OpenCloudOS-Kernel/include/drm/gpu_scheduler.h

218 lines
6.9 KiB
C
Raw Normal View History

/*
* Copyright 2015 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
#ifndef _DRM_GPU_SCHEDULER_H_
#define _DRM_GPU_SCHEDULER_H_
#include <drm/spsc_queue.h>
#include <linux/dma-fence.h>
struct drm_gpu_scheduler;
struct drm_sched_rq;
enum drm_sched_priority {
DRM_SCHED_PRIORITY_MIN,
DRM_SCHED_PRIORITY_LOW = DRM_SCHED_PRIORITY_MIN,
DRM_SCHED_PRIORITY_NORMAL,
DRM_SCHED_PRIORITY_HIGH_SW,
DRM_SCHED_PRIORITY_HIGH_HW,
DRM_SCHED_PRIORITY_KERNEL,
DRM_SCHED_PRIORITY_MAX,
DRM_SCHED_PRIORITY_INVALID = -1,
DRM_SCHED_PRIORITY_UNSET = -2
};
/**
* drm_sched_entity - A wrapper around a job queue (typically attached
* to the DRM file_priv).
*
* Entities will emit jobs in order to their corresponding hardware
* ring, and the scheduler will alternate between entities based on
* scheduling policy.
*/
struct drm_sched_entity {
struct list_head list;
struct drm_sched_rq *rq;
spinlock_t rq_lock;
struct drm_gpu_scheduler *sched;
struct spsc_queue job_queue;
atomic_t fence_seq;
uint64_t fence_context;
struct dma_fence *dependency;
struct dma_fence_cb cb;
atomic_t *guilty; /* points to ctx's guilty */
drm/gpu-sched: fix force APP kill hang(v4) issue: there are VMC page fault occurred if force APP kill during 3dmark test, the cause is in entity_fini we manually signal all those jobs in entity's queue which confuse the sync/dep mechanism: 1)page fault occurred in sdma's clear job which operate on shadow buffer, and shadow buffer's Gart table is cleaned by ttm_bo_release since the fence in its reservation was fake signaled by entity_fini() under the case of SIGKILL received. 2)page fault occurred in gfx' job because during the lifetime of gfx job we manually fake signal all jobs from its entity in entity_fini(), thus the unmapping/clear PTE job depend on those result fence is satisfied and sdma start clearing the PTE and lead to GFX page fault. fix: 1)should at least wait all jobs already scheduled complete in entity_fini() if SIGKILL is the case. 2)if a fence signaled and try to clear some entity's dependency, should set this entity guilty to prevent its job really run since the dependency is fake signaled. v2: splitting drm_sched_entity_fini() into two functions: 1)The first one is does the waiting, removes the entity from the runqueue and returns an error when the process was killed. 2)The second one then goes over the entity, install it as completion signal for the remaining jobs and signals all jobs with an error code. v3: 1)Replace the fini1 and fini2 with better name 2)Call the first part before the VM teardown in amdgpu_driver_postclose_kms() and the second part after the VM teardown 3)Keep the original function drm_sched_entity_fini to refine the code. v4: 1)Rename entity->finished to entity->last_scheduled; 2)Rename drm_sched_entity_fini_job_cb() to drm_sched_entity_kill_jobs_cb(); 3)Pass NULL to drm_sched_entity_fini_job_cb() if -ENOENT; 4)Replace the type of entity->fini_status with "int"; 5)Remove the check about entity->finished. Signed-off-by: Monk Liu <Monk.Liu@amd.com> Signed-off-by: Emily Deng <Emily.Deng@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2018-04-16 10:07:02 +08:00
int fini_status;
struct dma_fence *last_scheduled;
};
/**
* Run queue is a set of entities scheduling command submissions for
* one specific ring. It implements the scheduling policy that selects
* the next entity to emit commands from.
*/
struct drm_sched_rq {
spinlock_t lock;
struct list_head entities;
struct drm_sched_entity *current_entity;
};
struct drm_sched_fence {
struct dma_fence scheduled;
/* This fence is what will be signaled by the scheduler when
* the job is completed.
