up +* elsewhere, so skip them. However, bridges that are +* power-manageable may signal wakeup for themselves (for example, +* on a hotplug event) and they need to be covered here. */ -if (pci_has_subordinate (dev)) +if (!pci_power_manageable (dev))return 0 ;/* Don't do the same thing twice in a row for one device. */
Signature · 1 line
2.18.0
Bjorn Helgaas
AppliedRe: [PATCH RESEND v2 01/10] PCI: Do not skip power managed bridges in pci_enable_wake()
Quoted 9 lines · Mika Westerberg
Commit
baecc470d5fd("PCI / PM: Skip bridges inpci_enable_wake()") changedpci_enable_wake()so that all bridges are skipped when wakeup is enabled (or disabled) with the reasoning that bridges can only signal wakeup on behalf of their subordinate devices.However, there are bridges that can signal wakeup itself. For example PCIe downstream and root ports supporting hotplug may signal wakeup upon hotplug event.
Quoted 6 lines
For this reason change
pci_enable_wake()so that it skips all bridges except those that we power manage (->bridge_d3 is set). Those are the ones that can go into low power states and may need to signal wakeup.Signed-off-by Mika Westerberg <[email protected]>Reviewed-by Rafael J. Wysocki <[email protected]>
Applied to pci/enumeration for v4.20, thanks!
Quoted 31 lines
---
Resending this because for some reason it did not hit the mailing lists. Added tag from Rafael as well.drivers/pci/pci.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)drivers/pci/pci.c +6
-3@@ -2133,10 +2133,13 @@ static int __pci_enable_wake(struct pci_dev *dev, pci_power_t state, bool enableint ret =0 ;/* -* Bridges can only signal wakeup on behalf of subordinate devices, -* but that is set up elsewhere, so skip them. +* Bridges that are not power-manageable directly only signal +* wakeup on behalf of subordinate devices which is set up +* elsewhere, so skip them. However, bridges that are +* power-manageable may signal wakeup for themselves (for example, +* on a hotplug event) and they need to be covered here. */ -if (pci_has_subordinate (dev)) +if (!pci_power_manageable (dev))return 0 ;/* Don't do the same thing twice in a row for one device. */ Signature · 1 line
2.18.0
PATCH v2 10/10 PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
In reply to Mika WesterbergIn order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI _DSD attached to the root port:
Name (_DSD, Package () {
ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"),
Package () {
Package () {"HotPlugSupportInD3", 1}
}
})This is also documented in:
Do the same in Linux by introducing new firmware PM callback (->bridge_d3()) and then implement it for ACPI based systems so that the above property is checked.
There is one catch, though. The initial pci_dev->bridge_d3 is set before the root port has ACPI companion bound (the device is not added to the PCI bus either) so we need to look up the ACPI companion manually in that case in acpi_pci_bridge_d3().
--- drivers/acpi/property.c | 3 +++ drivers/pci/pci-acpi.c | 41 +++++++++++++++++++++++++++++++++++++++++ drivers/pci/pci.c | 9 +++++++++ drivers/pci/pci.h | 3 +++ 4 files changed, 56 insertions(+)
drivers/acpi/property.c +3 -0
@@ -28,6 +28,9 @@ static const guid_t prp_guids[] = {/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c ,0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +/* Hotplug in D3 GUID: 6211e2c0-58a3-4af3-90e1-927a4e0c55a4 */ +GUID_INIT (0x6211e2c0 ,0x58a3 ,0x4af3 , + 0x90, 0xe1, 0x92, 0x7a, 0x4e, 0x0c, 0x55, 0xa4), }; static const guid_t ads_guid =
drivers/pci/pci-acpi.c +41 -0
53 lines not on this page · raw message
drivers/pci/pci.c +9 -0
@@ -791,6 +791,11 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)return pci_platform_pm ? pci_platform_pm->need_resume (dev) : false; } +static inline bool platform_pci_bridge_d3 (struct pci_dev *dev) +{ +return pci_platform_pm ? pci_platform_pm->bridge_d3 (dev) : false; +} +/** * pci_raw_set_power_state - Use PCI PM registers to set the power state of * given PCI device
@@ -2514,6 +2519,10 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)if (bridge->is_thunderbolt)return true; +/* Platform might know better if the bridge supports D3 */ +if (platform_pci_bridge_d3 (bridge)) +return true; +/* * Hotplug ports handled natively by the OS were not validated * by vendors for runtime D3 at least until 2018 because there
drivers/pci/pci.h +3 -0
@@ -39,6 +39,8 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);/** * struct pci_platform_pm_ops - Firmware PM callbacks * +* @bridge_d3: Does the bridge allow entering into D3 +* * @is_manageable: returns 'true' if given device is power manageable by the * platform firmware *
@@ -60,6 +62,7 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);* these callbacks are mandatory. */ struct pci_platform_pm_ops { +bool (*bridge_d3)(struct pci_dev *dev);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);
Signature · 1 line
2.18.0
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })
Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this _DSD, the root port is "capable of handling hot plug events while in D3 state".
What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
Quoted 96 lines
Do the same in Linux by introducing new firmware PM callback (->bridge_d3()) and then implement it for ACPI based systems so that the above property is checked.
