DRM Internals¶

This chapter documents DRM internals relevant to driver authors and developers working to add support for the latest features to existing drivers.

First, we go over some typical driver initialization requirements, like setting up command buffers, creating an initial output configuration, and initializing core services. Subsequent sections cover core internals in more detail, providing implementation notes and examples.

The DRM layer provides several services to graphics drivers, many of them driven by the application interfaces it provides through libdrm, the library that wraps most of the DRM ioctls. These include vblank event handling, memory management, output management, framebuffer management, command submission & fencing, suspend/resume support, and DMA services.

Driver Initialization¶

At the core of every DRM driver is a struct drm_driver structure. Drivers typically statically initialize a drm_driver structure, and then pass it to drm_dev_alloc() to allocate a device instance. After the device instance is fully initialized it can be registered (which makes it accessible from userspace) using drm_dev_register().

The struct drm_driver structure contains static information that describes the driver and features it supports, and pointers to methods that the DRM core will call to implement the DRM API. We will first go through the struct drm_driver static information fields, and will then describe individual operations in details as they get used in later sections.

Driver Information¶

Driver Features¶

Drivers inform the DRM core about their requirements and supported features by setting appropriate flags in the driver_features field. Since those flags influence the DRM core behaviour since registration time, most of them must be set to registering the struct drm_driver instance.

u32 driver_features;

DRIVER_USE_AGP

Driver uses AGP interface, the DRM core will manage AGP resources.

DRIVER_LEGACY

Denote a legacy driver using shadow attach. Don’t use.

DRIVER_KMS_LEGACY_CONTEXT

Used only by nouveau for backwards compatibility with existing userspace. Don’t use.

DRIVER_PCI_DMA

Driver is capable of PCI DMA, mapping of PCI DMA buffers to userspace will be enabled. Deprecated.

DRIVER_SG

Driver can perform scatter/gather DMA, allocation and mapping of scatter/gather buffers will be enabled. Deprecated.

DRIVER_HAVE_DMA

Driver supports DMA, the userspace DMA API will be supported. Deprecated.

DRIVER_HAVE_IRQ; DRIVER_IRQ_SHARED

DRIVER_HAVE_IRQ indicates whether the driver has an IRQ handler managed by the DRM Core. The core will support simple IRQ handler installation when the flag is set. The installation process is described in ?.

DRIVER_IRQ_SHARED indicates whether the device & handler support shared IRQs (note that this is required of PCI drivers).

DRIVER_GEM

Driver use the GEM memory manager.

DRIVER_MODESET

Driver supports mode setting interfaces (KMS).

DRIVER_PRIME

Driver implements DRM PRIME buffer sharing.

DRIVER_RENDER

Driver supports dedicated render nodes.

DRIVER_ATOMIC

Driver supports atomic properties. In this case the driver must implement appropriate obj->atomic_get_property() vfuncs for any modeset objects with driver specific properties.

Major, Minor and Patchlevel¶

int major; int minor; int patchlevel; The DRM core identifies driver versions by a major, minor and patch level triplet. The information is printed to the kernel log at initialization time and passed to userspace through the DRM_IOCTL_VERSION ioctl.

The major and minor numbers are also used to verify the requested driver API version passed to DRM_IOCTL_SET_VERSION. When the driver API changes between minor versions, applications can call DRM_IOCTL_SET_VERSION to select a specific version of the API. If the requested major isn’t equal to the driver major, or the requested minor is larger than the driver minor, the DRM_IOCTL_SET_VERSION call will return an error. Otherwise the driver’s set_version() method will be called with the requested version.

Name, Description and Date¶

char *name; char *desc; char *date; The driver name is printed to the kernel log at initialization time, used for IRQ registration and passed to userspace through DRM_IOCTL_VERSION.

The driver description is a purely informative string passed to userspace through the DRM_IOCTL_VERSION ioctl and otherwise unused by the kernel.

The driver date, formatted as YYYYMMDD, is meant to identify the date of the latest modification to the driver. However, as most drivers fail to update it, its value is mostly useless. The DRM core prints it to the kernel log at initialization time and passes it to userspace through the DRM_IOCTL_VERSION ioctl.

