Many AI Processur Unit (APU) have a similar architecture. This driver intends helping supporting them. This relies on DRM and provides some abstractions useful for AI accelerators. Currently, this provides the infrastructure to alloc an APU device and register one or many cores. The driver will takes care to register itself to DRM.

Signed-off-by: Alexandre Bailon abailon@baylibre.com Reviewed-by: Julien Stephan jstephan@baylibre.com --- drivers/gpu/drm/Kconfig | 2 + drivers/gpu/drm/Makefile | 1 + drivers/gpu/drm/apu/Kconfig | 12 ++ drivers/gpu/drm/apu/Makefile | 5 + drivers/gpu/drm/apu/apu_drv.c | 272 +++++++++++++++++++++++++++++ drivers/gpu/drm/apu/apu_internal.h | 68 ++++++++ include/uapi/drm/apu_drm.h | 28 +++ 7 files changed, 388 insertions(+) create mode 100644 drivers/gpu/drm/apu/Kconfig create mode 100644 drivers/gpu/drm/apu/Makefile create mode 100644 drivers/gpu/drm/apu/apu_drv.c create mode 100644 drivers/gpu/drm/apu/apu_internal.h create mode 100644 include/uapi/drm/apu_drm.h

diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index ba3fb04bb691..32ffa66a8b54 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -371,6 +371,8 @@ source "drivers/gpu/drm/solomon/Kconfig"

source "drivers/gpu/drm/sprd/Kconfig"