There is one catch, though. The initial
pci_dev->bridge_d3is set before the root port has ACPI companion bound (the device is not added to the PCI bus either) so we need to look up the ACPI companion manually in that case inacpi_pci_bridge_d3().Signed-off-by Mika Westerberg <[email protected]>Reviewed-by Rafael J. Wysocki <[email protected]>--- drivers/acpi/property.c | 3 +++ drivers/pci/pci-acpi.c | 41 +++++++++++++++++++++++++++++++++++++++++ drivers/pci/pci.c | 9 +++++++++ drivers/pci/pci.h | 3 +++ 4 files changed, 56 insertions(+)drivers/acpi/property.c +3
-0@@ -28,6 +28,9 @@ static const guid_t prp_guids[] = {/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c ,0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +/* Hotplug in D3 GUID: 6211e2c0-58a3-4af3-90e1-927a4e0c55a4 */ +GUID_INIT (0x6211e2c0 ,0x58a3 ,0x4af3 , + 0x90, 0xe1, 0x92, 0x7a, 0x4e, 0x0c, 0x55, 0xa4), }; static const guid_t ads_guid =drivers/pci/pci-acpi.c +41
-0@@ -519,6 +519,46 @@ static pci_power_t acpi_pci_choose_state(struct pci_dev *pdev)return PCI_POWER_ERROR ; } +static struct acpi_device *acpi_pci_find_companion (struct device *dev); + +static bool acpi_pci_bridge_d3 (struct pci_dev *dev) +{ +const struct fwnode_handle *fwnode; +struct acpi_device *adev; +struct pci_dev *root; +u8 val; + +if (!dev->is_hotplug_bridge) +return false; + +/* +* Look for a special _DSD property for the root port and if it +* is set we know the hierarchy behind it supports D3 just fine. +*/ + root =pci_find_pcie_root_port (dev); +if (!root) +return false; + + adev =ACPI_COMPANION (&root->dev); +if (root == dev) { +/* +* It is possible that the ACPI companion is not yet bound +* for the root port so look it up manually here. +*/ +if (!adev && !pci_dev_is_added (root)) + adev =acpi_pci_find_companion (&root->dev); + } + +if (!adev) +return false; + + fwnode =acpi_fwnode_handle (adev); +if (fwnode_property_read_u8 (fwnode,"HotPlugSupportInD3" , &val)) +return false; + +return val ==1 ; +} +static bool acpi_pci_power_manageable (struct pci_dev *dev) {struct acpi_device *adev =ACPI_COMPANION (&dev->dev);@@ -635,6 +675,7 @@ static bool acpi_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops acpi_pci_platform_pm = { + .bridge_d3 = acpi_pci_bridge_d3, .is_manageable = acpi_pci_power_manageable, .set_state = acpi_pci_set_power_state, .get_state = acpi_pci_get_power_state,drivers/pci/pci.c +0
-0
drivers/pci/pci.c static inline bool platform_pci_need_resume(struct pci_dev *dev) · line 796
Quoted 6 lines
drivers/pci/pci.c +3
-0@@ -791,6 +791,11 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)return pci_platform_pm ? pci_platform_pm->need_resume (dev) : false; } +static inline bool platform_pci_bridge_d3 (struct pci_dev *dev) +{ +return pci_platform_pm ? pci_platform_pm->bridge_d3 (dev) : false;
This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling platform_pci_bridge_d3() on a MID platform and trying to call through a NULL pointer?
Shouldn't we do something like the patch attached below?
drivers/pci/pci.c bool pci_bridge_d3_possible(struct pci_dev *bridge) · line 2524
Quoted 12 lines
drivers/pci/pci.c +5
-0+} +/** * pci_raw_set_power_state - Use PCI PM registers to set the power state of * given PCI device @@ -2514,6 +2519,10 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)if (bridge->is_thunderbolt)return true; +/* Platform might know better if the bridge supports D3 */ +if (platform_pci_bridge_d3 (bridge)) +return true;
*All* devices trivially support D3. Obviously we're trying to learn something else here. What is it?
Quoted 18 lines
drivers/pci/pci.c +0
-0/* * Hotplug ports handled natively by the OS were not validated * by vendors for runtime D3 at least until 2018 because there drivers/pci/pci.h +2
-0@@ -39,6 +39,8 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);/** * struct pci_platform_pm_ops - Firmware PM callbacks * +* @bridge_d3: Does the bridge allow entering into D3 +* * @is_manageable: returns 'true' if given device is power manageable by the * platform firmware *
drivers/pci/pci.h int pci_bridge_secondary_bus_reset(struct pci_dev *dev); · line 68
Quoted 7 lines
drivers/pci/pci.h +1
-0@@ -60,6 +62,7 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);* these callbacks are mandatory. */ struct pci_platform_pm_ops { +bool (*bridge_d3)(struct pci_dev *dev);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);
drivers/pci/pci-mid.c +6 -0
@@ -16,6 +16,11 @@#include "pci.h" +static bool mid_pci_bridge_d3 (struct pci_dev *dev) +{ +return false; +} +static bool mid_pci_power_manageable (struct pci_dev *dev) {return true;
@@ -47,6 +52,7 @@ static bool mid_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops mid_pci_platform_pm = { + .bridge_d3 = mid_pci_bridge_d3, .is_manageable = mid_pci_power_manageable, .set_state = mid_pci_set_power_state, .get_state = mid_pci_get_power_state,
drivers/pci/pci.c +3 -2
@@ -820,8 +820,9 @@ static const struct pci_platform_pm_ops *pci_platform_pm;int pci_set_platform_pm (const struct pci_platform_pm_ops *ops) { -if (!ops->is_manageable || !ops->set_state || !ops->get_state || - !ops->choose_state || !ops->set_wakeup || !ops->need_resume) +if (!ops->bridge_d3 || !ops->is_manageable || !ops->set_state || + !ops->get_state || !ops->choose_state || !ops->set_wakeup || + !ops->need_resume)return -EINVAL ; pci_platform_pm = ops;return 0 ;
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 19 lines · Bjorn Helgaas
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this
_DSD, the root port is "capable of handling hot plug events while in D3 state".
Quoted 6 lines
What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
I think it simply gives the OS a hint that it can put PCIe hotplug capable port into D3 and expect it to wake up when device is detected.
Quoted 107 lines
Quoted 96 lines
Do the same in Linux by introducing new firmware PM callback (->bridge_d3()) and then implement it for ACPI based systems so that the above property is checked.
There is one catch, though. The initial
pci_dev->bridge_d3is set before the root port has ACPI companion bound (the device is not added to the PCI bus either) so we need to look up the ACPI companion manually in that case inacpi_pci_bridge_d3().Signed-off-by Mika Westerberg <[email protected]>Reviewed-by Rafael J. Wysocki <[email protected]>--- drivers/acpi/property.c | 3 +++ drivers/pci/pci-acpi.c | 41 +++++++++++++++++++++++++++++++++++++++++ drivers/pci/pci.c | 9 +++++++++ drivers/pci/pci.h | 3 +++ 4 files changed, 56 insertions(+)drivers/acpi/property.c +3
-0@@ -28,6 +28,9 @@ static const guid_t prp_guids[] = {/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c ,0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +/* Hotplug in D3 GUID: 6211e2c0-58a3-4af3-90e1-927a4e0c55a4 */ +GUID_INIT (0x6211e2c0 ,0x58a3 ,0x4af3 , + 0x90, 0xe1, 0x92, 0x7a, 0x4e, 0x0c, 0x55, 0xa4), }; static const guid_t ads_guid =drivers/pci/pci-acpi.c +41
-0@@ -519,6 +519,46 @@ static pci_power_t acpi_pci_choose_state(struct pci_dev *pdev)return PCI_POWER_ERROR ; } +static struct acpi_device *acpi_pci_find_companion (struct device *dev); + +static bool acpi_pci_bridge_d3 (struct pci_dev *dev) +{ +const struct fwnode_handle *fwnode; +struct acpi_device *adev; +struct pci_dev *root; +u8 val; + +if (!dev->is_hotplug_bridge) +return false; + +/* +* Look for a special _DSD property for the root port and if it +* is set we know the hierarchy behind it supports D3 just fine. +*/ + root =pci_find_pcie_root_port (dev); +if (!root) +return false; + + adev =ACPI_COMPANION (&root->dev); +if (root == dev) { +/* +* It is possible that the ACPI companion is not yet bound +* for the root port so look it up manually here. +*/ +if (!adev && !pci_dev_is_added (root)) + adev =acpi_pci_find_companion (&root->dev); + } + +if (!adev) +return false; + + fwnode =acpi_fwnode_handle (adev); +if (fwnode_property_read_u8 (fwnode,"HotPlugSupportInD3" , &val)) +return false; + +return val ==1 ; +} +static bool acpi_pci_power_manageable (struct pci_dev *dev) {struct acpi_device *adev =ACPI_COMPANION (&dev->dev);@@ -635,6 +675,7 @@ static bool acpi_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops acpi_pci_platform_pm = { + .bridge_d3 = acpi_pci_bridge_d3, .is_manageable = acpi_pci_power_manageable, .set_state = acpi_pci_set_power_state, .get_state = acpi_pci_get_power_state,drivers/pci/pci.c +0
-0drivers/pci/pci.c static inline bool platform_pci_need_resume(struct pci_dev *dev) · line 796
Quoted 6 lines
drivers/pci/pci.c +3
-0@@ -791,6 +791,11 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)return pci_platform_pm ? pci_platform_pm->need_resume (dev) : false; } +static inline bool platform_pci_bridge_d3 (struct pci_dev *dev) +{ +return pci_platform_pm ? pci_platform_pm->bridge_d3 (dev) : false;This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling
platform_pci_bridge_d3()on a MID platform and trying to call through a NULL pointer?