Device Instance and Driver Handling¶

A device instance for a drm driver is represented by struct drm_device. This is allocated with drm_dev_alloc(), usually from bus-specific ->:c:func:probe() callbacks implemented by the driver. The driver then needs to initialize all the various subsystems for the drm device like memory management, vblank handling, modesetting support and intial output configuration plus obviously initialize all the corresponding hardware bits. Finally when everything is up and running and ready for userspace the device instance can be published using drm_dev_register().

There is also deprecated support for initalizing device instances using bus-specific helpers and the ->:c:func:load() callback. But due to backwards-compatibility needs the device instance have to be published too early, which requires unpretty global locking to make safe and is therefore only support for existing drivers not yet converted to the new scheme.

When cleaning up a device instance everything needs to be done in reverse: First unpublish the device instance with drm_dev_unregister(). Then clean up any other resources allocated at device initialization and drop the driver’s reference to drm_device using drm_dev_unref().

Note that the lifetime rules for drm_device instance has still a lot of historical baggage. Hence use the reference counting provided by drm_dev_ref() and drm_dev_unref() only carefully.

Also note that embedding of drm_device is currently not (yet) supported (but it would be easy to add). Drivers can store driver-private data in the dev_priv field of drm_device.

void drm_put_dev(struct drm_device * dev)¶: Unregister and release a DRM device

Parameters

struct drm_device * dev: DRM device

Description

Called at module unload time or when a PCI device is unplugged.

Cleans up all DRM device, calling drm_lastclose().

Note

Use of this function is deprecated. It will eventually go away completely. Please use drm_dev_unregister() and drm_dev_unref() explicitly instead to make sure that the device isn’t userspace accessible any more while teardown is in progress, ensuring that userspace can’t access an inconsistent state.

int drm_dev_init(struct drm_device * dev, struct drm_driver * driver, struct device * parent)¶: Initialise new DRM device

Parameters

struct drm_device * dev: DRM device
struct drm_driver * driver: DRM driver
struct device * parent: Parent device object

Description

Initialize a new DRM device. No device registration is done. Call drm_dev_register() to advertice the device to user space and register it with other core subsystems. This should be done last in the device initialization sequence to make sure userspace can’t access an inconsistent state.

The initial ref-count of the object is 1. Use drm_dev_ref() and drm_dev_unref() to take and drop further ref-counts.

Note that for purely virtual devices parent can be NULL.

Drivers that do not want to allocate their own device struct embedding struct drm_device can call drm_dev_alloc() instead.

Return

0 on success, or error code on failure.

struct drm_device * drm_dev_alloc(struct drm_driver * driver, struct device * parent)¶: Allocate new DRM device

Parameters

struct drm_driver * driver: DRM driver to allocate device for
struct device * parent: Parent device object

Description

Allocate and initialize a new DRM device. No device registration is done. Call drm_dev_register() to advertice the device to user space and register it with other core subsystems. This should be done last in the device initialization sequence to make sure userspace can’t access an inconsistent state.

The initial ref-count of the object is 1. Use drm_dev_ref() and drm_dev_unref() to take and drop further ref-counts.

Note that for purely virtual devices parent can be NULL.

Drivers that wish to subclass or embed struct drm_device into their own struct should look at using drm_dev_init() instead.

Return

Pointer to new DRM device, or ERR_PTR on failure.

void drm_dev_ref(struct drm_device * dev)¶: Take reference of a DRM device

Parameters

struct drm_device * dev: device to take reference of or NULL

Description

This increases the ref-count of dev by one. You must already own a reference when calling this. Use drm_dev_unref() to drop this reference again.

This function never fails. However, this function does not provide any guarantee whether the device is alive or running. It only provides a reference to the object and the memory associated with it.

void drm_dev_unref(struct drm_device * dev)¶: Drop reference of a DRM device

Parameters

struct drm_device * dev: device to drop reference of or NULL

Description

This decreases the ref-count of dev by one. The device is destroyed if the ref-count drops to zero.

int drm_dev_register(struct drm_device * dev, unsigned long flags)¶: Register DRM device

Parameters

struct drm_device * dev: Device to register
unsigned long flags: Flags passed to the driver’s .:c:func:load() function

Description

Register the DRM device dev with the system, advertise device to user-space and start normal device operation. dev must be allocated via drm_dev_alloc() previously.