+source "drivers/gpu/drm/apu/Kconfig" + config DRM_HYPERV tristate "DRM Support for Hyper-V synthetic video device" depends on DRM && PCI && MMU && HYPERV diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index a33257d2bc7f..7cd8c0f3936a 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -191,6 +191,7 @@ obj-$(CONFIG_DRM_MCDE) += mcde/ obj-$(CONFIG_DRM_TIDSS) += tidss/ obj-y += xlnx/ obj-y += gud/ +obj-$(CONFIG_DRM_APU) += apu/ obj-$(CONFIG_DRM_HYPERV) += hyperv/ obj-y += solomon/ obj-$(CONFIG_DRM_SPRD) += sprd/ diff --git a/drivers/gpu/drm/apu/Kconfig b/drivers/gpu/drm/apu/Kconfig new file mode 100644 index 000000000000..226dcf072115 --- /dev/null +++ b/drivers/gpu/drm/apu/Kconfig @@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0-only +# + +config DRM_APU + tristate "APU (AI Processor Unit)" + select DRM_GEM_DMA_HELPER + select DRM_KMS_HELPER + help + This provides a DRM driver that provides some facilities to + communicate with an AI Processor Unit (APU). + The driver intends to provide a common infrastructure that may be + used to support many different APU. diff --git a/drivers/gpu/drm/apu/Makefile b/drivers/gpu/drm/apu/Makefile new file mode 100644 index 000000000000..ad85b88a8b52 --- /dev/null +++ b/drivers/gpu/drm/apu/Makefile @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: GPL-2.0 + +drm_apu-y += apu_drv.o + +obj-$(CONFIG_DRM_APU) += drm_apu.o diff --git a/drivers/gpu/drm/apu/apu_drv.c b/drivers/gpu/drm/apu/apu_drv.c new file mode 100644 index 000000000000..b420b13a9ffd --- /dev/null +++ b/drivers/gpu/drm/apu/apu_drv.c @@ -0,0 +1,272 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// Copyright 2020 BayLibre SAS + +#include <linux/list.h> +#include <linux/module.h> + +#include <drm/apu_drm.h> +#include <drm/drm_drv.h> +#include <drm/drm_gem_dma_helper.h> +#include <drm/drm_probe_helper.h> + +#include "apu_internal.h" + +static LIST_HEAD(apu_devices); + +static int ioctl_apu_state(struct drm_device *dev, void *data, + struct drm_file *file_priv); + +static const struct drm_ioctl_desc ioctls[] = { + DRM_IOCTL_DEF_DRV(APU_STATE, ioctl_apu_state, + DRM_RENDER_ALLOW), +}; + +DEFINE_DRM_GEM_DMA_FOPS(apu_drm_ops); + +static struct drm_driver apu_drm_driver = { + .driver_features = DRIVER_GEM | DRIVER_SYNCOBJ, + .name = "drm_apu", + .desc = "APU DRM driver", + .date = "20210319", + .major = 1, + .minor = 0, + .patchlevel = 0, + .ioctls = ioctls, + .num_ioctls = ARRAY_SIZE(ioctls), + .fops = &apu_drm_ops, + DRM_GEM_DMA_DRIVER_OPS_WITH_DUMB_CREATE(drm_gem_dma_dumb_create), +}; + +/** + * apu_dev_alloc() - Allocate a new APU device + * + * @dev: Pointer to the device instance. + + * This allocate an APU device. + * The APU describe a hardware accelerator that may have one or more + * core (or unit). + * + * Returns: A pointer or NULL in case of failure. + */ +struct apu_drm *apu_dev_alloc(struct device *dev) +{ + struct drm_device *drm; + struct apu_drm *apu; + + apu = devm_drm_dev_alloc(dev, &apu_drm_driver, typeof(*apu), base); + if (IS_ERR(apu)) + return NULL; + INIT_LIST_HEAD(&apu->cores); + + apu->dev = dev; + ida_init(&apu->ida); + drm = &apu->base; + drm->dev_private = apu; + + dev_set_drvdata(dev, drm); + + return apu; +} +EXPORT_SYMBOL_GPL(apu_dev_alloc); + +/** + * apu_dev_register() - Register the APU to DRM + * + * @apu: Pointer to APU device + * + * Register an APU device to DRM. + * On success, this creates everything required to use the APU. + * Note that at this step, the cores (or units) have not been + * registered so we can't yet perform any operations. + * + * Returns: Zero on success, non-zero value on failure. + */ +int apu_dev_register(struct apu_drm *apu) +{ + struct drm_device *drm = &apu->base; + int ret; + + ret = drm_dev_register(drm, 0); + if (ret) + return ret; + + list_add(&apu->node, &apu_devices); + + return 0; +} +EXPORT_SYMBOL_GPL(apu_dev_register); + +/** + * apu_dev_unregister() - Unregister the APU + * + * @apu: Pointer to APU device + * + * This undo what has been done by apu_dev_register(); + */ +void apu_dev_unregister(struct apu_drm *apu) +{ + struct drm_device *drm = &apu->base; + + list_del(&apu->node); + drm_dev_unregister(drm); +} +EXPORT_SYMBOL_GPL(apu_dev_unregister); + +/** + * apu_core_alloc() - Allocate an APU core + * + * @apu: Pointer to APU device + * @ops: The operation callbacks to use for this core + * @priv: + * + * Allocate an APU core. This represents a computing unit that could + * execute a job. The APU may be composed of different units that doesn't + * accept same kind of jobs so we may to use differents callbacks for each + * core. + * + * Returns: A pointer or NULL in case of failure. + */ +struct apu_core *apu_core_alloc(struct apu_drm *apu, struct apu_core_ops *ops, + void *priv) +{ + struct apu_core *core; + + if (!ops || !ops->is_ready) + return NULL; + + core = devm_kzalloc(apu->dev, sizeof(*core), GFP_KERNEL); + if (!core) + return NULL; + + core->device_id = ida_alloc(&apu->ida, GFP_KERNEL); + if (core->device_id < 0) + return NULL; + + core->apu = apu; + core->priv = priv; + core->ops = ops; + + list_add(&core->node, &apu->cores); + + return core; +} +EXPORT_SYMBOL_GPL(apu_core_alloc); + +/** + * apu_core_free() - Free an APU core allocated using apu_core_alloc() + * + * @core: The APU core to release + */ +void apu_core_free(struct apu_core *core) +{ + ida_free(&core->apu->ida, core->device_id); + list_del(&core->node); +} +EXPORT_SYMBOL_GPL(apu_core_free); + +/** + * apu_core_register() - Register a core to APU device + * + * @dev: Pointer to APU device + * @core: Pointer to APU core to register + * @priv: Private data attached to this core + * + * Register an APU core and make it available for computing. + * On success, userspace can start using this core. + * + * Returns: Zero on success, non-zero value on failure. + */ +int apu_core_register(struct device *dev, struct apu_core *core, void *priv) +{ + int ret; + + core->dev_priv = priv; + core->dev = dev; + + if (core->ops->register_core) { + ret = core->ops->register_core(core); + if (ret) + return ret; + } + + return 0; +} +EXPORT_SYMBOL_GPL(apu_core_register); + +/** + * apu_core_remove() - Remove a core from the APU device + * + * @core: Pointer to APU core to remove + */ +void apu_core_remove(struct apu_core *core) +{ + core->dev_priv = NULL; +} +EXPORT_SYMBOL_GPL(apu_core_remove); + +/** + * apu_find_core_by_priv() - Find a core allocated by apu_core_alloc() + * + * @priv: The pointer used to allocate the core + * + * All core allocated using apu_core_alloc() is registered to a list. + * This goes through the list to find the core using the @priv field. + * + * Returns: A pointer or NULL if no core has been found. + */ +struct apu_core *apu_find_core_by_priv(void *priv) +{ + struct apu_drm *apu; + struct apu_core *core; + + list_for_each_entry(apu, &apu_devices, node) { + list_for_each_entry(core, &apu->cores, node) { + if (core->priv == priv) + return core; + } + } + + return NULL; +} +EXPORT_SYMBOL_GPL(apu_find_core_by_priv); + +static struct apu_core *get_apu_core(struct apu_drm *apu, int device_id) +{ + struct apu_core *core; + + list_for_each_entry(core, &apu->cores, node) { + if (core->device_id == device_id) + return core; + } + + return NULL; +} + +static void apu_core_update_state(struct apu_core *core) +{ + if (!core->ops->is_ready(core)) + core->flags &= ~APU_ONLINE; +} + +static int ioctl_apu_state(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct apu_drm *apu = dev->dev_private; + struct drm_apu_state *args = data; + struct apu_core *core; + + args->flags = 0; + + core = get_apu_core(apu, args->device); + if (!core) + return -ENODEV; + + apu_core_update_state(core); + args->flags |= core->flags; + + return 0; +} + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Alexandre Bailon"); diff --git a/drivers/gpu/drm/apu/apu_internal.h b/drivers/gpu/drm/apu/apu_internal.h new file mode 100644 index 000000000000..58d93a16c68f --- /dev/null +++ b/drivers/gpu/drm/apu/apu_internal.h @@ -0,0 +1,68 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __APU_INTERNAL_H__ +#define __APU_INTERNAL_H__ + +#include <drm/drm_drv.h> + +struct apu_core { + int device_id; + struct device *dev; + struct apu_core_ops *ops; + struct apu_drm *apu; + + struct list_head node; + void *priv; + void *dev_priv; + + u32 flags; +}; + +struct apu_drm { + struct drm_device base; + struct device *dev; + + struct list_head cores; + struct list_head node; + + struct ida ida; +}; + +/** + * @apu_core_ops: Provides platform specific callbacks + */ +struct apu_core_ops { + /** + * @register_core: + * + * Optional. Platform specific APU core registration. + */ + int (*register_core)(struct apu_core *core); + + /** + * @is_ready: + * + * Implements platform specific code to test if APU is ready to receive + * commands. + * Basically, an APU core may be running but not be ready to handle + * commands. This allows checking if APU is ready and start executing + * requests. + * + * Returns: + * + * One if the APU is ready or zero. + */ + int (*is_ready)(struct apu_core *core); +}; + +struct apu_drm *apu_dev_alloc(struct device *dev); +int apu_dev_register(struct apu_drm *apu); +void apu_dev_unregister(struct apu_drm *apu); + +struct apu_core *apu_core_alloc(struct apu_drm *apu, struct apu_core_ops *ops, + void *priv); +void apu_core_free(struct apu_core *core); +int apu_core_register(struct device *dev, struct apu_core *core, void *priv); +void apu_core_remove(struct apu_core *core); +struct apu_core *apu_find_core_by_priv(void *priv); + +#endif /* __APU_INTERNAL_H__ */ diff --git a/include/uapi/drm/apu_drm.h b/include/uapi/drm/apu_drm.h new file mode 100644 index 000000000000..d50c63d1b813 --- /dev/null +++ b/include/uapi/drm/apu_drm.h @@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note*/ + +#ifndef __UAPI_APU_DRM_H__ +#define __UAPI_APU_DRM_H__ + +#include "drm.h" + +#if defined(__cplusplus) +extern "C" { +#endif + +#define APU_ONLINE BIT(0) + +struct drm_apu_state { + __u32 device; + __u32 flags; +}; + +#define DRM_APU_STATE 0x00 +#define DRM_APU_NUM_IOCTLS 0x01 + +#define DRM_IOCTL_APU_STATE DRM_IOWR(DRM_COMMAND_BASE + DRM_APU_STATE, struct drm_apu_state) + +#if defined(__cplusplus) +} +#endif + +#endif /* __UAPI_APU_DRM_H__ */

This updates the APU driver to with two new ioctls to queue and dequeue requests. This uses DRM scheduler to manage the requests. The requests allocation and send and receive operations are platform specifics and must be implemented as callback.