Quoted 1 line
Shouldn't we do something like the patch attached below?
IIRC MID devices in general don't have PCIe ports (so we never enable PM for them). Is this a real problem that crashes peoples kernels on MID systems? Then yes I think the patch makes sense to have.
I also remember testing some other stuff on one MID system (Edison) quite recently and did not see any issues.
Quoted 9 lines
Quoted 9 lines
drivers/pci/pci.c +2
-0+} +/** * pci_raw_set_power_state - Use PCI PM registers to set the power state of * given PCI device @@ -2514,6 +2519,10 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)if (bridge->is_thunderbolt)return true;
Quoted 6 lines
drivers/pci/pci.c bool pci_bridge_d3_possible(struct pci_dev *bridge) · line 2524
Quoted 3 lines
drivers/pci/pci.c +3
-0+/* Platform might know better if the bridge supports D3 */ +if (platform_pci_bridge_d3 (bridge)) +return true;*All* devices trivially support D3. Obviously we're trying to learn something else here. What is it?
D3 has been problematic with hotplug bridges which is the reason we have not put them in D3 until recently (and still don't do that for ACPI hotplug bridges).
BTW, this patch was merged over a year ago so I'm not sure why you comment it now. Or you want me to add incremental changes to it?
Quoted 67 lines
Quoted 25 lines
drivers/pci/pci.c +0
-0/* * Hotplug ports handled natively by the OS were not validated * by vendors for runtime D3 at least until 2018 because there drivers/pci/pci.h +3
-0@@ -39,6 +39,8 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);/** * struct pci_platform_pm_ops - Firmware PM callbacks * +* @bridge_d3: Does the bridge allow entering into D3 +* * @is_manageable: returns 'true' if given device is power manageable by the * platform firmware * @@ -60,6 +62,7 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);* these callbacks are mandatory. */ struct pci_platform_pm_ops { +bool (*bridge_d3)(struct pci_dev *dev);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);drivers/pci/pci-mid.c +6
-0@@ -16,6 +16,11 @@#include "pci.h" +static bool mid_pci_bridge_d3 (struct pci_dev *dev) +{ +return false; +} +static bool mid_pci_power_manageable (struct pci_dev *dev) {return true;@@ -47,6 +52,7 @@ static bool mid_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops mid_pci_platform_pm = { + .bridge_d3 = mid_pci_bridge_d3, .is_manageable = mid_pci_power_manageable, .set_state = mid_pci_set_power_state, .get_state = mid_pci_get_power_state,drivers/pci/pci.c +3
-2@@ -820,8 +820,9 @@ static const struct pci_platform_pm_ops *pci_platform_pm;int pci_set_platform_pm (const struct pci_platform_pm_ops *ops) { -if (!ops->is_manageable || !ops->set_state || !ops->get_state || - !ops->choose_state || !ops->set_wakeup || !ops->need_resume) +if (!ops->bridge_d3 || !ops->is_manageable || !ops->set_state || + !ops->get_state || !ops->choose_state || !ops->set_wakeup || + !ops->need_resume)return -EINVAL ; pci_platform_pm = ops;return 0 ;
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 27 lines · Mika Westerberg
Quoted 19 lines · Bjorn Helgaas
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this
_DSD, the root port is "capable of handling hot plug events while in D3 state".Quoted 6 lines
What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
Quoted 2 lines
I think it simply gives the OS a hint that it can put PCIe hotplug capable port into D3 and expect it to wake up when device is detected.
I'd really like more specific details than this. PCI power management is explicitly controlled by software, so I don't know what it means for a bridge to "wake up when device is detected."
Normally Linux would get some kind of notification like a PME, then execute an ACPI method and/or write PCI_PM_CTRL to put the device back in D0.
Are we talking about D3hot or D3cold?
The PCI PM capability already has a way to advertise that it can generate PME from D3cold. How is this different from that?
Is this _DSD something that *could* be advertised via PCI config space, i.e., is it completely determined by the Root Port? Or is it something that requires ACPI support, so it cannot be done directly by the hardware device?
These are all things we need to know in order to make power management reliable.
Quoted 107 lines
Quoted 107 lines
Quoted 96 lines
Do the same in Linux by introducing new firmware PM callback (->bridge_d3()) and then implement it for ACPI based systems so that the above property is checked.
