diff options
Diffstat (limited to 'drivers/gpu/drm/i915/i915_request.h')
-rw-r--r-- | drivers/gpu/drm/i915/i915_request.h | 397 |
1 files changed, 344 insertions, 53 deletions
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h index edbbdfec24ab..28b1f9db5487 100644 --- a/drivers/gpu/drm/i915/i915_request.h +++ b/drivers/gpu/drm/i915/i915_request.h @@ -26,28 +26,48 @@ #define I915_REQUEST_H #include <linux/dma-fence.h> +#include <linux/hrtimer.h> +#include <linux/irq_work.h> +#include <linux/llist.h> #include <linux/lockdep.h> +#include "gem/i915_gem_context_types.h" +#include "gt/intel_context_types.h" #include "gt/intel_engine_types.h" +#include "gt/intel_timeline_types.h" #include "i915_gem.h" #include "i915_scheduler.h" #include "i915_selftest.h" #include "i915_sw_fence.h" +#include "i915_vma_resource.h" #include <uapi/drm/i915_drm.h> struct drm_file; struct drm_i915_gem_object; +struct drm_printer; +struct i915_deps; struct i915_request; -struct i915_timeline; -struct i915_timeline_cacheline; +#if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR) struct i915_capture_list { + struct i915_vma_resource *vma_res; struct i915_capture_list *next; - struct i915_vma *vma; }; +void i915_request_free_capture_list(struct i915_capture_list *capture); +#else +#define i915_request_free_capture_list(_a) do {} while (0) +#endif + +#define RQ_TRACE(rq, fmt, ...) do { \ + const struct i915_request *rq__ = (rq); \ + ENGINE_TRACE(rq__->engine, "fence %llx:%lld, current %d " fmt, \ + rq__->fence.context, rq__->fence.seqno, \ + hwsp_seqno(rq__), ##__VA_ARGS__); \ +} while (0) + enum { /* * I915_FENCE_FLAG_ACTIVE - this request is currently submitted to HW. @@ -63,12 +83,93 @@ enum { I915_FENCE_FLAG_ACTIVE = DMA_FENCE_FLAG_USER_BITS, /* + * I915_FENCE_FLAG_PQUEUE - this request is ready for execution + * + * Using the scheduler, when a request is ready for execution it is put + * into the priority queue, and removed from that queue when transferred + * to the HW runlists. We want to track its membership within the + * priority queue so that we can easily check before rescheduling. + * + * See i915_request_in_priority_queue() + */ + I915_FENCE_FLAG_PQUEUE, + + /* + * I915_FENCE_FLAG_HOLD - this request is currently on hold + * + * This request has been suspended, pending an ongoing investigation. + */ + I915_FENCE_FLAG_HOLD, + + /* + * I915_FENCE_FLAG_INITIAL_BREADCRUMB - this request has the initial + * breadcrumb that marks the end of semaphore waits and start of the + * user payload. + */ + I915_FENCE_FLAG_INITIAL_BREADCRUMB, + + /* * I915_FENCE_FLAG_SIGNAL - this request is currently on signal_list * * Internal bookkeeping used by the breadcrumb code to track when * a request is on the various signal_list. */ I915_FENCE_FLAG_SIGNAL, + + /* + * I915_FENCE_FLAG_NOPREEMPT - this request should not be preempted + * + * The execution of some requests should not be interrupted. This is + * a sensitive operation as it makes the request super important, + * blocking other higher priority work. Abuse of this flag will + * lead to quality of service issues. + */ + I915_FENCE_FLAG_NOPREEMPT, + + /* + * I915_FENCE_FLAG_SENTINEL - this request should be last in the queue + * + * A high priority sentinel request may be submitted to clear the + * submission queue. As it will be the only request in-flight, upon + * execution all other active requests will have been preempted and + * unsubmitted. This preemptive pulse is used to re-evaluate the + * in-flight requests, particularly in cases where an active context + * is banned and those active requests need to be cancelled. + */ + I915_FENCE_FLAG_SENTINEL, + + /* + * I915_FENCE_FLAG_BOOST - upclock the gpu for this request + * + * Some requests are more important than