Signed-off-by: Alexandre Bailon abailon@baylibre.com Reviewed-by: Julien Stephan jstephan@baylibre.com --- drivers/gpu/drm/apu/Kconfig | 1 + drivers/gpu/drm/apu/Makefile | 1 + drivers/gpu/drm/apu/apu_drv.c | 52 +-- drivers/gpu/drm/apu/apu_internal.h | 93 ++++- drivers/gpu/drm/apu/apu_sched.c | 564 +++++++++++++++++++++++++++++ include/uapi/drm/apu_drm.h | 31 +- 6 files changed, 697 insertions(+), 45 deletions(-) create mode 100644 drivers/gpu/drm/apu/apu_sched.c

diff --git a/drivers/gpu/drm/apu/Kconfig b/drivers/gpu/drm/apu/Kconfig index 226dcf072115..a769df42091c 100644 --- a/drivers/gpu/drm/apu/Kconfig +++ b/drivers/gpu/drm/apu/Kconfig @@ -5,6 +5,7 @@ config DRM_APU tristate "APU (AI Processor Unit)" select DRM_GEM_DMA_HELPER select DRM_KMS_HELPER + select DRM_SCHED help This provides a DRM driver that provides some facilities to communicate with an AI Processor Unit (APU). diff --git a/drivers/gpu/drm/apu/Makefile b/drivers/gpu/drm/apu/Makefile index 91894250d4c1..fc8d6380fc38 100644 --- a/drivers/gpu/drm/apu/Makefile +++ b/drivers/gpu/drm/apu/Makefile @@ -2,5 +2,6 @@

drm_apu-y += apu_drv.o drm_apu-y += apu_gem.o +drm_apu-y += apu_sched.o

obj-$(CONFIG_DRM_APU) += drm_apu.o diff --git a/drivers/gpu/drm/apu/apu_drv.c b/drivers/gpu/drm/apu/apu_drv.c index 323e267b0f53..b6bd340b2bc8 100644 --- a/drivers/gpu/drm/apu/apu_drv.c +++ b/drivers/gpu/drm/apu/apu_drv.c @@ -14,14 +14,15 @@

static LIST_HEAD(apu_devices);

-static int ioctl_apu_state(struct drm_device *dev, void *data, - struct drm_file *file_priv); - static const struct drm_ioctl_desc ioctls[] = { DRM_IOCTL_DEF_DRV(APU_STATE, ioctl_apu_state, DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(APU_GEM_NEW, ioctl_gem_new, DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(APU_GEM_QUEUE, ioctl_gem_queue, + DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(APU_GEM_DEQUEUE, ioctl_gem_dequeue, + DRM_RENDER_ALLOW), };

DEFINE_DRM_GEM_DMA_FOPS(apu_drm_ops); @@ -134,7 +135,8 @@ struct apu_core *apu_core_alloc(struct apu_drm *apu, struct apu_core_ops *ops, { struct apu_core *core;

- if (!ops || !ops->is_ready) + if (!ops || !ops->alloc_prepare_request || !ops->send_request + || !ops->handle_request || !ops->is_ready) return NULL;

core = devm_kzalloc(apu->dev, sizeof(*core), GFP_KERNEL); @@ -148,6 +150,8 @@ struct apu_core *apu_core_alloc(struct apu_drm *apu, struct apu_core_ops *ops, core->apu = apu; core->priv = priv; core->ops = ops; + spin_lock_init(&core->ctx_lock); + INIT_LIST_HEAD(&core->requests);

list_add(&core->node, &apu->cores);

@@ -192,7 +196,7 @@ int apu_core_register(struct device *dev, struct apu_core *core, void *priv) return ret; }

- return 0; + return apu_drm_job_init(core); } EXPORT_SYMBOL_GPL(apu_core_register);

@@ -203,6 +207,7 @@ EXPORT_SYMBOL_GPL(apu_core_register); */ void apu_core_remove(struct apu_core *core) { + apu_sched_fini(core); core->dev_priv = NULL; } EXPORT_SYMBOL_GPL(apu_core_remove); @@ -233,42 +238,5 @@ struct apu_core *apu_find_core_by_priv(void *priv) } EXPORT_SYMBOL_GPL(apu_find_core_by_priv);

-static struct apu_core *get_apu_core(struct apu_drm *apu, int device_id) -{ - struct apu_core *core; - - list_for_each_entry(core, &apu->cores, node) { - if (core->device_id == device_id) - return core; - } - - return NULL; -} - -static void apu_core_update_state(struct apu_core *core) -{ - if (!core->ops->is_ready(core)) - core->flags &= ~APU_ONLINE; -} - -static int ioctl_apu_state(struct drm_device *dev, void *data, - struct drm_file *file_priv) -{ - struct apu_drm *apu = dev->dev_private; - struct drm_apu_state *args = data; - struct apu_core *core; - - args->flags = 0; - - core = get_apu_core(apu, args->device); - if (!core) - return -ENODEV; - - apu_core_update_state(core); - args->flags |= core->flags; - - return 0; -} - MODULE_LICENSE("GPL"); MODULE_AUTHOR("Alexandre Bailon"); diff --git a/drivers/gpu/drm/apu/apu_internal.h b/drivers/gpu/drm/apu/apu_internal.h index 203aadc58b72..021a3efdedf2 100644 --- a/drivers/gpu/drm/apu/apu_internal.h +++ b/drivers/gpu/drm/apu/apu_internal.h @@ -4,6 +4,7 @@

#include <drm/drm_drv.h> #include <drm/drm_gem_dma_helper.h> +#include <drm/gpu_scheduler.h>

struct apu_gem_object { struct drm_gem_dma_object base; @@ -12,16 +13,21 @@ struct apu_gem_object { u32 offset; };