Never call this twice on any device!

NOTE

To ensure backward compatibility with existing drivers method this function calls the ->:c:func:load() method after registering the device nodes, creating race conditions. Usage of the ->:c:func:load() methods is therefore deprecated, drivers must perform all initialization before calling drm_dev_register().

Return

0 on success, negative error code on failure.

void drm_dev_unregister(struct drm_device * dev)¶: Unregister DRM device

Parameters

struct drm_device * dev: Device to unregister

Description

Unregister the DRM device from the system. This does the reverse of drm_dev_register() but does not deallocate the device. The caller must call drm_dev_unref() to drop their final reference.

This should be called first in the device teardown code to make sure userspace can’t access the device instance any more.

Driver Load¶

IRQ Registration¶

The DRM core tries to facilitate IRQ handler registration and unregistration by providing drm_irq_install() and drm_irq_uninstall() functions. Those functions only support a single interrupt per device, devices that use more than one IRQs need to be handled manually.

Managed IRQ Registration¶

drm_irq_install() starts by calling the irq_preinstall driver operation. The operation is optional and must make sure that the interrupt will not get fired by clearing all pending interrupt flags or disabling the interrupt.

The passed-in IRQ will then be requested by a call to request_irq(). If the DRIVER_IRQ_SHARED driver feature flag is set, a shared (IRQF_SHARED) IRQ handler will be requested.

The IRQ handler function must be provided as the mandatory irq_handler driver operation. It will get passed directly to request_irq() and thus has the same prototype as all IRQ handlers. It will get called with a pointer to the DRM device as the second argument.

Finally the function calls the optional irq_postinstall driver operation. The operation usually enables interrupts (excluding the vblank interrupt, which is enabled separately), but drivers may choose to enable/disable interrupts at a different time.

drm_irq_uninstall() is similarly used to uninstall an IRQ handler. It starts by waking up all processes waiting on a vblank interrupt to make sure they don’t hang, and then calls the optional irq_uninstall driver operation. The operation must disable all hardware interrupts. Finally the function frees the IRQ by calling free_irq().

Manual IRQ Registration¶

Drivers that require multiple interrupt handlers can’t use the managed IRQ registration functions. In that case IRQs must be registered and unregistered manually (usually with the request_irq() and free_irq() functions, or their devm_request_irq() and devm_free_irq() equivalents).

When manually registering IRQs, drivers must not set the DRIVER_HAVE_IRQ driver feature flag, and must not provide the irq_handler driver operation. They must set the struct drm_device irq_enabled field to 1 upon registration of the IRQs, and clear it to 0 after unregistering the IRQs.

Memory Manager Initialization¶

Every DRM driver requires a memory manager which must be initialized at load time. DRM currently contains two memory managers, the Translation Table Manager (TTM) and the Graphics Execution Manager (GEM). This document describes the use of the GEM memory manager only. See ? for details.

Miscellaneous Device Configuration¶

Another task that may be necessary for PCI devices during configuration is mapping the video BIOS. On many devices, the VBIOS describes device configuration, LCD panel timings (if any), and contains flags indicating device state. Mapping the BIOS can be done using the pci_map_rom() call, a convenience function that takes care of mapping the actual ROM, whether it has been shadowed into memory (typically at address 0xc0000) or exists on the PCI device in the ROM BAR. Note that after the ROM has been mapped and any necessary information has been extracted, it should be unmapped; on many devices, the ROM address decoder is shared with other BARs, so leaving it mapped could cause undesired behaviour like hangs or memory corruption.

Bus-specific Device Registration and PCI Support¶

A number of functions are provided to help with device registration. The functions deal with PCI and platform devices respectively and are only provided for historical reasons. These are all deprecated and shouldn’t be used in new drivers. Besides that there’s a few helpers for pci drivers.

drm_dma_handle_t * drm_pci_alloc(struct drm_device * dev, size_t size, size_t align)¶: Allocate a PCI consistent memory block, for DMA.