There is one catch, though. The initial
pci_dev->bridge_d3is set before the root port has ACPI companion bound (the device is not added to the PCI bus either) so we need to look up the ACPI companion manually in that case inacpi_pci_bridge_d3().Signed-off-by Mika Westerberg <[email protected]>Reviewed-by Rafael J. Wysocki <[email protected]>--- drivers/acpi/property.c | 3 +++ drivers/pci/pci-acpi.c | 41 +++++++++++++++++++++++++++++++++++++++++ drivers/pci/pci.c | 9 +++++++++ drivers/pci/pci.h | 3 +++ 4 files changed, 56 insertions(+)drivers/acpi/property.c +3
-0@@ -28,6 +28,9 @@ static const guid_t prp_guids[] = {/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c ,0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +/* Hotplug in D3 GUID: 6211e2c0-58a3-4af3-90e1-927a4e0c55a4 */ +GUID_INIT (0x6211e2c0 ,0x58a3 ,0x4af3 , + 0x90, 0xe1, 0x92, 0x7a, 0x4e, 0x0c, 0x55, 0xa4), }; static const guid_t ads_guid =drivers/pci/pci-acpi.c +41
-0@@ -519,6 +519,46 @@ static pci_power_t acpi_pci_choose_state(struct pci_dev *pdev)return PCI_POWER_ERROR ; } +static struct acpi_device *acpi_pci_find_companion (struct device *dev); + +static bool acpi_pci_bridge_d3 (struct pci_dev *dev) +{ +const struct fwnode_handle *fwnode; +struct acpi_device *adev; +struct pci_dev *root; +u8 val; + +if (!dev->is_hotplug_bridge) +return false; + +/* +* Look for a special _DSD property for the root port and if it +* is set we know the hierarchy behind it supports D3 just fine. +*/ + root =pci_find_pcie_root_port (dev); +if (!root) +return false; + + adev =ACPI_COMPANION (&root->dev); +if (root == dev) { +/* +* It is possible that the ACPI companion is not yet bound +* for the root port so look it up manually here. +*/ +if (!adev && !pci_dev_is_added (root)) + adev =acpi_pci_find_companion (&root->dev); + } + +if (!adev) +return false; + + fwnode =acpi_fwnode_handle (adev); +if (fwnode_property_read_u8 (fwnode,"HotPlugSupportInD3" , &val)) +return false; + +return val ==1 ; +} +static bool acpi_pci_power_manageable (struct pci_dev *dev) {struct acpi_device *adev =ACPI_COMPANION (&dev->dev);@@ -635,6 +675,7 @@ static bool acpi_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops acpi_pci_platform_pm = { + .bridge_d3 = acpi_pci_bridge_d3, .is_manageable = acpi_pci_power_manageable, .set_state = acpi_pci_set_power_state, .get_state = acpi_pci_get_power_state,drivers/pci/pci.c +0
-0drivers/pci/pci.c static inline bool platform_pci_need_resume(struct pci_dev *dev) · line 796
Quoted 6 lines
drivers/pci/pci.c +3
-0@@ -791,6 +791,11 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)return pci_platform_pm ? pci_platform_pm->need_resume (dev) : false; } +static inline bool platform_pci_bridge_d3 (struct pci_dev *dev) +{ +return pci_platform_pm ? pci_platform_pm->bridge_d3 (dev) : false;This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling
platform_pci_bridge_d3()on a MID platform and trying to call through a NULL pointer?
Quoted 8 lines
Quoted 1 line
Shouldn't we do something like the patch attached below?
IIRC MID devices in general don't have PCIe ports (so we never enable PM for them). Is this a real problem that crashes peoples kernels on MID systems? Then yes I think the patch makes sense to have.
I also remember testing some other stuff on one MID system (Edison) quite recently and did not see any issues.
I have not seen reports of crashes, but I do think this is a real problem. The problem is that the code as-is relies on assumptions ("MID does not have PCI and never will") that are implicit and impossible to verify, which is a maintenance problem.
Quoted 16 lines
Quoted 9 lines
Quoted 9 lines
drivers/pci/pci.c +2
-0+} +/** * pci_raw_set_power_state - Use PCI PM registers to set the power state of * given PCI device @@ -2514,6 +2519,10 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)if (bridge->is_thunderbolt)return true;Quoted 6 lines
drivers/pci/pci.c bool pci_bridge_d3_possible(struct pci_dev *bridge) · line 2524
Quoted 3 lines
drivers/pci/pci.c +3
-0+/* Platform might know better if the bridge supports D3 */ +if (platform_pci_bridge_d3 (bridge)) +return true;*All* devices trivially support D3. Obviously we're trying to learn something else here. What is it?
Quoted 6 lines
D3 has been problematic with hotplug bridges which is the reason we have not put them in D3 until recently (and still don't do that for ACPI hotplug bridges).
BTW, this patch was merged over a year ago so I'm not sure why you comment it now. Or you want me to add incremental changes to it?
I was reviewing [1], which updates pci_find_pcie_root_port(), which is used by acpi_pci_bridge_d3(). I noticed a potential NULL pointer dereference, which was a distraction.
Unless you object, I'll merge something like the patch below to prevent that distraction for others.
[1] https://lore.kernel.org/r/[email protected]
Quoted 67 lines
Quoted 67 lines
Quoted 25 lines
drivers/pci/pci.c +0
-0/* * Hotplug ports handled natively by the OS were not validated * by vendors for runtime D3 at least until 2018 because there drivers/pci/pci.h +3
-0@@ -39,6 +39,8 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);/** * struct pci_platform_pm_ops - Firmware PM callbacks * +* @bridge_d3: Does the bridge allow entering into D3 +* * @is_manageable: returns 'true' if given device is power manageable by the * platform firmware * @@ -60,6 +62,7 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);* these callbacks are mandatory. */ struct pci_platform_pm_ops { +bool (*bridge_d3)(struct pci_dev *dev);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);drivers/pci/pci-mid.c +6
-0@@ -16,6 +16,11 @@#include "pci.h" +static bool mid_pci_bridge_d3 (struct pci_dev *dev) +{ +return false; +} +static bool mid_pci_power_manageable (struct pci_dev *dev) {return true;@@ -47,6 +52,7 @@ static bool mid_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops mid_pci_platform_pm = { + .bridge_d3 = mid_pci_bridge_d3, .is_manageable = mid_pci_power_manageable, .set_state = mid_pci_set_power_state, .get_state = mid_pci_get_power_state,drivers/pci/pci.c +3
-2@@ -820,8 +820,9 @@ static const struct pci_platform_pm_ops *pci_platform_pm;int pci_set_platform_pm (const struct pci_platform_pm_ops *ops) { -if (!ops->is_manageable || !ops->set_state || !ops->get_state || - !ops->choose_state || !ops->set_wakeup || !ops->need_resume) +if (!ops->bridge_d3 || !ops->is_manageable || !ops->set_state || + !ops->get_state || !ops->choose_state || !ops->set_wakeup || + !ops->need_resume)return -EINVAL ; pci_platform_pm = ops;return 0 ;
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 28 lines · Bjorn Helgaas
Quoted 27 lines · Mika Westerberg
Quoted 25 lines · Bjorn Helgaas
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this
_DSD, the root port is "capable of handling hot plug events while in D3 state".What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
Quoted 6 lines
Quoted 2 lines
I think it simply gives the OS a hint that it can put PCIe hotplug capable port into D3 and expect it to wake up when device is detected.