+struct apu_sched; struct apu_core { int device_id; struct device *dev; struct apu_core_ops *ops; struct apu_drm *apu;

+ spinlock_t ctx_lock; + struct list_head requests; + struct list_head node; void *priv; void *dev_priv;

+ struct apu_sched *sched; u32 flags; };

@@ -35,6 +41,43 @@ struct apu_drm { struct ida ida; };

+struct apu_job { + struct drm_sched_job base; + + struct kref refcount; + + struct apu_core *core; + struct apu_drm *apu; + + /* Fence to be signaled by IRQ handler when the job is complete. */ + struct dma_fence *done_fence; + + __u32 cmd; + + /* Exclusive fences we have taken from the BOs to wait for */ + struct dma_fence **implicit_fences; + struct drm_gem_object **bos; + u32 bo_count; + + /* Fence to be signaled by drm-sched once its done with the job */ + struct dma_fence *render_done_fence; + + void *data_in; + uint16_t size_in; + void *data_out; + uint16_t size_out; + uint16_t result; + uint16_t id; + + struct list_head node; + struct drm_syncobj *sync_out; + + struct apu_event *event; + + void *request_data; + int request_len; +}; + /** * @apu_core_ops: Provides platform specific callbacks */ @@ -46,6 +89,40 @@ struct apu_core_ops { */ int (*register_core)(struct apu_core *core);

+ /** + * @alloc_prepare_request: + * + * Allocate and initialize the data to be sent to the APU. + * Basically, this must convert the job object to something + * that could be understand by the hardware accelerator. + * + * Returns: + * + * Zero on success, non-zero value on failure. + */ + int (*alloc_prepare_request)(struct apu_job *job); + + /** + * @send_data: + * + * Implements platform specific code to send the data previously + * allocated using @alloc_data. + * + * Returns: + * + * Zero on success, non-zero value on failure. + */ + int (*send_request)(struct apu_job *job); + + /** + * @handle_request: + * + * Implements platform specific code to handle incoming request. + * This must decode the data encoded using platform specific format + * and convert it to generic format. + */ + int (*handle_request)(struct apu_job *job, void *data, int len); + /** * @is_ready: * @@ -68,9 +145,9 @@ static inline struct apu_gem_object *to_apu_bo(struct drm_gem_object *obj) base); }

-static inline void *apu_drm_priv(struct apu_core *apu_core) +static inline void *apu_drm_priv(struct apu_core *core) { - return apu_core->dev_priv; + return core->dev_priv; }

struct apu_drm *apu_dev_alloc(struct device *dev); @@ -94,5 +171,17 @@ int ioctl_gem_user_new(struct drm_device *dev, void *data, struct drm_file *file_priv); struct dma_buf *apu_gem_prime_export(struct drm_gem_object *gem, int flags); +int ioctl_gem_queue(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int ioctl_gem_dequeue(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int ioctl_apu_state(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int apu_drm_callback(struct apu_core *core, uint16_t job_id, void *data, int len); + +struct apu_job; + +int apu_drm_job_init(struct apu_core *core); +void apu_sched_fini(struct apu_core *core);

