140 lines · plain
1config DRM_I915_REQUEST_TIMEOUT2 int "Default timeout for requests (ms)"3 default 20000 # milliseconds4 help5 Configures the default timeout after which any user submissions will6 be forcefully terminated.7 8 Beware setting this value lower, or close to heartbeat interval9 rounded to whole seconds times three, in order to avoid allowing10 misbehaving applications causing total rendering failure in unrelated11 clients.12 13 May be 0 to disable the timeout.14 15config DRM_I915_FENCE_TIMEOUT16 int "Timeout for unsignaled foreign fences (ms, jiffy granularity)"17 default 10000 # milliseconds18 help19 When listening to a foreign fence, we install a supplementary timer20 to ensure that we are always signaled and our userspace is able to21 make forward progress. This value specifies the timeout used for an22 unsignaled foreign fence.23 24 May be 0 to disable the timeout, and rely on the foreign fence being25 eventually signaled.26 27config DRM_I915_USERFAULT_AUTOSUSPEND28 int "Runtime autosuspend delay for userspace GGTT mmaps (ms)"29 default 250 # milliseconds30 help31 On runtime suspend, as we suspend the device, we have to revoke32 userspace GGTT mmaps and force userspace to take a pagefault on33 their next access. The revocation and subsequent recreation of34 the GGTT mmap can be very slow and so we impose a small hysteris35 that complements the runtime-pm autosuspend and provides a lower36 floor on the autosuspend delay.37 38 May be 0 to disable the extra delay and solely use the device level39 runtime pm autosuspend delay tunable.40 41config DRM_I915_HEARTBEAT_INTERVAL42 int "Interval between heartbeat pulses (ms)"43 default 2500 # milliseconds44 help45 The driver sends a periodic heartbeat down all active engines to46 check the health of the GPU and undertake regular house-keeping of47 internal driver state.48 49 This is adjustable via50 /sys/class/drm/card?/engine/*/heartbeat_interval_ms51 52 May be 0 to disable heartbeats and therefore disable automatic GPU53 hang detection.54 55config DRM_I915_PREEMPT_TIMEOUT56 int "Preempt timeout (ms, jiffy granularity)"57 default 640 # milliseconds58 help59 How long to wait (in milliseconds) for a preemption event to occur60 when submitting a new context. If the current context does not hit61 an arbitration point and yield to HW before the timer expires, the62 HW will be reset to allow the more important context to execute.63 64 This is adjustable via65 /sys/class/drm/card?/engine/*/preempt_timeout_ms66 67 May be 0 to disable the timeout.68 69 The compiled in default may get overridden at driver probe time on70 certain platforms and certain engines which will be reflected in the71 sysfs control.72 73config DRM_I915_PREEMPT_TIMEOUT_COMPUTE74 int "Preempt timeout for compute engines (ms, jiffy granularity)"75 default 7500 # milliseconds76 help77 How long to wait (in milliseconds) for a preemption event to occur78 when submitting a new context to a compute capable engine. If the79 current context does not hit an arbitration point and yield to HW80 before the timer expires, the HW will be reset to allow the more81 important context to execute.82 83 This is adjustable via84 /sys/class/drm/card?/engine/*/preempt_timeout_ms85 86 May be 0 to disable the timeout.87 88 The compiled in default may get overridden at driver probe time on89 certain platforms and certain engines which will be reflected in the90 sysfs control.91 92config DRM_I915_MAX_REQUEST_BUSYWAIT93 int "Busywait for request completion limit (ns)"94 default 8000 # nanoseconds95 help96 Before sleeping waiting for a request (GPU operation) to complete,97 we may spend some time polling for its completion. As the IRQ may98 take a non-negligible time to setup, we do a short spin first to99 check if the request will complete in the time it would have taken100 us to enable the interrupt.101 102 This is adjustable via103 /sys/class/drm/card?/engine/*/max_busywait_duration_ns104 105 May be 0 to disable the initial spin. In practice, we estimate106 the cost of enabling the interrupt (if currently disabled) to be107 a few microseconds.108 109config DRM_I915_STOP_TIMEOUT110 int "How long to wait for an engine to quiesce gracefully before reset (ms)"111 default 100 # milliseconds112 help113 By stopping submission and sleeping for a short time before resetting114 the GPU, we allow the innocent contexts also on the system to quiesce.115 It is then less likely for a hanging context to cause collateral116 damage as the system is reset in order to recover. The corollary is117 that the reset itself may take longer and so be more disruptive to118 interactive or low latency workloads.119 120 This is adjustable via121 /sys/class/drm/card?/engine/*/stop_timeout_ms122 123config DRM_I915_TIMESLICE_DURATION124 int "Scheduling quantum for userspace batches (ms, jiffy granularity)"125 default 1 # milliseconds126 help127 When two user batches of equal priority are executing, we will128 alternate execution of each batch to ensure forward progress of129 all users. This is necessary in some cases where there may be130 an implicit dependency between those batches that requires131 concurrent execution in order for them to proceed, e.g. they132 interact with each other via userspace semaphores. Each context133 is scheduled for execution for the timeslice duration, before134 switching to the next context.135 136 This is adjustable via137 /sys/class/drm/card?/engine/*/timeslice_duration_ms138 139 May be 0 to disable timeslicing.140