Parameters

struct drm_device * dev: DRM device
size_t size: size of block to allocate
size_t align: alignment of block

Return

A handle to the allocated memory block on success or NULL on failure.

void drm_pci_free(struct drm_device * dev, drm_dma_handle_t * dmah)¶: Free a PCI consistent memory block

Parameters

struct drm_device * dev: DRM device
drm_dma_handle_t * dmah: handle to memory block

int drm_get_pci_dev(struct pci_dev * pdev, const struct pci_device_id * ent, struct drm_driver * driver)¶: Register a PCI device with the DRM subsystem

Parameters

struct pci_dev * pdev: PCI device
const struct pci_device_id * ent: entry from the PCI ID table that matches pdev
struct drm_driver * driver: DRM device driver

Description

Attempt to gets inter module “drm” information. If we are first then register the character device and inter module information. Try and register, if we fail to register, backout previous work.

NOTE

This function is deprecated, please use drm_dev_alloc() and drm_dev_register() instead and remove your ->:c:func:load() callback.

Return

0 on success or a negative error code on failure.

int drm_pci_init(struct drm_driver * driver, struct pci_driver * pdriver)¶: Register matching PCI devices with the DRM subsystem

Parameters

struct drm_driver * driver: DRM device driver
struct pci_driver * pdriver: PCI device driver

Description

Initializes a drm_device structures, registering the stubs and initializing the AGP device.

NOTE

This function is deprecated. Modern modesetting drm drivers should use pci_register_driver() directly, this function only provides shadow-binding support for old legacy drivers on top of that core pci function.

Return

0 on success or a negative error code on failure.

void drm_pci_exit(struct drm_driver * driver, struct pci_driver * pdriver)¶: Unregister matching PCI devices from the DRM subsystem

Parameters

struct drm_driver * driver: DRM device driver
struct pci_driver * pdriver: PCI device driver

Description

Unregisters one or more devices matched by a PCI driver from the DRM subsystem.

NOTE

This function is deprecated. Modern modesetting drm drivers should use pci_unregister_driver() directly, this function only provides shadow-binding support for old legacy drivers on top of that core pci function.

int drm_platform_init(struct drm_driver * driver, struct platform_device * platform_device)¶: Register a platform device with the DRM subsystem

Parameters

struct drm_driver * driver: DRM device driver
struct platform_device * platform_device: platform device to register

Description

Registers the specified DRM device driver and platform device with the DRM subsystem, initializing a drm_device structure and calling the driver’s .:c:func:load() function.

NOTE

This function is deprecated, please use drm_dev_alloc() and drm_dev_register() instead and remove your ->:c:func:load() callback.

Return

0 on success or a negative error code on failure.

Open/Close, File Operations and IOCTLs¶

Open and Close¶

Open and close handlers. None of those methods are mandatory:

int (*firstopen) (struct drm_device *);
void (*lastclose) (struct drm_device *);
int (*open) (struct drm_device *, struct drm_file *);
void (*preclose) (struct drm_device *, struct drm_file *);
void (*postclose) (struct drm_device *, struct drm_file *);

The firstopen method is called by the DRM core for legacy UMS (User Mode Setting) drivers only when an application opens a device that has no other opened file handle. UMS drivers can implement it to acquire device resources. KMS drivers can’t use the method and must acquire resources in the load method instead.

Similarly the lastclose method is called when the last application holding a file handle opened on the device closes it, for both UMS and KMS drivers. Additionally, the method is also called at module unload time or, for hot-pluggable devices, when the device is unplugged. The firstopen and lastclose calls can thus be unbalanced.

The open method is called every time the device is opened by an application. Drivers can allocate per-file private data in this method and store them in the struct struct drm_file driver_priv field. Note that the open method is called before firstopen.

The close operation is split into preclose and postclose methods. Drivers must stop and cleanup all per-file operations in the preclose method. For instance pending vertical blanking and page flip events must be cancelled. No per-file operation is allowed on the file handle after returning from the preclose method.

Finally the postclose method is called as the last step of the close operation, right before calling the lastclose method if no other open file handle exists for the device. Drivers that have allocated per-file private data in the open method should free it here.