I'd really like more specific details than this. PCI power management is explicitly controlled by software, so I don't know what it means for a bridge to "wake up when device is detected."
Note these are for PCIe which is not the same as the paraller PCI.
What I mean here is that there is some sort of wake depending whether the link is in L1 or L2/L3 and then resulting the link to go to L0 and PME message being send over to the root complex.
Quoted 3 lines
Normally Linux would get some kind of notification like a PME, then execute an ACPI method and/or write
PCI_PM_CTRLto put the device back in D0.
Right.
Quoted 1 line
Are we talking about D3hot or D3cold?
My interpretation is that both (D3 implies both D3hot and D3cold) but I did not write that spec.
Systems where this is used typically go down to D3cold with the PCIe topology (links are in L2).
Quoted 2 lines
The PCI PM capability already has a way to advertise that it can generate PME from D3cold. How is this different from that?
Well you always need some platform (ACPI) help in order to even to D3cold. This applies to waking up as well. The PCIe device may advertise that it supports this but I don't think it can be sure that the system it is connected to has this plumbing.
For PCIe hotplug ports there have been issues that have prevented doing power management for those. The _DSD is there to provide a hint to the OS saying that yeah, this port actually is expected to work even if it goes into D3 (cold/hot).
Quoted 4 lines
Is this
_DSDsomething that *could* be advertised via PCI config space, i.e., is it completely determined by the Root Port? Or is it something that requires ACPI support, so it cannot be done directly by the hardware device?
You always need help from platform (ACPI) to get into D3cold.
Quoted 116 lines
These are all things we need to know in order to make power management reliable.
Quoted 113 lines
Quoted 107 lines
Quoted 96 lines
Do the same in Linux by introducing new firmware PM callback (->bridge_d3()) and then implement it for ACPI based systems so that the above property is checked.
There is one catch, though. The initial
pci_dev->bridge_d3is set before the root port has ACPI companion bound (the device is not added to the PCI bus either) so we need to look up the ACPI companion manually in that case inacpi_pci_bridge_d3().Signed-off-by Mika Westerberg <[email protected]>Reviewed-by Rafael J. Wysocki <[email protected]>--- drivers/acpi/property.c | 3 +++ drivers/pci/pci-acpi.c | 41 +++++++++++++++++++++++++++++++++++++++++ drivers/pci/pci.c | 9 +++++++++ drivers/pci/pci.h | 3 +++ 4 files changed, 56 insertions(+)drivers/acpi/property.c +3
-0@@ -28,6 +28,9 @@ static const guid_t prp_guids[] = {/* ACPI _DSD device properties GUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 */ GUID_INIT (0xdaffd814 ,0x6eba ,0x4d8c ,0x8a ,0x91 ,0xbc ,0x9b ,0xbf ,0x4a ,0xa3 ,0x01 ), +/* Hotplug in D3 GUID: 6211e2c0-58a3-4af3-90e1-927a4e0c55a4 */ +GUID_INIT (0x6211e2c0 ,0x58a3 ,0x4af3 , + 0x90, 0xe1, 0x92, 0x7a, 0x4e, 0x0c, 0x55, 0xa4), }; static const guid_t ads_guid =drivers/pci/pci-acpi.c +41
-0@@ -519,6 +519,46 @@ static pci_power_t acpi_pci_choose_state(struct pci_dev *pdev)return PCI_POWER_ERROR ; } +static struct acpi_device *acpi_pci_find_companion (struct device *dev); + +static bool acpi_pci_bridge_d3 (struct pci_dev *dev) +{ +const struct fwnode_handle *fwnode; +struct acpi_device *adev; +struct pci_dev *root; +u8 val; + +if (!dev->is_hotplug_bridge) +return false; + +/* +* Look for a special _DSD property for the root port and if it +* is set we know the hierarchy behind it supports D3 just fine. +*/ + root =pci_find_pcie_root_port (dev); +if (!root) +return false; + + adev =ACPI_COMPANION (&root->dev); +if (root == dev) { +/* +* It is possible that the ACPI companion is not yet bound +* for the root port so look it up manually here. +*/ +if (!adev && !pci_dev_is_added (root)) + adev =acpi_pci_find_companion (&root->dev); + } + +if (!adev) +return false; + + fwnode =acpi_fwnode_handle (adev); +if (fwnode_property_read_u8 (fwnode,"HotPlugSupportInD3" , &val)) +return false; + +return val ==1 ; +} +static bool acpi_pci_power_manageable (struct pci_dev *dev) {struct acpi_device *adev =ACPI_COMPANION (&dev->dev);@@ -635,6 +675,7 @@ static bool acpi_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops acpi_pci_platform_pm = { + .bridge_d3 = acpi_pci_bridge_d3, .is_manageable = acpi_pci_power_manageable, .set_state = acpi_pci_set_power_state, .get_state = acpi_pci_get_power_state,drivers/pci/pci.c +0
-0drivers/pci/pci.c static inline bool platform_pci_need_resume(struct pci_dev *dev) · line 796
Quoted 6 lines
drivers/pci/pci.c +3
-0@@ -791,6 +791,11 @@ static inline bool platform_pci_need_resume(struct pci_dev *dev)return pci_platform_pm ? pci_platform_pm->need_resume (dev) : false; } +static inline bool platform_pci_bridge_d3 (struct pci_dev *dev) +{ +return pci_platform_pm ? pci_platform_pm->bridge_d3 (dev) : false;This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling
platform_pci_bridge_d3()on a MID platform and trying to call through a NULL pointer?Quoted 1 line
Shouldn't we do something like the patch attached below?
IIRC MID devices in general don't have PCIe ports (so we never enable PM for them). Is this a real problem that crashes peoples kernels on MID systems? Then yes I think the patch makes sense to have.
Quoted 7 lines
Quoted 2 lines
I also remember testing some other stuff on one MID system (Edison) quite recently and did not see any issues.
I have not seen reports of crashes, but I do think this is a real problem. The problem is that the code as-is relies on assumptions ("MID does not have PCI and never will") that are implicit and impossible to verify, which is a maintenance problem.
Well, there won't be any new MID devices and the only one we support partially is Edison and it does not have a single PCIe port that could even be power managed.
Quoted 23 lines
Quoted 16 lines
Quoted 9 lines
Quoted 9 lines
drivers/pci/pci.c +2
-0+} +/** * pci_raw_set_power_state - Use PCI PM registers to set the power state of * given PCI device @@ -2514,6 +2519,10 @@ bool pci_bridge_d3_possible(struct pci_dev *bridge)if (bridge->is_thunderbolt)return true;Quoted 6 lines
drivers/pci/pci.c bool pci_bridge_d3_possible(struct pci_dev *bridge) · line 2524
Quoted 3 lines
drivers/pci/pci.c +3
-0+/* Platform might know better if the bridge supports D3 */ +if (platform_pci_bridge_d3 (bridge)) +return true;*All* devices trivially support D3. Obviously we're trying to learn something else here. What is it?