#endif /* __APU_INTERNAL_H__ */ diff --git a/drivers/gpu/drm/apu/apu_sched.c b/drivers/gpu/drm/apu/apu_sched.c new file mode 100644 index 000000000000..13b6fbd00bd8 --- /dev/null +++ b/drivers/gpu/drm/apu/apu_sched.c @@ -0,0 +1,564 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// Copyright 2020 BayLibre SAS + +#include <drm/apu_drm.h> +#include <drm/drm_syncobj.h> +#include <drm/gpu_scheduler.h> + +#include "apu_internal.h" + +struct apu_queue_state { + struct drm_gpu_scheduler sched; + u64 fence_context; + u64 seqno; +}; + +struct apu_request { + struct list_head node; + struct apu_job *job; +}; + +struct apu_sched { + struct apu_queue_state apu_queue; + spinlock_t job_lock; + struct drm_sched_entity sched_entity; +}; + +struct apu_event { + struct drm_pending_event pending_event; + union { + struct drm_event base; + struct apu_job_event job_event; + }; +}; + +static DEFINE_IDA(req_ida); +static LIST_HEAD(complete_node); + +static void apu_core_update_state(struct apu_core *core) +{ + if (!core->ops->is_ready(core)) + core->flags &= ~APU_ONLINE; +} + +static int apu_core_is_running(struct apu_core *core) +{ + apu_core_update_state(core); + + return core->flags & APU_ONLINE; +} + +static void apu_set_online(struct apu_core *core) +{ + core->flags |= APU_ONLINE; +} + +static void apu_set_offline(struct apu_core *core) +{ + core->flags &= ~APU_ONLINE; +} + +/* + * apu_drm_callback() - Handle the data coming from accelerator + * + * @core: The pointer to the APU core + * @job_id: The job id + * @data: The data coming from the accelerator + * @len: The size of the data + * + * Returns: Zero on success, non-zero value on failure. + */ +int apu_drm_callback(struct apu_core *core, uint16_t job_id, void *data, int len) +{ + struct apu_request *apu_req, *tmp; + unsigned long flags; + int ret = -EINVAL; + + spin_lock_irqsave(&core->ctx_lock, flags); + list_for_each_entry_safe(apu_req, tmp, &core->requests, node) { + struct apu_job *job = apu_req->job; + + if (job && job_id == job->id) { + kref_get(&job->refcount); + ret = core->ops->handle_request(job, data, len); + list_add(&job->node, &complete_node); + list_del(&apu_req->node); + ida_simple_remove(&req_ida, job->id); + kfree(apu_req); + drm_send_event(&job->apu->base, + &job->event->pending_event); + dma_fence_signal_locked(job->done_fence); + break; + } + } + spin_unlock_irqrestore(&core->ctx_lock, flags); + + return ret; +} +EXPORT_SYMBOL_GPL(apu_drm_callback); + +static void apu_job_cleanup(struct kref *ref) +{ + struct apu_job *job = container_of(ref, struct apu_job, + refcount); + unsigned int i; + + if (job->implicit_fences) { + for (i = 0; i < job->bo_count; i++) + dma_fence_put(job->implicit_fences[i]); + kvfree(job->implicit_fences); + } + dma_fence_put(job->done_fence); + dma_fence_put(job->render_done_fence); + + if (job->bos) { + for (i = 0; i < job->bo_count; i++) { + struct apu_gem_object *apu_obj; + + apu_obj = to_apu_bo(job->bos[i]); + drm_gem_object_put(job->bos[i]); + } + + kvfree(job->bos); + } + + kfree(job->data_out); + kfree(job->data_in); + kfree(job); +} + +static void apu_job_put(struct apu_job *job) +{ + kref_put(&job->refcount, apu_job_cleanup); +} + +static void apu_acquire_object_fences(struct drm_gem_object **bos, + int bo_count, + struct dma_fence **implicit_fences) +{ + int i; + + for (i = 0; i < bo_count; i++) + dma_resv_get_singleton(bos[i]->resv, DMA_RESV_USAGE_KERNEL, + &implicit_fences[i]); +} + +static void apu_attach_object_fences(struct drm_gem_object **bos, + int bo_count, struct dma_fence *fence) +{ + int i; + + for (i = 0; i < bo_count; i++) { + dma_resv_reserve_fences(bos[i]->resv, 1); + dma_resv_add_fence(bos[i]->resv, fence, DMA_RESV_USAGE_KERNEL); + } +} + +static int apu_job_push(struct apu_job *job) +{ + struct drm_sched_entity *entity = &job->core->sched->sched_entity; + struct ww_acquire_ctx acquire_ctx; + int ret = 0; + int i; + + ret = drm_gem_lock_reservations(job->bos, job->bo_count, &acquire_ctx); + if (ret) + return ret; + + ret = drm_sched_job_init(&job->base, entity, NULL); + if (ret) + goto unlock; + + drm_sched_job_arm(&job->base); + job->render_done_fence = dma_fence_get(&job->base.s_fence->finished); + + kref_get(&job->refcount); /* put by scheduler job completion */ + apu_acquire_object_fences(job->bos, job->bo_count, + job->implicit_fences); + + drm_sched_entity_push_job(&job->base); + + apu_attach_object_fences(job->bos, job->bo_count, + job->render_done_fence); + + for (i = 0; i < job->bo_count; i++) + ret = drm_sched_job_add_implicit_dependencies(&job->base, job->bos[i], + true); +unlock: + drm_gem_unlock_reservations(job->bos, job->bo_count, &acquire_ctx); + + return ret; +} + +static const char *apu_fence_get_driver_name(struct dma_fence *fence) +{ + return "apu"; +} + +static const char *apu_fence_get_timeline_name(struct dma_fence *fence) +{ + return "apu-0"; +} + +static void apu_fence_release(struct dma_fence *f) +{ + kfree(f); +} + +static const struct dma_fence_ops apu_fence_ops = { + .get_driver_name = apu_fence_get_driver_name, + .get_timeline_name = apu_fence_get_timeline_name, + .release = apu_fence_release, +}; + +static struct dma_fence *apu_fence_create(struct apu_sched *sched) +{ + struct dma_fence *fence; + struct apu_queue_state *apu_queue = &sched->apu_queue; + + fence = kzalloc(sizeof(*fence), GFP_KERNEL); + if (!fence) + return ERR_PTR(-ENOMEM); + + dma_fence_init(fence, &apu_fence_ops, &sched->job_lock, + apu_queue->fence_context, apu_queue->seqno++); + + return fence; +} + +static struct apu_job *to_apu_job(struct drm_sched_job *sched_job) +{ + return container_of(sched_job, struct apu_job, base); +} + +static int apu_job_hw_submit(struct apu_job *job) +{ + int ret; + struct apu_core *core = job->core; + struct apu_request *apu_req; + unsigned long flags; + + ret = ida_simple_get(&req_ida, 0, 0xffff, GFP_KERNEL); + if (ret < 0) + return ret; + job->id = ret; + + ret = core->ops->alloc_prepare_request(job); + if (ret || !job->request_data || !job->request_len) { + ret = -ENOMEM; + goto err_ida_remove; + } + + apu_req = kzalloc(sizeof(*apu_req), GFP_KERNEL); + if (!apu_req) { + ret = -ENOMEM; + goto err_free_data; + } + + apu_req->job = job; + spin_lock_irqsave(&core->ctx_lock, flags); + list_add(&apu_req->node, &core->requests); + spin_unlock_irqrestore(&core->ctx_lock, flags); + ret = core->ops->send_request(job); + if (ret < 0) + goto err; + kfree(job->request_data); + + return 0; + +err: + list_del(&apu_req->node); + kfree(apu_req); +err_free_data: + kfree(job->request_data); +err_ida_remove: + ida_simple_remove(&req_ida, job->id); + + return