PCI Power Management ~~~~~~~~~~~~~~~~~~~~ An overview of the concepts and the related functions in the Linux kernel Patrick Mochel (and others) --------------------------------------------------------------------------- 1. Overview 2. How the PCI Subsystem Does Power Management 3. PCI Utility Functions 4. PCI Device Drivers 5. Resources 1. Overview ~~~~~~~~~~~ The PCI Power Management Specification was introduced between the PCI 2.1 and PCI 2.2 Specifications. It a standard interface for controlling various power management operations. Implementation of the PCI PM Spec is optional, as are several sub-components of it. If a device supports the PCI PM Spec, the device will have an 8 byte capability field in its PCI configuration space. This field is used to describe and control the standard PCI power management features. The PCI PM spec defines 4 operating states for devices (D0 - D3) and for buses (B0 - B3). The higher the number, the less power the device consumes. However, the higher the number, the longer the latency is for the device to return to an operational state (D0). There are actually two D3 states. When someone talks about D3, they usually mean D3hot, which corresponds to an ACPI D2 state (power is reduced, the device may lose some context). But they may also mean D3cold, which is an ACPI D3 state (power is fully off, all state was discarded); or both. Bus power management is not covered in this version of this document. Note that all PCI devices support D0 and D3cold by default, regardless of whether or not they implement any of the PCI PM spec. The possible state transitions that a device can undergo are: +---------------------------+ | Current State | New State | +---------------------------+ | D0 | D1, D2, D3| +---------------------------+ | D1 | D2, D3 | +---------------------------+ | D2 | D3 | +---------------------------+ | D1, D2, D3 | D0 | +---------------------------+ Note that when the system is entering a global suspend state, all devices will be placed into D3 and when resuming, all devices will be placed into D0. However, when the system is running, other state transitions are possible. 2. How The PCI Subsystem Handles Power Management ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The PCI suspend/resume functionality is accessed indirectly via the Power Management subsystem. At boot, the PCI driver registers a power management callback with that layer. Upon entering a suspend state, the PM layer iterates through all of its registered callbacks. This currently takes place only during APM state transitions. Upon going to sleep, the PCI subsystem walks its device tree twice. Both times, it does a depth first walk of the device tree. The first walk saves each of the device's state and checks for devices that will prevent the system from entering a global power state. The next walk then places the devices in a low power state. The first walk allows a graceful recovery in the event of a failure, since none of the devices have actually been powered down. In both walks, in particular the second, all children of a bridge are touched before the actual bridge itself. This allows the bridge to retain power while its children are being accessed. Upon resuming from sleep, just the opposite must be true: all bridges must be powered on and restored before their children are powered on. This is easily accomplished with a breadth-first walk of the PCI device tree. 3. PCI Utility Functions ~~~~~~~~~~~~~~~~~~~~~~~~ These are helper functions designed to be called by individual device drivers. Assuming that a device behaves as advertised, these should be applicable in most cases. However, results may vary. Note that these functions are never implicitly called for the driver. The driver is always responsible for deciding when and if to call these. pci_save_state -------------- Usage: pci_save_state(dev, buffer); Description: Save first 64 bytes of PCI config space. Buffer must be allocated by caller. pci_restore_state ----------------- Usage: pci_restore_state(dev, buffer); Description: Restore previously saved config space. (First 64 bytes only); If buffer is NULL, then restore what information we know about the device from bootup: BARs and interrupt line. pci_set_power_state ------------------- Usage: pci_set_power_state(dev, state); Description: Transition device to low power state using PCI PM Capabilities registers. Will fail under one of the following conditions: - If state is less than current state, but not D0 (illegal transition) - Device doesn't support PM Capabilities - Device does not support requested state pci_enable_wake --------------- Usage: pci_enable_wake(dev, state, enable); Description: Enable device to generate PME# during low power state using PCI PM Capabilities. Checks whether if device supports generating PME# from requested state and fail if it does not, unless enable == 0 (request is to disable wake events, which is implicit if it doesn't even support it in the first place). Note that the PMC Register in the device's PM Capabilties has a bitmask of the states it supports generating PME# from. D3hot is bit 3 and D3cold is bit 4. So, while a value of 4 as the state may not seem semantically correct, it is. 4. PCI Device Drivers ~~~~~~~~~~~~~~~~~~~~~ These functions are intended for use by individual drivers, and are defined