The lastclose method should restore CRTC and plane properties to default value, so that a subsequent open of the device will not inherit state from the previous user. It can also be used to execute delayed power switching state changes, e.g. in conjunction with the VGA Switcheroo infrastructure. Beyond that KMS drivers should not do any further cleanup. Only legacy UMS drivers might need to clean up device state so that the vga console or an independent fbdev driver could take over.

File Operations¶

Drivers must define the file operations structure that forms the DRM userspace API entry point, even though most of those operations are implemented in the DRM core. The mandatory functions are drm_open(), drm_read(), drm_ioctl() and drm_compat_ioctl if CONFIG_COMPAT is enabled. Drivers which implement private ioctls that require 32/64 bit compatibility support must provided their onw .:c:func:compat_ioctl() handler that processes private ioctls and calls drm_compat_ioctl() for core ioctls.

In addition drm_read() and drm_poll() provide support for DRM events. DRM events are a generic and extensible means to send asynchronous events to userspace through the file descriptor. They are used to send vblank event and page flip completions by the KMS API. But drivers can also use it for their own needs, e.g. to signal completion of rendering.

The memory mapping implementation will vary depending on how the driver manages memory. Legacy drivers will use the deprecated drm_legacy_mmap() function, modern drivers should use one of the provided memory-manager specific implementations. For GEM-based drivers this is drm_gem_mmap().

No other file operations are supported by the DRM userspace API. Overall the following is an example #file_operations structure:

static const example_drm_fops = {
        .owner = THIS_MODULE,
        .open = drm_open,
        .release = drm_release,
        .unlocked_ioctl = drm_ioctl,
#ifdef CONFIG_COMPAT
        .compat_ioctl = drm_compat_ioctl,
#endif
        .poll = drm_poll,
        .read = drm_read,
        .llseek = no_llseek,
        .mmap = drm_gem_mmap,
};

int drm_open(struct inode * inode, struct file * filp)¶: open method for DRM file

Parameters

struct inode * inode: device inode
struct file * filp: file pointer.

Description

This function must be used by drivers as their .:c:func:open() #file_operations method. It looks up the correct DRM device and instantiates all the per-file resources for it.

Return

0 on success or negative errno value on falure.

int drm_release(struct inode * inode, struct file * filp)¶: release method for DRM file

Parameters

struct inode * inode: device inode
struct file * filp: file pointer.

Description

This function must be used by drivers as their .:c:func:release() #file_operations method. It frees any resources associated with the open file, and if this is the last open file for the DRM device also proceeds to call drm_lastclose().

Return

Always succeeds and returns 0.

ssize_t drm_read(struct file * filp, char __user * buffer, size_t count, loff_t * offset)¶: read method for DRM file

Parameters

struct file * filp: file pointer
char __user * buffer: userspace destination pointer for the read
size_t count: count in bytes to read
loff_t * offset: offset to read

Description

This function must be used by drivers as their .:c:func:read() #file_operations method iff they use DRM events for asynchronous signalling to userspace. Since events are used by the KMS API for vblank and page flip completion this means all modern display drivers must use it.

offset is ignore, DRM events are read like a pipe. Therefore drivers also must set the .:c:func:llseek() #file_operation to no_llseek(). Polling support is provided by drm_poll().

This function will only ever read a full event. Therefore userspace must supply a big enough buffer to fit any event to ensure forward progress. Since the maximum event space is currently 4K it’s recommended to just use that for safety.

Return

Number of bytes read (always aligned to full events, and can be 0) or a negative error code on failure.

unsigned int drm_poll(struct file * filp, struct poll_table_struct * wait)¶: poll method for DRM file

Parameters

struct file * filp: file pointer
struct poll_table_struct * wait: poll waiter table

Description

This function must be used by drivers as their .:c:func:read() #file_operations method iff they use DRM events for asynchronous signalling to userspace. Since events are used by the KMS API for vblank and page flip completion this means all modern display drivers must use it.

IOCTLs¶

struct drm_ioctl_desc *ioctls; int num_ioctls;: Driver-specific ioctls descriptors table.

Driver-specific ioctls numbers start at DRM_COMMAND_BASE. The ioctls descriptors table is indexed by the ioctl number offset from the base value. Drivers can use the DRM_IOCTL_DEF_DRV() macro to initialize the table entries.