others! In particular, a + * request that the user is waiting on is typically required for + * interactive latency, for which we want to minimise by upclocking + * the GPU. Here we track such boost requests on a per-request basis. + */ + I915_FENCE_FLAG_BOOST, + + /* + * I915_FENCE_FLAG_SUBMIT_PARALLEL - request with a context in a + * parent-child relationship (parallel submission, multi-lrc) should + * trigger a submission to the GuC rather than just moving the context + * tail. + */ + I915_FENCE_FLAG_SUBMIT_PARALLEL, + + /* + * I915_FENCE_FLAG_SKIP_PARALLEL - request with a context in a + * parent-child relationship (parallel submission, multi-lrc) that + * hit an error while generating requests in the execbuf IOCTL. + * Indicates this request should be skipped as another request in + * submission / relationship encoutered an error. + */ + I915_FENCE_FLAG_SKIP_PARALLEL, + + /* + * I915_FENCE_FLAG_COMPOSITE - Indicates fence is part of a composite + * fence (dma_fence_array) and i915 generated for parallel submission. + */ + I915_FENCE_FLAG_COMPOSITE, }; /** @@ -95,9 +196,6 @@ struct i915_request { struct dma_fence fence; spinlock_t lock; - /** On Which ring this request was generated */ - struct drm_i915_private *i915; - /** * Context and ring buffer related to this request * Contexts are refcounted, so when this request is associated with a @@ -108,12 +206,13 @@ struct i915_request { * i915_request_free() will then decrement the refcount on the * context. */ - struct i915_gem_context *gem_context; struct intel_engine_cs *engine; - struct intel_context *hw_context; + struct intel_context *context; struct intel_ring *ring; - struct i915_timeline *timeline; + struct intel_timeline __rcu *timeline; + struct list_head signal_link; + struct llist_node signal_node; /* * The rcu epoch of when this request was allocated. Used to judiciously @@ -143,9 +242,18 @@ struct i915_request { union { wait_queue_entry_t submitq; struct i915_sw_dma_fence_cb dmaq; + struct i915_request_duration_cb { + struct dma_fence_cb cb; + ktime_t emitted; + } duration; }; - struct list_head execute_cb; + struct llist_head execute_cb; struct i915_sw_fence semaphore; + /** + * @submit_work: complete submit fence from an IRQ if needed for + * locking hierarchy reasons. + */ + struct irq_work submit_work; /* * A list of everyone we wait upon, and everyone who waits upon us. @@ -167,16 +275,6 @@ struct i915_request { */ const u32 *hwsp_seqno; - /* - * If we need to access the timeline's seqno for this request in - * another request, we need to keep a read reference to this associated - * cacheline, so that we do not free and recycle it before the foreign - * observers have completed. Hence, we keep a pointer to the cacheline - * inside the timeline's HWSP vma, but it is only valid while this - * request has not completed and guarded by the timeline mutex. - */ - struct i915_timeline_cacheline *hwsp_cacheline; - /** Position in the ring of the start of the request */ u32 head; @@ -199,10 +297,12 @@ struct i915_request { /** Preallocate space in the ring for the emitting the request */ u32 reserved_space; - /** Batch buffer related to this request if any (used for - * error state dump only). - */ - struct i915_vma *batch; + /** Batch buffer pointer for selftest internal use. */ + I915_SELFTEST_DECLARE(struct i915_vma *batch); + + struct i915_vma_resource *batch_res; + +#if IS_ENABLED(CONFIG_DRM_I915_CAPTURE_ERROR) /** * Additional buffers requested by userspace to be captured upon * a GPU hang. The vma/obj on this list are protected by their @@ -210,22 +310,41 @@ struct i915_request { * on the active_list (of their final request). */ struct i915_capture_list *capture_list; - struct list_head active_list; +#endif /** Time at which this