ret; +} + +static struct dma_fence *apu_job_run(struct drm_sched_job *sched_job) +{ + struct apu_job *job = to_apu_job(sched_job); + struct dma_fence *fence = NULL; + + if (unlikely(job->base.s_fence->finished.error)) + return NULL; + + fence = apu_fence_create(job->core->sched); + if (IS_ERR(fence)) + return NULL; + + job->done_fence = dma_fence_get(fence); + + apu_job_hw_submit(job); + + return fence; +} + +static enum drm_gpu_sched_stat apu_job_timedout(struct drm_sched_job *sched_job) +{ + struct apu_request *apu_req, *tmp; + struct apu_job *job = to_apu_job(sched_job); + + if (dma_fence_is_signaled(job->done_fence)) + return DRM_GPU_SCHED_STAT_NOMINAL; + + list_for_each_entry_safe(apu_req, tmp, &job->core->requests, node) { + /* Remove the request and notify user about timeout */ + if (apu_req->job == job) { + kref_get(&job->refcount); + job->result = ETIMEDOUT; + list_add(&job->node, &complete_node); + list_del(&apu_req->node); + ida_simple_remove(&req_ida, job->id); + kfree(apu_req); + drm_send_event(&job->apu->base, + &job->event->pending_event); + dma_fence_signal_locked(job->done_fence); + } + } + + return DRM_GPU_SCHED_STAT_NOMINAL; +} + +static void apu_job_free(struct drm_sched_job *sched_job) +{ + struct apu_job *job = to_apu_job(sched_job); + + drm_sched_job_cleanup(sched_job); + + apu_job_put(job); +} + +static const struct drm_sched_backend_ops apu_sched_ops = { + .run_job = apu_job_run, + .timedout_job = apu_job_timedout, + .free_job = apu_job_free +}; + +int apu_drm_job_init(struct apu_core *core) +{ + int ret; + struct apu_sched *apu_sched; + struct drm_gpu_scheduler *sched; + + apu_sched = devm_kzalloc(core->dev, sizeof(*apu_sched), GFP_KERNEL); + if (!apu_sched) + return -ENOMEM; + + sched = &apu_sched->apu_queue.sched; + apu_sched->apu_queue.fence_context = dma_fence_context_alloc(1); + ret = drm_sched_init(sched, &apu_sched_ops, + 1, 0, msecs_to_jiffies(500), + NULL, NULL, "apu_js", core->dev); + if (ret) { + dev_err(core->dev, "Failed to create scheduler"); + return ret; + } + + ret = drm_sched_entity_init(&apu_sched->sched_entity, + DRM_SCHED_PRIORITY_NORMAL, + &sched, 1, NULL); + if (ret) { + dev_err(core->dev, "Failed to initialize scheduler entity"); + drm_sched_fini(&core->sched->apu_queue.sched); + return ret; + } + + core->sched = apu_sched; + apu_set_online(core); + + return 0; +} + +void apu_sched_fini(struct apu_core *core) +{ + apu_set_offline(core); + drm_sched_fini(&core->sched->apu_queue.sched); + devm_kfree(core->dev, core->sched); + core->sched = NULL; +} + +static struct apu_core *get_apu_core(struct apu_drm *apu, int device_id) +{ + struct apu_core *core; + + list_for_each_entry(core, &apu->cores, node) { + if (core->device_id == device_id) + return core; + } + + return NULL; +} + +static int apu_lookup_bos(struct drm_device *dev, struct drm_file *file_priv, + struct drm_apu_gem_queue *args, struct apu_job *job) +{ + void __user *bo_handles; + int ret; + + job->bo_count = args->bo_handle_count; + + if (!job->bo_count) + return 0; + + job->implicit_fences = kvmalloc_array(job->bo_count, + sizeof(struct dma_fence *), + GFP_KERNEL | __GFP_ZERO); + if (!job->implicit_fences) + return -ENOMEM; + + bo_handles = (void __user *)(uintptr_t) args->bo_handles; + ret = drm_gem_objects_lookup(file_priv, bo_handles, + job->bo_count, &job->bos); + + return ret; +} + +int ioctl_gem_queue(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct apu_drm *apu = dev->dev_private; + struct drm_apu_gem_queue *args = data; + struct apu_event *event; + struct apu_core *core; + struct drm_syncobj *sync_out = NULL; + struct apu_job *job; + int ret = 0; + + core = get_apu_core(apu, args->device); + if (!apu_core_is_running(core)) + return -ENODEV; + + if (args->out_sync > 0) { + sync_out = drm_syncobj_find(file_priv, args->out_sync); + if (!sync_out) + return -ENODEV; + } + + job = kzalloc(sizeof(*job), GFP_KERNEL); + if (!job) { + ret = -ENOMEM; + goto fail_out_sync; + } + + kref_init(&job->refcount); + + job->apu = apu; + job->core = core; + job->cmd = args->cmd; + job->size_in = args->size_in; + job->size_out = args->size_out; + job->sync_out = sync_out; + if (job->size_in) { + job->data_in = kmalloc(job->size_in, GFP_KERNEL); + if (!job->data_in) { + ret = -ENOMEM; + goto fail_job; + } + + ret = + copy_from_user(job->data_in, + (void __user *)(uintptr_t) args->data, + job->size_in); + if (ret) + goto fail_job; + } + + if (job->size_out) { + job->data_out = kmalloc(job->size_out, GFP_KERNEL); + if (!job->data_out) { + ret = -ENOMEM; + goto fail_job; + } + } + + ret = apu_lookup_bos(dev, file_priv, args, job); + if (ret) + goto fail_job; + + event = kzalloc(sizeof(*event), GFP_KERNEL); + event->base.length = sizeof(struct apu_job_event); + event->base.type = APU_JOB_COMPLETED; + event->job_event.out_sync = args->out_sync; + job->event = event; + ret = drm_event_reserve_init(dev, file_priv, &job->event->pending_event, + &job->event->base); + if (ret) + goto fail_job; + + ret = apu_job_push(job); + if (ret) { + drm_event_cancel_free(dev, &job->event->pending_event); + goto fail_job; + } + + if (sync_out) + drm_syncobj_replace_fence(sync_out, job->render_done_fence); + +fail_job: + apu_job_put(job); +fail_out_sync: + if (sync_out) + drm_syncobj_put(sync_out); + + return ret; +} + +int ioctl_gem_dequeue(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_apu_gem_dequeue *args = data; + struct drm_syncobj *sync_out = NULL; + struct apu_job *job; + int ret = 0; + + if (args->out_sync > 0) { + sync_out = drm_syncobj_find(file_priv, args->out_sync); + if (!sync_out) + return -ENODEV; + } + + list_for_each_entry(job, &complete_node, node) { + if (job->sync_out == sync_out) { + if (job->data_out) { + ret = copy_to_user((void __user *)(uintptr_t) + args->data, job->data_out, + job->size_out); + args->size = job->size_out; + } + args->result = job->result; + list_del(&job->node); + apu_job_put(job); + drm_syncobj_put(sync_out); + + return ret; + } + } + + if (sync_out) + drm_syncobj_put(sync_out); + + return 0; +} + +int ioctl_apu_state(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct apu_drm *apu = dev->dev_private; + struct drm_apu_state *args = data; + struct apu_core *core; + + args->flags = 0; + + core = get_apu_core(apu, args->device); + if (!core) + return -ENODEV; + + apu_core_update_state(core); + args->flags |= core->flags; + + return 0; +} diff --git a/include/uapi/drm/apu_drm.h b/include/uapi/drm/apu_drm.h index 14fc478f45dc..c47000097040 100644 --- a/include/uapi/drm/apu_drm.h +++ b/include/uapi/drm/apu_drm.h @@ -9,6 +9,8 @@ extern "C" { #endif