Quoted 6 lines
D3 has been problematic with hotplug bridges which is the reason we have not put them in D3 until recently (and still don't do that for ACPI hotplug bridges).
BTW, this patch was merged over a year ago so I'm not sure why you comment it now. Or you want me to add incremental changes to it?
Quoted 6 lines
I was reviewing [1], which updates
pci_find_pcie_root_port(), which is used byacpi_pci_bridge_d3(). I noticed a potential NULL pointer dereference, which was a distraction.Unless you object, I'll merge something like the patch below to prevent that distraction for others.
No objections :)
However, in order to make this more maintainable, what if we check in the platform_pci_bridge_d3() itself whether the pointer is NULL and return false? That would work for any future platforms and avoid adding the dummy implementation each time.
Quoted 69 lines
[1] https://lore.kernel.org/r/[email protected]
Quoted 67 lines
Quoted 67 lines
Quoted 25 lines
drivers/pci/pci.c +0
-0/* * Hotplug ports handled natively by the OS were not validated * by vendors for runtime D3 at least until 2018 because there drivers/pci/pci.h +3
-0@@ -39,6 +39,8 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);/** * struct pci_platform_pm_ops - Firmware PM callbacks * +* @bridge_d3: Does the bridge allow entering into D3 +* * @is_manageable: returns 'true' if given device is power manageable by the * platform firmware * @@ -60,6 +62,7 @@ int pci_bridge_secondary_bus_reset(struct pci_dev *dev);* these callbacks are mandatory. */ struct pci_platform_pm_ops { +bool (*bridge_d3)(struct pci_dev *dev);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);drivers/pci/pci-mid.c +6
-0@@ -16,6 +16,11 @@#include "pci.h" +static bool mid_pci_bridge_d3 (struct pci_dev *dev) +{ +return false; +} +static bool mid_pci_power_manageable (struct pci_dev *dev) {return true;@@ -47,6 +52,7 @@ static bool mid_pci_need_resume(struct pci_dev *dev) }static const struct pci_platform_pm_ops mid_pci_platform_pm = { + .bridge_d3 = mid_pci_bridge_d3, .is_manageable = mid_pci_power_manageable, .set_state = mid_pci_set_power_state, .get_state = mid_pci_get_power_state,drivers/pci/pci.c +3
-2@@ -820,8 +820,9 @@ static const struct pci_platform_pm_ops *pci_platform_pm;int pci_set_platform_pm (const struct pci_platform_pm_ops *ops) { -if (!ops->is_manageable || !ops->set_state || !ops->get_state || - !ops->choose_state || !ops->set_wakeup || !ops->need_resume) +if (!ops->bridge_d3 || !ops->is_manageable || !ops->set_state || + !ops->get_state || !ops->choose_state || !ops->set_wakeup || + !ops->need_resume)return -EINVAL ; pci_platform_pm = ops;return 0 ;
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
[breadcrumbs for the archives: this thread is about 26ad34d510a8 ("PCI / ACPI: Whitelist D3 for more PCIe hotplug ports"),
https://git.kernel.org/linus/26ad34d510a8]
Quoted 29 lines · Mika Westerberg
Quoted 28 lines · Bjorn Helgaas
Quoted 27 lines · Mika Westerberg
Quoted 25 lines · Bjorn Helgaas
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this
_DSD, the root port is "capable of handling hot plug events while in D3 state".What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
I think it simply gives the OS a hint that it can put PCIe hotplug capable port into D3 and expect it to wake up when device is detected.
I'd really like more specific details than this. PCI power management is explicitly controlled by software, so I don't know what it means for a bridge to "wake up when device is detected."
Note these are for PCIe which is not the same as the parallel PCI.
Sure, but the PCI power management framework doesn't need to know about PCI/PCIe differences.
Quoted 3 lines
What I mean here is that there is some sort of wake depending whether the link is in L1 or L2/L3 and then resulting the link to go to L0 and PME message being send over to the root complex.
This is compatible with conventional PCI behavior. PME_Support in the PM Capability tells us whether the device can generate a PME from D0, D1, D2, D3hot, or D3cold. AFAIK, this applies to both conventional PCI and PCIe, so I think this _DSD must be telling us something more than just "this PCIe device can assert PME from D3hot and D3cold."
PME_Support tells us whether the device can generate PMEs from D3cold, but of course it doesn't say anything about whether hotplug events cause those PMEs.
PCIe r5.0, sec 6.7.3.4, says ports must support PMEs for hotplug events while the device is in D1, D2, or D3hot. It clearly leaves out D3cold.
So maybe this _DSD tells us that this device can generate *hotplug* event PMEs from D3cold?
Quoted 14 lines
Quoted 3 lines
Normally Linux would get some kind of notification like a PME, then execute an ACPI method and/or write
PCI_PM_CTRLto put the device back in D0.Right.
Quoted 1 line
Are we talking about D3hot or D3cold?
My interpretation is that both (D3 implies both D3hot and D3cold) but I did not write that spec.
Systems where this is used typically go down to D3cold with the PCIe topology (links are in L2).
Quoted 7 lines
Quoted 2 lines
The PCI PM capability already has a way to advertise that it can generate PME from D3cold. How is this different from that?
Well you always need some platform (ACPI) help in order to even to D3cold. This applies to waking up as well. The PCIe device may advertise that it supports this but I don't think it can be sure that the system it is connected to has this plumbing.
Yes, transitions to/from D3cold require platform support. But those transitions are done by ACPI methods, so any plumbing is ACPI's problem, and I don't think this _DSD helps the OS with that.
Quoted 4 lines
For PCIe hotplug ports there have been issues that have prevented doing power management for those. The
_DSDis there to provide a hint to the OS saying that yeah, this port actually is expected to work even if it goes into D3 (cold/hot).
That doesn't really help fix bugs in this area or help with future refactoring, because it's not specific about what the issues were or what the _DSD means.
I think the combination of PCIe r5.0 7.5.2.1 (PME_Support from D0, D1, D2, D3hot, D3cold) and 6.7.3.4 (hotplug PME events from D1, D2, D3hot) covers everything except hotplug PME events from D3cold.
So my guess is that this _DSD is a way to tell the OS that this device generates PMEs for hotplug events even when the device is in D3cold. Does that make any sense?