request was emitted, in jiffies. */ unsigned long emitted_jiffies; - bool waitboost; - /** timeline->request entry for this request */ struct list_head link; - /** ring->request_list entry for this request */ - struct list_head ring_link; + /** Watchdog support fields. */ + struct i915_request_watchdog { + struct llist_node link; + struct hrtimer timer; + } watchdog; + + /** + * @guc_fence_link: Requests may need to be stalled when using GuC + * submission waiting for certain GuC operations to complete. If that is + * the case, stalled requests are added to a per context list of stalled + * requests. The below list_head is the link in that list. Protected by + * ce->guc_state.lock. + */ + struct list_head guc_fence_link; - struct drm_i915_file_private *file_priv; - /** file_priv list entry for this request */ - struct list_head client_link; + /** + * @guc_prio: Priority level while the request is in flight. Differs + * from i915 scheduler priority. See comment above + * I915_SCHEDULER_CAP_STATIC_PRIORITY_MAP for details. Protected by + * ce->guc_active.lock. Two special values (GUC_PRIO_INIT and + * GUC_PRIO_FINI) outside the GuC priority range are used to indicate + * if the priority has not been initialized yet or if no more updates + * are possible because the request has completed. + */ +#define GUC_PRIO_INIT 0xff +#define GUC_PRIO_FINI 0xfe + u8 guc_prio; I915_SELFTEST_DECLARE(struct { struct list_head link; @@ -242,13 +361,23 @@ static inline bool dma_fence_is_i915(const struct dma_fence *fence) return fence->ops == &i915_fence_ops; } +struct kmem_cache *i915_request_slab_cache(void); + struct i915_request * __must_check __i915_request_create(struct intel_context *ce, gfp_t gfp); struct i915_request * __must_check i915_request_create(struct intel_context *ce); +void __i915_request_skip(struct i915_request *rq); +bool i915_request_set_error_once(struct i915_request *rq, int error); +struct i915_request *i915_request_mark_eio(struct i915_request *rq); + struct i915_request *__i915_request_commit(struct i915_request *request); +void __i915_request_queue(struct i915_request *rq, + const struct i915_sched_attr *attr); +void __i915_request_queue_bh(struct i915_request *rq); +bool i915_request_retire(struct i915_request *rq); void i915_request_retire_upto(struct i915_request *rq); static inline struct i915_request * @@ -283,34 +412,37 @@ int i915_request_await_object(struct i915_request *to, bool write); int i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence); +int i915_request_await_deps(struct i915_request *rq, const struct i915_deps *deps); int i915_request_await_execution(struct i915_request *rq, - struct dma_fence *fence, - void (*hook)(struct i915_request *rq, - struct dma_fence *signal)); + struct dma_fence *fence); void i915_request_add(struct i915_request *rq); -void __i915_request_submit(struct i915_request *request); +bool __i915_request_submit(struct i915_request *request); void i915_request_submit(struct i915_request *request); -void i915_request_skip(struct i915_request *request, int error); - void __i915_request_unsubmit(struct i915_request *request); void i915_request_unsubmit(struct i915_request *request); -/* Note: part of the intel_breadcrumbs family */ -bool i915_request_enable_breadcrumb(struct i915_request *request); -void i915_request_cancel_breadcrumb(struct i915_request *request); +void i915_request_cancel(struct i915_request *rq, int error); + +long i915_request_wait_timeout(struct i915_request *rq, + unsigned int flags, + long timeout) + __attribute__((nonnull(1))); long i915_request_wait(struct i915_request *rq, unsigned int flags, long timeout) __attribute__((nonnull(1))); #define I915_WAIT_INTERRUPTIBLE BIT(0) -#define I915_WAIT_LOCKED BIT(1) /* struct_mutex