*
* When setting up an out fence for the job, you should use
* this, since it's available immediately upon
* drm_sched_job_init(), and the fence returned by the driver
* from run_job() won't be created until the dependencies have
* resolved.
*/
struct dma_fence finished;
struct dma_fence_cb cb;
struct dma_fence *parent;
struct drm_gpu_scheduler *sched;
spinlock_t lock;
void *owner;
};
struct drm_sched_fence *to_drm_sched_fence(struct dma_fence *f);
/**
* drm_sched_job - A job to be run by an entity.
*
* A job is created by the driver using drm_sched_job_init(), and
* should call drm_sched_entity_push_job() once it wants the scheduler
* to schedule the job.
*/
struct drm_sched_job {
struct spsc_node queue_node;
struct drm_gpu_scheduler *sched;
struct drm_sched_fence *s_fence;
struct dma_fence_cb finish_cb;
struct work_struct finish_work;
struct list_head node;
struct delayed_work work_tdr;
uint64_t id;
atomic_t karma;
enum drm_sched_priority s_priority;
drm/gpu-sched: fix force APP kill hang(v4) issue: there are VMC page fault occurred if force APP kill during 3dmark test, the cause is in entity_fini we manually signal all those jobs in entity's queue which confuse the sync/dep mechanism: 1)page fault occurred in sdma's clear job which operate on shadow buffer, and shadow buffer's Gart table is cleaned by ttm_bo_release since the fence in its reservation was fake signaled by entity_fini() under the case of SIGKILL received. 2)page fault occurred in gfx' job because during the lifetime of gfx job we manually fake signal all jobs from its entity in entity_fini(), thus the unmapping/clear PTE job depend on those result fence is satisfied and sdma start clearing the PTE and lead to GFX page fault. fix: 1)should at least wait all jobs already scheduled complete in entity_fini() if SIGKILL is the case. 2)if a fence signaled and try to clear some entity's dependency, should set this entity guilty to prevent its job really run since the dependency is fake signaled. v2: splitting drm_sched_entity_fini() into two functions: 1)The first one is does the waiting, removes the entity from the runqueue and returns an error when the process was killed. 2)The second one then goes over the entity, install it as completion signal for the remaining jobs and signals all jobs with an error code. v3: 1)Replace the fini1 and fini2 with better name 2)Call the first part before the VM teardown in amdgpu_driver_postclose_kms() and the second part after the VM teardown 3)Keep the original function drm_sched_entity_fini to refine the code. v4: 1)Rename entity->finished to entity->last_scheduled; 2)Rename drm_sched_entity_fini_job_cb() to drm_sched_entity_kill_jobs_cb(); 3)Pass NULL to drm_sched_entity_fini_job_cb() if -ENOENT; 4)Replace the type of entity->fini_status with "int"; 5)Remove the check about entity->finished. Signed-off-by: Monk Liu <Monk.Liu@amd.com> Signed-off-by: Emily Deng <Emily.Deng@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2018-04-16 10:07:02 +08:00
struct drm_sched_entity *entity;
};
static inline bool drm_sched_invalidate_job(struct drm_sched_job *s_job,
int threshold)
{
return (s_job && atomic_inc_return(&s_job->karma) > threshold);
}
/**
* Define the backend operations called by the scheduler,
* these functions should be implemented in driver side
*/
struct drm_sched_backend_ops {
/* Called when the scheduler is considering scheduling this
* job next, to get another struct dma_fence for this job to
* block on. Once it returns NULL, run_job() may be called.
*/
struct dma_fence *(*dependency)(struct drm_sched_job *sched_job,
struct drm_sched_entity *s_entity);
/* Called to execute the job once all of the dependencies have
* been resolved. This may be called multiple times, if
* timedout_job() has happened and drm_sched_job_recovery()
* decides to try it again.
*/
struct dma_fence *(*run_job)(struct drm_sched_job *sched_job);
/* Called when a job has taken too long to execute, to trigger
* GPU recovery.
*/
void (*timedout_job)(struct drm_sched_job *sched_job);
/* Called once the job's finished fence has been signaled and
* it's time to clean it up.
