STRUCT DRM_GPU_SCHEDULER(9)

struct drm_gpu_scheduler(9) 💖

okay so bestie, this one is from the Kernel API Manual and it is literally the blueprint for a scheduler instance’s very own data. no cap, it’s giving “one struct to hold all the scheduler things.” main character energy fr fr. 🎀

---

🌸 NAME 🌸

struct drm_gpu_scheduler — scheduler instance-specific data. that’s the whole name vibe, and yes, every little field matters. 💅

---

✨ SYNOPSIS ✨

here’s the struct in all its glory — copy it exactly, bestie, no “fixing” the spacing or renaming anything:

       struct drm_gpu_scheduler {
           const struct drm_sched_backend_ops      *ops;
           u32 credit_limit;
           atomic_t credit_count;
           long timeout;
           const char                      *name;
           u32 num_rqs;
           u32 num_user_rqs;
           struct drm_sched_rq             **sched_rq;
           wait_queue_head_t job_scheduled;
           atomic64_t job_id_count;
           struct workqueue_struct         *submit_wq;
           struct workqueue_struct         *timeout_wq;
           struct ewma_drm_sched_avgtime   avg_job_us;
           struct work_struct              work_run_job;
           struct work_struct              work_free_job;
           struct delayed_work             work_tdr;
           struct list_head                pending_list;
           spinlock_t job_list_lock;
           int hang_limit;
           atomic_t *score;
           atomic_t _score;
           bool ready;
           bool free_guilty;
           bool pause_submit;
           bool own_submit_wq;
           struct device                   *dev;
        };

---

🫶 Members 🫶

• ops — backend operations provided by the driver. the driver’s lil’ checklist of backend behavior, basically. 💖
• credit_limit — the credit limit of this scheduler. 🔥
• credit_count — the current credit count of this scheduler. ⭐
• timeout — the time after which a job is removed from the scheduler. yes, the clock is ticking, bestie. 😭
• name — name of the ring for which this scheduler is being used. iconic little name tag energy. 🎀
• num_rqs — Equal to num_user_rqs for FIFO and RR and 1 for the FAIR policy. periodt. 💅
• num_user_rqs — Number of run-queues. This is at most DRM_SCHED_PRIORITY_COUNT, as there’s usually one run-queue per priority, but could be less. so close to the max, but not always. 👑
• sched_rq — An allocated array of run-queues of size num_rqs; yes, that semicolon is in the building. 🦋
• job_scheduled — once drm_sched_entity_flush is called the scheduler waits on this wait queue until all the scheduled jobs are finished. the “wait for everybody to finish” queue, fr fr. 🌸
• job_id_count — used to assign unique id to the each job. every job gets its own little id moment. ✨
• submit_wq — workqueue used to queue work_run_job and work_free_job. 🧚
• timeout_wq — workqueue used to queue work_tdr. ⏰
• avg_job_us — Average job duration. the literal average-job-time icon. 📊
• work_run_job — work which calls run_job op of each scheduler. go off, run that job. 🔥
• work_free_job — work which calls free_job op of each scheduler. cleanup era. 🧹
• work_tdr — schedules a delayed call to drm_sched_job_timedout after the timeout interval is over. the dramatic delayed entrance, honestly. 💀
• pending_list — the list of jobs which are currently in the job queue. the “currently pending” clique. 💜
• job_list_lock — lock to protect the pending_list. protect the list at all costs. 🛡️
• hang_limit — once the hangs by a job crosses this limit then it is marked guilty and it will no longer be considered for scheduling. not the guilty verdict 😭 — it’s no longer considered for scheduling, periodt.
• score — score to help loadbalancer pick a idle sched. yes, “a idle sched” — we’re keeping the details, bestie. 🌈
• _score — score used when the driver doesn’t provide one. the backup-score main character. ⭐
• ready — marks if the underlying HW is ready to work. vibe check for the hardware. ✅
• free_guilty — A hit to time out handler to free the guilty job. 🫶
• pause_submit — pause queuing of work_run_job on submit_wq. hold up, bestie — pause the queuing. ✋
• own_submit_wq — scheduler owns allocation of submit_wq. it’s not borrowing it; it owns it. 👑
• dev — system struct device. the device link-up, fr fr. 🔌

---

✨ Description ✨

One scheduler is implemented for each hardware ring. that’s the big-picture tea: one scheduler per hardware ring, no cap. 🌀

---

👑 SEE ALSO 👑

• Kernel file ./include/drm/gpu_scheduler.h
• DRM_SCHED_FENCE_DONT_PIPELINE(9)
• DRM_SCHED_FENCE_FLAG_HAS_DEADLINE_BIT(9)
• struct drm_sched_entity(9)
• struct drm_sched_rq(9)
• struct drm_sched_fence(9)
• struct drm_sched_job(9)
• enum drm_gpu_sched_stat(9)
• struct drm_sched_backend_ops(9)
• struct drm_sched_init_args(9)
• struct drm_sched_pending_job_iter(9)
• drm_sched_for_each_pending_job(9)

---

mwah 💋 source/version: ./include/drm, date: October 2026. that’s the cute lil’ sign-off, bestie — now go read the whole scheduler struct like the icon you are. 🫶✨