Quoted 29 lines
Quoted 4 lines
Is this
_DSDsomething that *could* be advertised via PCI config space, i.e., is it completely determined by the Root Port? Or is it something that requires ACPI support, so it cannot be done directly by the hardware device?You always need help from platform (ACPI) to get into D3cold.
Quoted 22 lines
These are all things we need to know in order to make power management reliable.
Quoted 18 lines
Quoted 10 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+static inline bool platform_pci_bridge_d3(struct pci_dev *dev) +{ + return pci_platform_pm ? pci_platform_pm->bridge_d3(dev) : false;This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling
platform_pci_bridge_d3()on a MID platform and trying to call through a NULL pointer?Shouldn't we do something like the patch attached below?
IIRC MID devices in general don't have PCIe ports (so we never enable PM for them). Is this a real problem that crashes peoples kernels on MID systems? Then yes I think the patch makes sense to have.
I also remember testing some other stuff on one MID system (Edison) quite recently and did not see any issues.
Quoted 8 lines
Quoted 4 lines
I have not seen reports of crashes, but I do think this is a real problem. The problem is that the code as-is relies on assumptions ("MID does not have PCI and never will") that are implicit and impossible to verify, which is a maintenance problem.
Well, there won't be any new MID devices and the only one we support partially is Edison and it does not have a single PCIe port that could even be power managed.
The problem I'm trying to solve is making this code understandable. It doesn't help to assert that "there are no devices that exercise this path" because one cannot figure that out by looking at the code.
Quoted 10 lines
Quoted 9 lines
Quoted 2 lines
BTW, this patch was merged over a year ago so I'm not sure why you comment it now. Or you want me to add incremental changes to it?
I was reviewing [1], which updates
pci_find_pcie_root_port(), which is used byacpi_pci_bridge_d3(). I noticed a potential NULL pointer dereference, which was a distraction.Unless you object, I'll merge something like the patch below to prevent that distraction for others.
Quoted 6 lines
No objections :)
However, in order to make this more maintainable, what if we check in the
platform_pci_bridge_d3()itself whether the pointer is NULL and return false? That would work for any future platforms and avoid adding the dummy implementation each time.
That's a much better idea, thank you! I'll cc you when I revise the patch.
Bjorn
Re: [PATCH v2 10/10] PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 37 lines · Bjorn Helgaas
[breadcrumbs for the archives: this thread is about
26ad34d510a8("PCI / ACPI: Whitelist D3 for more PCIe hotplug ports"),
https://git.kernel.org/linus/26ad34d510a8]Quoted 32 lines · Mika Westerberg
Quoted 31 lines · Bjorn Helgaas
Quoted 29 lines · Mika Westerberg
Quoted 19 lines · Bjorn Helgaas
Quoted 11 lines · Mika Westerberg
In order to have better power management for Thunderbolt PCIe chains, Windows enables power management for native PCIe hotplug ports if there is following ACPI
_DSDattached to the root port:Name (_DSD, Package () { ToUUID ("6211e2c0-58a3-4af3-90e1-927a4e0c55a4"), Package () { Package () {"HotPlugSupportInD3", 1} } })Quoted 3 lines
This is also documented in:
This doc basically says that if the platform supplies this
_DSD, the root port is "capable of handling hot plug events while in D3 state".Quoted 6 lines
What does that mean? That statement is not really actionable. I *assume* it's telling us about some specific hardware or firmware functionality, like maybe we'll get a notification for hotplug events when the device is in D3? D3hot? D3cold? What is the notification? Is it immediate or when the device comes back to D0? How do we control and field the notification?
I think it simply gives the OS a hint that it can put PCIe hotplug capable port into D3 and expect it to wake up when device is detected.
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
I'd really like more specific details than this. PCI power management is explicitly controlled by software, so I don't know what it means for a bridge to "wake up when device is detected."
Note these are for PCIe which is not the same as the parallel PCI.
Sure, but the PCI power management framework doesn't need to know about PCI/PCIe differences.
Well it needs to deal with certain differences anyway like the required delays etc.
Quoted 14 lines
Quoted 3 lines
What I mean here is that there is some sort of wake depending whether the link is in L1 or L2/L3 and then resulting the link to go to L0 and PME message being send over to the root complex.
This is compatible with conventional PCI behavior.
PME_Supportin the PM Capability tells us whether the device can generate a PME from D0, D1, D2, D3hot, or D3cold. AFAIK, this applies to both conventional PCI and PCIe, so I think this_DSDmust be telling us something more than just "this PCIe device can assert PME from D3hot and D3cold."
PME_Supporttells us whether the device can generate PMEs from D3cold, but of course it doesn't say anything about whether hotplug events cause those PMEs.
Quoted 6 lines
PCIe r5.0, sec 6.7.3.4, says ports must support PMEs for hotplug events while the device is in D1, D2, or D3hot. It clearly leaves out D3cold.
So maybe this
_DSDtells us that this device can generate *hotplug* event PMEs from D3cold?
Yes, that sounds reasonable.
Quoted 36 lines
Quoted 14 lines
Quoted 3 lines
Normally Linux would get some kind of notification like a PME, then execute an ACPI method and/or write
PCI_PM_CTRLto put the device back in D0.Right.
Quoted 1 line
Are we talking about D3hot or D3cold?
My interpretation is that both (D3 implies both D3hot and D3cold) but I did not write that spec.
Systems where this is used typically go down to D3cold with the PCIe topology (links are in L2).
Quoted 7 lines
Quoted 2 lines
The PCI PM capability already has a way to advertise that it can generate PME from D3cold. How is this different from that?
Well you always need some platform (ACPI) help in order to even to D3cold. This applies to waking up as well. The PCIe device may advertise that it supports this but I don't think it can be sure that the system it is connected to has this plumbing.
Yes, transitions to/from D3cold require platform support. But those transitions are done by ACPI methods, so any plumbing is ACPI's problem, and I don't think this
_DSDhelps the OS with that.Quoted 4 lines
For PCIe hotplug ports there have been issues that have prevented doing power management for those. The
_DSDis there to provide a hint to the OS saying that yeah, this port actually is expected to work even if it goes into D3 (cold/hot).That doesn't really help fix bugs in this area or help with future refactoring, because it's not specific about what the issues were or what the
_DSDmeans.
Quoted 7 lines
I think the combination of PCIe r5.0 7.5.2.1 (
PME_Supportfrom D0, D1, D2, D3hot, D3cold) and 6.7.3.4 (hotplug PME events from D1, D2, D3hot) covers everything except hotplug PME events from D3cold.So my guess is that this
_DSDis a way to tell the OS that this device generates PMEs for hotplug events even when the device is in D3cold. Does that make any sense?
Yes it does.