*/
void (*free_job)(struct drm_sched_job *sched_job);
};
/**
* One scheduler is implemented for each hardware ring
*/
struct drm_gpu_scheduler {
const struct drm_sched_backend_ops *ops;
uint32_t hw_submission_limit;
long timeout;
const char *name;
struct drm_sched_rq sched_rq[DRM_SCHED_PRIORITY_MAX];
wait_queue_head_t wake_up_worker;
wait_queue_head_t job_scheduled;
atomic_t hw_rq_count;
atomic64_t job_id_count;
struct task_struct *thread;
struct list_head ring_mirror_list;
spinlock_t job_list_lock;
int hang_limit;
};
int drm_sched_init(struct drm_gpu_scheduler *sched,
const struct drm_sched_backend_ops *ops,
uint32_t hw_submission, unsigned hang_limit, long timeout,
const char *name);
void drm_sched_fini(struct drm_gpu_scheduler *sched);
int drm_sched_entity_init(struct drm_gpu_scheduler *sched,
struct drm_sched_entity *entity,
struct drm_sched_rq *rq,
atomic_t *guilty);
drm/gpu-sched: fix force APP kill hang(v4) issue: there are VMC page fault occurred if force APP kill during 3dmark test, the cause is in entity_fini we manually signal all those jobs in entity's queue which confuse the sync/dep mechanism: 1)page fault occurred in sdma's clear job which operate on shadow buffer, and shadow buffer's Gart table is cleaned by ttm_bo_release since the fence in its reservation was fake signaled by entity_fini() under the case of SIGKILL received. 2)page fault occurred in gfx' job because during the lifetime of gfx job we manually fake signal all jobs from its entity in entity_fini(), thus the unmapping/clear PTE job depend on those result fence is satisfied and sdma start clearing the PTE and lead to GFX page fault. fix: 1)should at least wait all jobs already scheduled complete in entity_fini() if SIGKILL is the case. 2)if a fence signaled and try to clear some entity's dependency, should set this entity guilty to prevent its job really run since the dependency is fake signaled. v2: splitting drm_sched_entity_fini() into two functions: 1)The first one is does the waiting, removes the entity from the runqueue and returns an error when the process was killed. 2)The second one then goes over the entity, install it as completion signal for the remaining jobs and signals all jobs with an error code. v3: 1)Replace the fini1 and fini2 with better name 2)Call the first part before the VM teardown in amdgpu_driver_postclose_kms() and the second part after the VM teardown 3)Keep the original function drm_sched_entity_fini to refine the code. v4: 1)Rename entity->finished to entity->last_scheduled; 2)Rename drm_sched_entity_fini_job_cb() to drm_sched_entity_kill_jobs_cb(); 3)Pass NULL to drm_sched_entity_fini_job_cb() if -ENOENT; 4)Replace the type of entity->fini_status with "int"; 5)Remove the check about entity->finished. Signed-off-by: Monk Liu <Monk.Liu@amd.com> Signed-off-by: Emily Deng <Emily.Deng@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2018-04-16 10:07:02 +08:00
void drm_sched_entity_do_release(struct drm_gpu_scheduler *sched,
struct drm_sched_entity *entity);
void drm_sched_entity_cleanup(struct drm_gpu_scheduler *sched,
struct drm_sched_entity *entity);
void drm_sched_entity_fini(struct drm_gpu_scheduler *sched,
struct drm_sched_entity *entity);
void drm_sched_entity_push_job(struct drm_sched_job *sched_job,
struct drm_sched_entity *entity);
void drm_sched_entity_set_rq(struct drm_sched_entity *entity,
struct drm_sched_rq *rq);
struct drm_sched_fence *drm_sched_fence_create(
struct drm_sched_entity *s_entity, void *owner);
void drm_sched_fence_scheduled(struct drm_sched_fence *fence);
void drm_sched_fence_finished(struct drm_sched_fence *fence);
int drm_sched_job_init(struct drm_sched_job *job,
struct drm_gpu_scheduler *sched,
struct drm_sched_entity *entity,
void *owner);
void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched,
struct drm_sched_job *job);
void drm_sched_job_recovery(struct drm_gpu_scheduler *sched);
bool drm_sched_dependency_optimized(struct dma_fence* fence,
struct drm_sched_entity *entity);
void drm_sched_job_kickout(struct drm_sched_job *s_job);
#endif