1config DRM_I915_FENCE_TIMEOUT 2 int "Timeout for unsignaled foreign fences (ms, jiffy granularity)" 3 default 10000 # milliseconds 4 help 5 When listening to a foreign fence, we install a supplementary timer 6 to ensure that we are always signaled and our userspace is able to 7 make forward progress. This value specifies the timeout used for an 8 unsignaled foreign fence. 9 10 May be 0 to disable the timeout, and rely on the foreign fence being 11 eventually signaled. 12 13config DRM_I915_USERFAULT_AUTOSUSPEND 14 int "Runtime autosuspend delay for userspace GGTT mmaps (ms)" 15 default 250 # milliseconds 16 help 17 On runtime suspend, as we suspend the device, we have to revoke 18 userspace GGTT mmaps and force userspace to take a pagefault on 19 their next access. The revocation and subsequent recreation of 20 the GGTT mmap can be very slow and so we impose a small hysteris 21 that complements the runtime-pm autosuspend and provides a lower 22 floor on the autosuspend delay. 23 24 May be 0 to disable the extra delay and solely use the device level 25 runtime pm autosuspend delay tunable. 26 27config DRM_I915_HEARTBEAT_INTERVAL 28 int "Interval between heartbeat pulses (ms)" 29 default 2500 # milliseconds 30 help 31 The driver sends a periodic heartbeat down all active engines to 32 check the health of the GPU and undertake regular house-keeping of 33 internal driver state. 34 35 This is adjustable via 36 /sys/class/drm/card?/engine/*/heartbeat_interval_ms 37 38 May be 0 to disable heartbeats and therefore disable automatic GPU 39 hang detection. 40 41config DRM_I915_PREEMPT_TIMEOUT 42 int "Preempt timeout (ms, jiffy granularity)" 43 default 640 # milliseconds 44 help 45 How long to wait (in milliseconds) for a preemption event to occur 46 when submitting a new context via execlists. If the current context 47 does not hit an arbitration point and yield to HW before the timer 48 expires, the HW will be reset to allow the more important context 49 to execute. 50 51 This is adjustable via 52 /sys/class/drm/card?/engine/*/preempt_timeout_ms 53 54 May be 0 to disable the timeout. 55 56 The compiled in default may get overridden at driver probe time on 57 certain platforms and certain engines which will be reflected in the 58 sysfs control. 59 60config DRM_I915_MAX_REQUEST_BUSYWAIT 61 int "Busywait for request completion limit (ns)" 62 default 8000 # nanoseconds 63 help 64 Before sleeping waiting for a request (GPU operation) to complete, 65 we may spend some time polling for its completion. As the IRQ may 66 take a non-negligible time to setup, we do a short spin first to 67 check if the request will complete in the time it would have taken 68 us to enable the interrupt. 69 70 This is adjustable via 71 /sys/class/drm/card?/engine/*/max_busywait_duration_ns 72 73 May be 0 to disable the initial spin. In practice, we estimate 74 the cost of enabling the interrupt (if currently disabled) to be 75 a few microseconds. 76 77config DRM_I915_STOP_TIMEOUT 78 int "How long to wait for an engine to quiesce gracefully before reset (ms)" 79 default 100 # milliseconds 80 help 81 By stopping submission and sleeping for a short time before resetting 82 the GPU, we allow the innocent contexts also on the system to quiesce. 83 It is then less likely for a hanging context to cause collateral 84 damage as the system is reset in order to recover. The corollary is 85 that the reset itself may take longer and so be more disruptive to 86 interactive or low latency workloads. 87 88 This is adjustable via 89 /sys/class/drm/card?/engine/*/stop_timeout_ms 90 91config DRM_I915_TIMESLICE_DURATION 92 int "Scheduling quantum for userspace batches (ms, jiffy granularity)" 93 default 1 # milliseconds 94 help 95 When two user batches of equal priority are executing, we will 96 alternate execution of each batch to ensure forward progress of 97 all users. This is necessary in some cases where there may be 98 an implicit dependency between those batches that requires 99 concurrent execution in order for them to proceed, e.g. they 100 interact with each other via userspace semaphores. Each context 101 is scheduled for execution for the timeslice duration, before 102 switching to the next context. 103 104 This is adjustable via 105 /sys/class/drm/card?/engine/*/timeslice_duration_ms 106 107 May be 0 to disable timeslicing. 108