+#define APU_JOB_COMPLETED 0x80000000 + /* * Please note that modifications to all structs defined here are * subject to backwards-compatibility constraints. @@ -21,6 +23,29 @@ struct drm_apu_gem_new { __u64 offset; /* out */ };

+struct drm_apu_gem_queue { + __u32 device; + __u32 cmd; + __u32 out_sync; + __u64 bo_handles; + __u32 bo_handle_count; + __u16 size_in; + __u16 size_out; + __u64 data; +}; + +struct drm_apu_gem_dequeue { + __u32 out_sync; + __u16 result; + __u16 size; + __u64 data; +}; + +struct apu_job_event { + struct drm_event base; + __u32 out_sync; +}; + #define APU_ONLINE BIT(0)

struct drm_apu_state { @@ -30,10 +55,14 @@ struct drm_apu_state {

#define DRM_APU_STATE 0x00 #define DRM_APU_GEM_NEW 0x01 -#define DRM_APU_NUM_IOCTLS 0x02 +#define DRM_APU_GEM_QUEUE 0x02 +#define DRM_APU_GEM_DEQUEUE 0x03 +#define DRM_APU_NUM_IOCTLS 0x04

#define DRM_IOCTL_APU_STATE DRM_IOWR(DRM_COMMAND_BASE + DRM_APU_STATE, struct drm_apu_state) #define DRM_IOCTL_APU_GEM_NEW DRM_IOWR(DRM_COMMAND_BASE + DRM_APU_GEM_NEW, struct drm_apu_gem_new) +#define DRM_IOCTL_APU_GEM_QUEUE DRM_IOWR(DRM_COMMAND_BASE + DRM_APU_GEM_QUEUE, struct drm_apu_gem_queue) +#define DRM_IOCTL_APU_GEM_DEQUEUE DRM_IOWR(DRM_COMMAND_BASE + DRM_APU_GEM_DEQUEUE, struct drm_apu_gem_dequeue)

#if defined(__cplusplus) }

From: Julien Stephan jstephan@baylibre.com

This implements a driver to use with a simulation APU.

This is useful for testing purpose and can be used as a basis to implement real platform driver. Communication between the simulated APU and the driver is done using netlink socket.

Signed-off-by: Julien Stephan jstephan@baylibre.com --- drivers/gpu/drm/apu/Kconfig | 9 + drivers/gpu/drm/apu/Makefile | 3 + drivers/gpu/drm/apu/simu_apu.c | 313 +++++++++++++++++++++++++++++++++ 3 files changed, 325 insertions(+) create mode 100644 drivers/gpu/drm/apu/simu_apu.c

diff --git a/drivers/gpu/drm/apu/Kconfig b/drivers/gpu/drm/apu/Kconfig index a769df42091c..e0ffc166497c 100644 --- a/drivers/gpu/drm/apu/Kconfig +++ b/drivers/gpu/drm/apu/Kconfig @@ -11,3 +11,12 @@ config DRM_APU communicate with an AI Processor Unit (APU). The driver intends to provide a common infrastructure that may be used to support many different APU. + +config DRM_SIMU_APU + tristate "SIMULATION APU DRM driver" + depends on DRM_APU + default n + help + This provides a driver using netlink socket to communicate + with a simu APU. + This is useful for simulation and testing of libAPU stack. diff --git a/drivers/gpu/drm/apu/Makefile b/drivers/gpu/drm/apu/Makefile index fc8d6380fc38..0b007854a07f 100644 --- a/drivers/gpu/drm/apu/Makefile +++ b/drivers/gpu/drm/apu/Makefile @@ -4,4 +4,7 @@ drm_apu-y += apu_drv.o drm_apu-y += apu_gem.o drm_apu-y += apu_sched.o