Quoted 34 lines
Quoted 34 lines
Quoted 4 lines
Is this
_DSDsomething that *could* be advertised via PCI config space, i.e., is it completely determined by the Root Port? Or is it something that requires ACPI support, so it cannot be done directly by the hardware device?You always need help from platform (ACPI) to get into D3cold.
Quoted 26 lines
These are all things we need to know in order to make power management reliable.
Quoted 11 lines
Quoted 10 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+static inline bool platform_pci_bridge_d3(struct pci_dev *dev) +{ + return pci_platform_pm ? pci_platform_pm->bridge_d3(dev) : false;This patch added a .bridge_d3() implementation for ACPI but not for MID. What prevents us from calling
platform_pci_bridge_d3()on a MID platform and trying to call through a NULL pointer?Shouldn't we do something like the patch attached below?
Quoted 7 lines
IIRC MID devices in general don't have PCIe ports (so we never enable PM for them). Is this a real problem that crashes peoples kernels on MID systems? Then yes I think the patch makes sense to have.
I also remember testing some other stuff on one MID system (Edison) quite recently and did not see any issues.
I have not seen reports of crashes, but I do think this is a real problem. The problem is that the code as-is relies on assumptions ("MID does not have PCI and never will") that are implicit and impossible to verify, which is a maintenance problem.
Quoted 7 lines
Quoted 3 lines
Well, there won't be any new MID devices and the only one we support partially is Edison and it does not have a single PCIe port that could even be power managed.
The problem I'm trying to solve is making this code understandable. It doesn't help to assert that "there are no devices that exercise this path" because one cannot figure that out by looking at the code.
Fair enough.
Quoted 12 lines
Quoted 12 lines
Quoted 9 lines
Quoted 2 lines
BTW, this patch was merged over a year ago so I'm not sure why you comment it now. Or you want me to add incremental changes to it?
I was reviewing [1], which updates
pci_find_pcie_root_port(), which is used byacpi_pci_bridge_d3(). I noticed a potential NULL pointer dereference, which was a distraction.Unless you object, I'll merge something like the patch below to prevent that distraction for others.
No objections :)
Quoted 7 lines
Quoted 4 lines
However, in order to make this more maintainable, what if we check in the
platform_pci_bridge_d3()itself whether the pointer is NULL and return false? That would work for any future platforms and avoid adding the dummy implementation each time.That's a much better idea, thank you! I'll cc you when I revise the patch.
OK, thanks!
Quoted 79 lines · Mika Westerberg
Hi all,
This patch series aims to allow PCIe hierarchies enter D3cold both during system sleep and runtime, including PCIe hotplug ports.
The motivation of this series are recent systems such as Lenovo Thinkpad X1 Carbon 6th gen and upcoming Dell laptops where the Thunderbolt controller is always present in the system (pretty much like what Apple systems have been doing for years). Because it is always present it consumes energy so it is important to power it off when idle.
The PCIe root port hosting the Thunderbolt controller and all the connected external PCIe devices includes a standard ACPI power resource and related methods
_PR3,_PR0,_PRWand_DSWthat take care of the actual transition to D3cold and back to D0.With current kernels entering system sleep states leaves the root port into D3hot which means the ACPI power resource is still on. This is bad when system enters suspend-to-idle (all these systems are s2idle only) because the BIOS is not involved and the devices are effectively left on, consuming battery unnecessarily.
Patches 1-4 add D3cold support for system sleep and the subsequent patches do the same for runtime PM.
In case someone wants to try out runtime PM, the xHCI controller that is part of the PCIe switch below the root port needs to have runtime PM "unblocked" manually (this will be automatic in future).
For Thinkpad:
# echo auto > /sys/bus/pci/devices/0000:3b:00.0/power/controlFor Dell:
# echo auto > /sys/bus/pci/devices/0000:39:00.0/power/controlNote if the root port enters D3cold, running things like 'lspci' brings it back t0 D0 (because PCI config space is not accessible in D3cold) so if one wants to check out the power state without accidentally bringing the device back into D0 needs to do that indirectly. If the root port is 1d.0 power state can be checked:
# cat /sys/bus/pci/devices/0000:00:1d.0/power/runtime_status suspended # cat /sys/bus/pci/devices/0000:00:1d.0/firmware_node/real_power_state D3coldEven if this again revolves around Thunderbolt I think these apply to any PCIe system supporting D3cold.
These patches apply on top of pci.git pci/hotplug. The previus version of the series can be viewed here:
https://www.spinics.net/lists/linux-acpi/msg83841.html
Changes from v1:
* Updated comment in patch [1/10] according what Rafael suggested
* Added empty line in patch [2/10]
* Check for !pciehp_poll_mode in [4/10]
* UseDPM_FLAG_NEVER_SKIPinstead ofDPM_FLAG_SMART_PREPAREwhich simplifies patch [5/10] signicantly
* Use const inacpi_data_get_property()and change int -> unsigned intin patch [9/10]* Added tags
Mika Westerberg (10):
PCI: Do not skip power managed bridges inpci_enable_wake()
PCI / ACPI: Enable wake automatically for power managed bridges
PCI: pciehp: Disable hotplug interrupt during suspend
PCI: pciehp: Do not handle events if interrupts are masked
PCI: portdrv: Resume upon exit from system suspend if left runtime suspended
PCI: portdrv: Add runtime PM hooks for port service drivers
PCI: pciehp: Implement runtime PM callbacks
PCI/PME: Implement runtime PM callbacks
ACPI / property: Allow multiple property compatible_DSDentries PCI / ACPI: Whitelist D3 for more PCIe hotplug ports
Quoted 16 lines
drivers/acpi/property.c | 97 ++++++++++++++++++++++--------- drivers/acpi/x86/apple.c | 2 +- drivers/gpio/gpiolib-acpi.c | 2 +- drivers/pci/hotplug/pciehp.h | 2 + drivers/pci/hotplug/pciehp_core.c | 37 ++++++++++++ drivers/pci/hotplug/pciehp_hpc.c | 16 ++++- drivers/pci/pci-acpi.c | 57 +++++++++++++++++- drivers/pci/pci.c | 18 +++++- drivers/pci/pci.h | 3 + drivers/pci/pcie/pme.c | 27 +++++++++ drivers/pci/pcie/portdrv.h | 4 ++ drivers/pci/pcie/portdrv_core.c | 20 +++++++ drivers/pci/pcie/portdrv_pci.c | 14 ++--- include/acpi/acpi_bus.h | 8 ++- include/linux/acpi.h | 9 +++ 15 files changed, 273 insertions(+), 43 deletions(-)
Applied the whole series (not just 01/10, which I applied by itself for some reason) to pci/hotplug for v4.20, thanks!