held, handle GPU reset */ -#define I915_WAIT_PRIORITY BIT(2) /* small priority bump for the request */ -#define I915_WAIT_ALL BIT(3) /* used by i915_gem_object_wait() */ -#define I915_WAIT_FOR_IDLE_BOOST BIT(4) +#define I915_WAIT_PRIORITY BIT(1) /* small priority bump for the request */ +#define I915_WAIT_ALL BIT(2) /* used by i915_gem_object_wait() */ + +void i915_request_show(struct drm_printer *m, + const struct i915_request *rq, + const char *prefix, + int indent); static inline bool i915_request_signaled(const struct i915_request *rq) { @@ -323,6 +455,17 @@ static inline bool i915_request_is_active(const struct i915_request *rq) return test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags); } +static inline bool i915_request_in_priority_queue(const struct i915_request *rq) +{ + return test_bit(I915_FENCE_FLAG_PQUEUE, &rq->fence.flags); +} + +static inline bool +i915_request_has_initial_breadcrumb(const struct i915_request *rq) +{ + return test_bit(I915_FENCE_FLAG_INITIAL_BREADCRUMB, &rq->fence.flags); +} + /** * Returns true if seq1 is later than seq2. */ @@ -333,7 +476,9 @@ static inline bool i915_seqno_passed(u32 seq1, u32 seq2) static inline u32 __hwsp_seqno(const struct i915_request *rq) { - return READ_ONCE(*rq->hwsp_seqno); + const u32 *hwsp = READ_ONCE(rq->hwsp_seqno); + + return READ_ONCE(*hwsp); } /** @@ -362,7 +507,7 @@ static inline u32 hwsp_seqno(const struct i915_request *rq) static inline bool __i915_request_has_started(const struct i915_request *rq) { - return i915_seqno_passed(hwsp_seqno(rq), rq->fence.seqno - 1); + return i915_seqno_passed(__hwsp_seqno(rq), rq->fence.seqno - 1); } /** @@ -393,11 +538,19 @@ static inline bool __i915_request_has_started(const struct i915_request *rq) */ static inline bool i915_request_started(const struct i915_request *rq) { + bool result; + if (i915_request_signaled(rq)) return true; - /* Remember: started but may have since been preempted! */ - return __i915_request_has_started(rq); + result = true; + rcu_read_lock(); /* the HWSP may be freed at runtime */ + if (likely(!i915_request_signaled(rq))) + /* Remember: started but may have since been preempted! */ + result = __i915_request_has_started(rq); + rcu_read_unlock(); + + return result; } /** @@ -410,25 +563,163 @@ static inline bool i915_request_started(const struct i915_request *rq) */ static inline bool i915_request_is_running(const struct i915_request *rq) { + bool result; + if (!i915_request_is_active(rq)) return false; - return __i915_request_has_started(rq); + rcu_read_lock(); + result = __i915_request_has_started(rq) && i915_request_is_active(rq); + rcu_read_unlock(); + + return result; +} + +/** + * i915_request_is_ready - check if the request is ready for execution + * @rq: the request + * + * Upon construction, the request is instructed to wait upon various + * signals before it is ready to be executed by the HW. That is, we do + * not want to start execution and read data before it is written. In practice, + * this is controlled with a mixture of interrupts and semaphores. Once + * the submit fence is completed, the backend scheduler will place the + * request into its queue and from there submit it for execution. So we + * can detect when a request is eligible for execution (and is under control + * of the scheduler) by querying where it is in any of the scheduler's lists. + * + * Returns true if the request is ready for execution (it may be inflight), + * false otherwise. + */ +static inline bool i915_request_is_ready(const struct i915_request *rq) +{ + return !list_empty(&rq->sched.link); +} + +static inline bool __i915_request_is_complete(const struct i915_request *rq) +{ + return i915_seqno_passed(__hwsp_seqno(rq), rq->fence.seqno); } static inline bool i915_request_completed(const