in struct pci_driver: int (*suspend) (struct pci_dev *dev, pm_message_t state); int (*resume) (struct pci_dev *dev); int (*enable_wake) (struct pci_dev *dev, pci_power_t state, int enable); suspend ------- Usage: if (dev->driver && dev->driver->suspend) dev->driver->suspend(dev,state); A driver uses this function to actually transition the device into a low power state. This should include disabling I/O, IRQs, and bus-mastering, as well as physically transitioning the device to a lower power state; it may also include calls to pci_enable_wake(). Bus mastering may be disabled by doing: pci_disable_device(dev); For devices that support the PCI PM Spec, this may be used to set the device's power state to match the suspend() parameter: pci_set_power_state(dev,state); The driver is also responsible for disabling any other device-specific features (e.g blanking screen, turning off on-card memory, etc). The driver should be sure to track the current state of the device, as it may obviate the need for some operations. The driver should update the current_state field in its pci_dev structure in this function, except for PM-capable devices when pci_set_power_state is used. resume ------ Usage: if (dev->driver && dev->driver->suspend) dev->driver->resume(dev) The resume callback may be called from any power state, and is always meant to transition the device to the D0 state. The driver is responsible for reenabling any features of the device that had been disabled during previous suspend calls, such as IRQs and bus mastering, as well as calling pci_restore_state(). If the device is currently in D3, it may need to be reinitialized in resume(). * Some types of devices, like bus controllers, will preserve context in D3hot (using Vcc power). Their drivers will often want to avoid re-initializing them after re-entering D0 (perhaps to avoid resetting downstream devices). * Other kinds of devices in D3hot will discard device context as part of a soft reset when re-entering the D0 state. * Devices resuming from D3cold always go through a power-on reset. Some device context can also be preserved using Vaux power. * Some systems hide D3cold resume paths from drivers. For example, on PCs the resume path for suspend-to-disk often runs BIOS powerup code, which will sometimes re-initialize the device. To handle resets during D3 to D0 transitions, it may be convenient to share device initialization code between probe() and resume(). Device parameters can also be saved before the driver suspends into D3, avoiding re-probe. If the device supports the PCI PM Spec, it can use this to physically transition the device to D0: pci_set_power_state(dev,0); Note that if the entire system is transitioning out of a global sleep state, all devices will be placed in the D0 state, so this is not necessary. However, in the event that the device is placed in the D3 state during normal operation, this call is necessary. It is impossible to determine which of the two events is taking place in the driver, so it is always a good idea to make that call. The driver should take note of the state that it is resuming from in order to ensure correct (and speedy) operation. The driver should update the current_state field in its pci_dev structure in this function, except for PM-capable devices when pci_set_power_state is used. enable_wake ----------- Usage: if (dev->driver && dev->driver->enable_wake) dev->driver->enable_wake(dev,state,enable); This callback is generally only relevant for devices that support the PCI PM spec and have the ability to generate a PME# (Power Management Event Signal) to wake the system up. (However, it is possible that a device may support some non-standard way of generating a wake event on sleep.) Bits 15:11 of the PMC (Power Mgmt Capabilities) Register in a device's PM Capabilties describe what power states the device supports generating a wake event from: +------------------+ | Bit | State | +------------------+ | 11 | D0 | | 12 | D1 | | 13 | D2 | | 14 | D3hot | | 15 | D3cold | +------------------+ A device can use this to enable wake events: pci_enable_wake(dev,state,enable); Note that to enable PME# from D3cold, a value of 4 should be passed to pci_enable_wake (since it uses an index into a bitmask). If a driver gets a request to enable wake events from D3, two calls should be made to pci_enable_wake (one for both D3hot and D3cold). A reference implementation ------------------------- .suspend() { /* driver specific operations */ /* Disable IRQ */ free_irq(); /* If using MSI */ pci_disable_msi(); pci_save_state(); pci_enable_wake(); /* Disable IO/bus master/irq router */ pci_disable_device(); pci_set_power_state(pci_choose_state()); } .resume() { pci_set_power_state(PCI_D0); pci_restore_state(); /* device's irq possibly is changed, driver should take care */ pci_enable_device(); pci_set_master(); /* if using MSI, device's vector possibly is changed */ pci_enable_msi(); request_irq(); /* driver specific operations; */ } This is a typical implementation. Drivers can slightly change the order of the operations in the implementation, ignore some operations or add more deriver specific operations in it, but drivers should do something like this on the whole. 5. Resources ~~~~~~~~~~~~ PCI Local Bus Specification PCI Bus Power Management Interface Specification http://www.pcisig.com