DRM_IOCTL_DEF_DRV(ioctl, func, flags)

ioctl is the ioctl name. Drivers must define the DRM_##ioctl and DRM_IOCTL_##ioctl macros to the ioctl number offset from DRM_COMMAND_BASE and the ioctl number respectively. The first macro is private to the device while the second must be exposed to userspace in a public header.

func is a pointer to the ioctl handler function compatible with the drm_ioctl_t type.

typedef int drm_ioctl_t(struct drm_device *dev, void *data,
        struct drm_file *file_priv);

flags is a bitmask combination of the following values. It restricts how the ioctl is allowed to be called.

DRM_AUTH - Only authenticated callers allowed
DRM_MASTER - The ioctl can only be called on the master file handle
DRM_ROOT_ONLY - Only callers with the SYSADMIN capability allowed
DRM_CONTROL_ALLOW - The ioctl can only be called on a control device
DRM_UNLOCKED - The ioctl handler will be called without locking the DRM global mutex. This is the enforced default for kms drivers (i.e. using the DRIVER_MODESET flag) and hence shouldn’t be used any more for new drivers.

int drm_noop(struct drm_device * dev, void * data, struct drm_file * file_priv)¶: DRM no-op ioctl implemntation

Parameters

struct drm_device * dev: DRM device for the ioctl
void * data: data pointer for the ioctl
struct drm_file * file_priv: DRM file for the ioctl call

Description

This no-op implementation for drm ioctls is useful for deprecated functionality where we can’t return a failure code because existing userspace checks the result of the ioctl, but doesn’t care about the action.

Always returns successfully with 0.

int drm_invalid_op(struct drm_device * dev, void * data, struct drm_file * file_priv)¶: DRM invalid ioctl implemntation

Parameters

struct drm_device * dev: DRM device for the ioctl
void * data: data pointer for the ioctl
struct drm_file * file_priv: DRM file for the ioctl call

Description

This no-op implementation for drm ioctls is useful for deprecated functionality where we really don’t want to allow userspace to call the ioctl any more. This is the case for old ums interfaces for drivers that transitioned to kms gradually and so kept the old legacy tables around. This only applies to radeon and i915 kms drivers, other drivers shouldn’t need to use this function.

Always fails with a return value of -EINVAL.

long drm_ioctl(struct file * filp, unsigned int cmd, unsigned long arg)¶: ioctl callback implementation for DRM drivers

Parameters

struct file * filp: file this ioctl is called on
unsigned int cmd: ioctl cmd number
unsigned long arg: user argument

Description

Looks up the ioctl function in the ::ioctls table, checking for root previleges if so required, and dispatches to the respective function.

Return

Zero on success, negative error code on failure.

bool drm_ioctl_flags(unsigned int nr, unsigned int * flags)¶: Check for core ioctl and return ioctl permission flags

Parameters

unsigned int nr: ioctl number
unsigned int * flags: where to return the ioctl permission flags

Description

This ioctl is only used by the vmwgfx driver to augment the access checks done by the drm core and insofar a pretty decent layering violation. This shouldn’t be used by any drivers.

Return

True if the nr corresponds to a DRM core ioctl number, false otherwise.

Legacy Support Code¶

The section very briefly covers some of the old legacy support code which is only used by old DRM drivers which have done a so-called shadow-attach to the underlying device instead of registering as a real driver. This also includes some of the old generic buffer management and command submission code. Do not use any of this in new and modern drivers.

Legacy Suspend/Resume¶

The DRM core provides some suspend/resume code, but drivers wanting full suspend/resume support should provide save() and restore() functions. These are called at suspend, hibernate, or resume time, and should perform any state save or restore required by your device across suspend or hibernate states.

int (*suspend) (struct drm_device *, pm_message_t state); int (*resume) (struct drm_device *); Those are legacy suspend and resume methods which only work with the legacy shadow-attach driver registration functions. New driver should use the power management interface provided by their bus type (usually through the struct device_driver dev_pm_ops) and set these methods to NULL.

Legacy DMA Services¶

This should cover how DMA mapping etc. is supported by the core. These functions are deprecated and should not be used.