struct i915_request *rq) { + bool result; + if (i915_request_signaled(rq)) return true; - return i915_seqno_passed(hwsp_seqno(rq), rq->fence.seqno); + result = true; + rcu_read_lock(); /* the HWSP may be freed at runtime */ + if (likely(!i915_request_signaled(rq))) + result = __i915_request_is_complete(rq); + rcu_read_unlock(); + + return result; } static inline void i915_request_mark_complete(struct i915_request *rq) { - rq->hwsp_seqno = (u32 *)&rq->fence.seqno; /* decouple from HWSP */ + WRITE_ONCE(rq->hwsp_seqno, /* decouple from HWSP */ + (u32 *)&rq->fence.seqno); +} + +static inline bool i915_request_has_waitboost(const struct i915_request *rq) +{ + return test_bit(I915_FENCE_FLAG_BOOST, &rq->fence.flags); +} + +static inline bool i915_request_has_nopreempt(const struct i915_request *rq) +{ + /* Preemption should only be disabled very rarely */ + return unlikely(test_bit(I915_FENCE_FLAG_NOPREEMPT, &rq->fence.flags)); +} + +static inline bool i915_request_has_sentinel(const struct i915_request *rq) +{ + return unlikely(test_bit(I915_FENCE_FLAG_SENTINEL, &rq->fence.flags)); +} + +static inline bool i915_request_on_hold(const struct i915_request *rq) +{ + return unlikely(test_bit(I915_FENCE_FLAG_HOLD, &rq->fence.flags)); +} + +static inline void i915_request_set_hold(struct i915_request *rq) +{ + set_bit(I915_FENCE_FLAG_HOLD, &rq->fence.flags); +} + +static inline void i915_request_clear_hold(struct i915_request *rq) +{ + clear_bit(I915_FENCE_FLAG_HOLD, &rq->fence.flags); +} + +static inline struct intel_timeline * +i915_request_timeline(const struct i915_request *rq) +{ + /* Valid only while the request is being constructed (or retired). */ + return rcu_dereference_protected(rq->timeline, + lockdep_is_held(&rcu_access_pointer(rq->timeline)->mutex) || + test_bit(CONTEXT_IS_PARKING, &rq->context->flags)); +} + +static inline struct i915_gem_context * +i915_request_gem_context(const struct i915_request *rq) +{ + /* Valid only while the request is being constructed (or retired). */ + return rcu_dereference_protected(rq->context->gem_context, true); +} + +static inline struct intel_timeline * +i915_request_active_timeline(const struct i915_request *rq) +{ + /* + * When in use during submission, we are protected by a guarantee that + * the context/timeline is pinned and must remain pinned until after + * this submission. + */ + return rcu_dereference_protected(rq->timeline, + lockdep_is_held(&rq->engine->sched_engine->lock)); +} + +static inline u32 +i915_request_active_seqno(const struct i915_request *rq) +{ + u32 hwsp_phys_base = + page_mask_bits(i915_request_active_timeline(rq)->hwsp_offset); + u32 hwsp_relative_offset = offset_in_page(rq->hwsp_seqno); + + /* + * Because of wraparound, we cannot simply take tl->hwsp_offset, + * but instead use the fact that the relative for vaddr is the + * offset as for hwsp_offset. Take the top bits from tl->hwsp_offset + * and combine them with the relative offset in rq->hwsp_seqno. + * + * As rw->hwsp_seqno is rewritten when signaled, this only works + * when the request isn't signaled yet, but at that point you + * no longer need the offset. + */ + + return hwsp_phys_base + hwsp_relative_offset; } -bool i915_retire_requests(struct drm_i915_private *i915); +bool +i915_request_active_engine(struct i915_request *rq, + struct intel_engine_cs **active); + +void i915_request_notify_execute_cb_imm(struct i915_request *rq); + +enum i915_request_state { + I915_REQUEST_UNKNOWN = 0, + I915_REQUEST_COMPLETE, + I915_REQUEST_PENDING, + I915_REQUEST_QUEUED, + I915_REQUEST_ACTIVE, +}; + +enum i915_request_state i915_test_request_state(struct i915_request *rq); + +void i915_request_module_exit(void); +int i915_request_module_init(void); #endif /* I915_REQUEST_H */ |