+drm_simu_apu-y += simu_apu.o + obj-$(CONFIG_DRM_APU) += drm_apu.o +obj-$(CONFIG_DRM_SIMU_APU) += drm_simu_apu.o diff --git a/drivers/gpu/drm/apu/simu_apu.c b/drivers/gpu/drm/apu/simu_apu.c new file mode 100644 index 000000000000..5557f8b78a83 --- /dev/null +++ b/drivers/gpu/drm/apu/simu_apu.c @@ -0,0 +1,313 @@ +// SPDX-License-Identifier: GPL-2.0 +// +// Copyright 2023 BayLibre SAS + +#include <linux/module.h> +#include <linux/netlink.h> +#include <linux/platform_device.h> +#include <linux/skbuff.h> + +#include <net/sock.h> + +#include <drm/apu_drm.h> + +#include "apu_internal.h" + + +#define MYPROTO 17 +#define MYGRP 17 + +#define DRIVER_NAME "SIMU APU DRIVER" + +/* + * Firmware request, must be aligned with the one defined in firmware. + * @id: Request id, used in the case of reply, to find the pending request + * @cmd: The command id to execute in the firmware + * @result: The result of the command executed on the firmware + * @size: The size of the data available in this request + * @count: The number of shared buffer + * @data: Contains the data attached with the request if size is greater than + * zero, and the addresses of shared buffers if count is greater than + * zero. Both the data and the shared buffer could be read and write + * by the APU. + */ +struct apu_dev_request { + u16 id; + u16 cmd; + u16 result; + u16 size_in; + u16 size_out; + u16 count; + u8 data[0]; +} __packed; + +struct platform_device *platform; +struct apu_core *apu_core; +static int pid = -1; +struct sock *nl_sock; + +static int apu_netlink_read(struct sk_buff *skb, struct apu_dev_request **msg_ptr, int *pid) +{ + struct nlmsghdr *nlh; + + nlh = (struct nlmsghdr *)skb->data; + *pid = nlh->nlmsg_pid; /* pid of sending process */ + *msg_ptr = nlmsg_data(nlh); + + return nlh->nlmsg_len - NLMSG_HDRLEN; +} + +static int apu_netlink_write(void *msg_ptr, int msg_size, int pid) +{ + struct sk_buff *skb_out; + struct nlmsghdr *nlh; + int res; + + skb_out = nlmsg_new(msg_size, 0); + if (!skb_out) + return -ENOMEM; + + nlh = nlmsg_put(skb_out, 0, 0, NLMSG_DONE, msg_size, 0); + NETLINK_CB(skb_out).dst_group = 0; /* not in multicast group */ + memcpy(nlmsg_data(nlh), msg_ptr, msg_size); + + res = nlmsg_unicast(nl_sock, skb_out, pid); + + if (res < 0) + return res; + else + return nlh->nlmsg_len - NLMSG_HDRLEN; +} + +static void netlink_recv_msg(struct sk_buff *skb) +{ + int msg_size; + struct apu_dev_request *hdr; + int nlmsg_pid; + + msg_size = apu_netlink_read(skb, &hdr, &nlmsg_pid); + + if (pid == -1) { + // No device registered yet, the first message should be + // "READY" + if (!strncmp((char *)hdr, "READY", strlen("READY"))) { + + pid = nlmsg_pid; + if (apu_core_register(&platform->dev, apu_core, apu_core->apu)) + pr_err("cannot register SIMU APU\n"); + } + } else if (pid == nlmsg_pid) { + if (!strncmp((char *)hdr, "STOP", strlen("STOP"))) { + pid = -1; + apu_core_remove(apu_core); + } else + apu_drm_callback(apu_core, hdr->id, hdr, msg_size); + } else { + pr_err("%s: Only one core is supported for now\n", DRIVER_NAME); + } +} + +static int netlink_setup(void) +{ + int ret = 0; + struct netlink_kernel_cfg cfg = { + .input = netlink_recv_msg, + }; + + nl_sock = netlink_kernel_create(&init_net, MYPROTO, &cfg); + if (!nl_sock) + ret = -ENOMEM; + + return ret; +} + +static int simu_apu_send(struct apu_job *job) +{ + return apu_netlink_write((void *)(job->request_data), job->request_len, pid); +} + +static int simu_apu_handle_request(struct apu_job *job, void *data, int len) +{ + struct apu_dev_request *hdr = data; + + job->result = hdr->result; + if (job->size_out) + memcpy(job->data_out, hdr->data + job->size_in, + min(job->size_out, hdr->size_out)); + job->size_out = hdr->size_out; + return 0; +} + +static int simu_apu_alloc_request(struct apu_job *job) +{ + struct apu_dev_request *dev_req; + + int size; + u64 *dev_req_da; + u32 *dev_req_buffer_size; + int i; + + size = sizeof(*dev_req) + (sizeof(u64) + sizeof(u32)) * job->bo_count * 2 + + job->size_in + job->size_out; + dev_req = kmalloc(size, GFP_KERNEL); + if (!dev_req) + return -ENOMEM; + + dev_req->cmd = job->cmd; + dev_req->size_in = job->size_in; + dev_req->size_out = job->size_out; + dev_req->count = job->bo_count; + dev_req_da = + (u64 *) (dev_req->data + dev_req->size_in + dev_req->size_out); + dev_req_buffer_size = (u32 *) (dev_req_da + dev_req->count); + memcpy(dev_req->data, job->data_in, job->size_in); + + for (i = 0; i < job->bo_count; i++) { + struct apu_gem_object *obj = to_apu_bo(job->bos[i]); + + dev_req_da[i] = drm_vma_node_offset_addr(&obj->base.base.vma_node); + dev_req_buffer_size[i] = obj->size; + } + + dev_req->id = job->id; + + job->request_data = dev_req; + job->request_len = size; + return 0; +} + +static int simu_apu_ready(struct apu_core *core) +{ + if (pid == -1) + return 0; + + return 1; +} + +/** + * simu_apu_gem_mmap + * + * this is directly based on drm_gem_mmap() function but removing the permission + * check before mapping a buffer. This is useful here to be able to easily + * share buffers between libapu host application and libapu device application + * (simulation use case) + * + */ +static int simu_apu_gem_mmap(struct file *filp, struct vm_area_struct *vma) +{ + struct drm_file *priv = filp->private_data; + struct drm_device *dev = priv->minor->dev; + struct drm_gem_object *obj = NULL; + struct drm_vma_offset_node *node; + int ret; + + if (drm_dev_is_unplugged(dev)) + return -ENODEV; + + drm_vma_offset_lock_lookup(dev->vma_offset_manager); + node = drm_vma_offset_exact_lookup_locked(dev->vma_offset_manager, + vma->vm_pgoff, + vma_pages(vma)); + if (likely(node)) { + obj = container_of(node, struct drm_gem_object, vma_node); + /* + * When the object is being freed, after it hits 0-refcnt it + * proceeds to tear down the object. In the process it will + * attempt to remove the VMA offset and so acquire this + * mgr->vm_lock. Therefore if we find an object with a 0-refcnt + * that matches our range, we know it is in the process of being + * destroyed and will be freed as soon as we release the lock - + * so we have to check for the 0-refcnted object and treat it as + * invalid. + */ + if (!kref_get_unless_zero(&obj->refcount)) { + obj = NULL; + pr_err("DTC: %s: %d\n", __func__, __LINE__); + } + } + drm_vma_offset_unlock_lookup(dev->vma_offset_manager); + + if (!obj) + return -EINVAL; + + ret = drm_gem_mmap_obj(obj, drm_vma_node_size(node) << PAGE_SHIFT, + vma); + + drm_gem_object_put(obj); + + return ret; +} + +static struct apu_core_ops simu_apu_ops = { + .alloc_prepare_request = simu_apu_alloc_request, + .send_request = simu_apu_send, + .handle_request = simu_apu_handle_request, + .is_ready = simu_apu_ready, +}; + +static int __init apu_platform_init(void) +{ + int ret; + struct apu_drm *apu; + + platform = platform_device_register_simple(DRIVER_NAME, -1, NULL, 0); + if (IS_ERR(platform)) + return PTR_ERR(platform); + + if (!devres_open_group(&platform->dev, NULL, GFP_KERNEL)) { + ret = -ENOMEM; + goto out_unregister; + } + + apu = apu_dev_alloc(&platform->dev); + if (!apu) { + ret = -ENOMEM; + goto out_devres; + } + + apu_core = apu_core_alloc(apu, &simu_apu_ops, apu); + if (!apu_core) { + ret = -ENOMEM; + goto out_devres; + } + + ret = apu_dev_register(apu); + if (ret) + goto out_apu_core_free; + + apu->mmap = simu_apu_gem_mmap; + + ret = netlink_setup(); + if (ret) + goto out_apu_dev_unregister; + + return 0; + +out_apu_dev_unregister: + apu_dev_unregister(apu); +out_apu_core_free: + apu_core_free(apu_core); +out_devres: + devres_release_group(&platform->dev, NULL); +out_unregister: + platform_device_unregister(platform); + return ret; +} + +static void __exit apu_platform_exit(void) +{ + netlink_kernel_release(nl_sock); + apu_core_remove(apu_core); + apu_core_free(apu_core); + apu_dev_unregister((struct apu_drm *)apu_core->apu); + devres_release_group(&platform->dev, NULL); + platform_device_unregister(platform); +} + + +module_init(apu_platform_init); +module_exit(apu_platform_exit); + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Julien Stephan"); +MODULE_DESCRIPTION(DRIVER_NAME);