bool valid;
/* Software tracked state */
bool waiting;
+ int num_waiters;
int hangcheck_score;
enum intel_ring_hangcheck_action hangcheck_action;
int num_requests;
u32 tail;
} *requests;
+ struct drm_i915_error_waiter {
+ char comm[TASK_COMM_LEN];
+ pid_t pid;
+ u32 seqno;
+ } *waiters;
+
struct {
u32 gfx_mode;
union {
/** LRU list of objects with fence regs on them. */
struct list_head fence_list;
- /**
- * We leave the user IRQ off as much as possible,
- * but this means that requests will finish and never
- * be retired once the system goes idle. Set a timer to
- * fire periodically while the ring is running. When it
- * fires, go retire requests.
- */
- struct delayed_work retire_work;
-
- /**
- * When we detect an idle GPU, we want to turn on
- * powersaving features. So once we see that there
- * are no more requests outstanding and no more
- * arrive within a small period of time, we fire
- * off the idle_work.
- */
- struct delayed_work idle_work;
-
/**
* Are we in a non-interruptible section of code like
* modesetting?
*/
bool interruptible;
- /**
- * Is the GPU currently considered idle, or busy executing userspace
- * requests? Whilst idle, we attempt to power down the hardware and
- * display clocks. In order to reduce the effect on performance, there
- * is a slight delay before we do so.
- */
- bool busy;
-
/* the indicator for dispatch video commands on two BSD rings */
unsigned int bsd_ring_dispatch_index;
/* Hang gpu twice in this window and your context gets banned */
#define DRM_I915_CTX_BAN_PERIOD DIV_ROUND_UP(8*DRM_I915_HANGCHECK_PERIOD, 1000)
- struct workqueue_struct *hangcheck_wq;
struct delayed_work hangcheck_work;
/* For reset and error_state handling. */
#define I915_RESET_IN_PROGRESS_FLAG 1
#define I915_WEDGED (1 << 31)
+ /**
+ * Waitqueue to signal when a hang is detected. Used to for waiters
+ * to release the struct_mutex for the reset to procede.
+ */
+ wait_queue_head_t wait_queue;
+
/**
* Waitqueue to signal when the reset has completed. Used by clients
* that wait for dev_priv->mm.wedged to settle.
*/
wait_queue_head_t reset_queue;
- /* Userspace knobs for gpu hang simulation;
- * combines both a ring mask, and extra flags
- */
- u32 stop_rings;
-#define I915_STOP_RING_ALLOW_BAN (1 << 31)
-#define I915_STOP_RING_ALLOW_WARN (1 << 30)
-
/* For missed irq/seqno simulation. */
- unsigned int test_irq_rings;
+ unsigned long test_irq_rings;
};
enum modeset_restore {
int (*init_engines)(struct drm_device *dev);
void (*cleanup_engine)(struct intel_engine_cs *engine);
void (*stop_engine)(struct intel_engine_cs *engine);
+
+ /**
+ * Is the GPU currently considered idle, or busy executing
+ * userspace requests? Whilst idle, we allow runtime power
+ * management to power down the hardware and display clocks.
+ * In order to reduce the effect on performance, there
+ * is a slight delay before we do so.
+ */
+ unsigned int active_engines;
+ bool awake;
+
+ /**
+ * We leave the user IRQ off as much as possible,
+ * but this means that requests will finish and never
+ * be retired once the system goes idle. Set a timer to
+ * fire periodically while the ring is running. When it
+ * fires, go retire requests.
+ */
+ struct delayed_work retire_work;
+
+ /**
+ * When we detect an idle GPU, we want to turn on
+ * powersaving features. So once we see that there
+ * are no more requests outstanding and no more
+ * arrive within a small period of time, we fire
+ * off the idle_work.
+ */
+ struct delayed_work idle_work;
} gt;
/* perform PHY state sanity checks? */
/** On Which ring this request was generated */
struct drm_i915_private *i915;
struct intel_engine_cs *engine;
- unsigned reset_counter;
+ struct intel_signal_node signaling;
/** GEM sequence number associated with the previous request,
* when the HWS breadcrumb is equal to this the GPU is processing
bool intel_hpd_pin_to_port(enum hpd_pin pin, enum port *port);
/* i915_irq.c */
-void i915_queue_hangcheck(struct drm_i915_private *dev_priv);
+static inline void i915_queue_hangcheck(struct drm_i915_private *dev_priv)
+{
+ unsigned long delay;
+
+ if (unlikely(!i915.enable_hangcheck))
+ return;
+
+ /* Don't continually defer the hangcheck so that it is always run at
+ * least once after work has been scheduled on any ring. Otherwise,
+ * we will ignore a hung ring if a second ring is kept busy.
+ */
+
+ delay = round_jiffies_up_relative(DRM_I915_HANGCHECK_JIFFIES);
+ queue_delayed_work(system_long_wq,
+ &dev_priv->gpu_error.hangcheck_work, delay);
+}
+
__printf(3, 4)
void i915_handle_error(struct drm_i915_private *dev_priv,
u32 engine_mask,
void assert_forcewakes_inactive(struct drm_i915_private *dev_priv);
+int intel_wait_for_register(struct drm_i915_private *dev_priv,
+ i915_reg_t reg,
+ const u32 mask,
+ const u32 value,
+ const unsigned long timeout_ms);
+int intel_wait_for_register_fw(struct drm_i915_private *dev_priv,
+ i915_reg_t reg,
+ const u32 mask,
+ const u32 value,
+ const unsigned long timeout_ms);
+
static inline bool intel_gvt_active(struct drm_i915_private *dev_priv)
{
return dev_priv->gvt.initialized;
ibx_display_interrupt_update(dev_priv, bits, 0);
}
-
/* i915_gem.c */
int i915_gem_create_ioctl(struct drm_device *dev, void *data,
struct drm_file *file_priv);
return (int32_t)(seq1 - seq2) >= 0;
}
-static inline bool i915_gem_request_started(struct drm_i915_gem_request *req,
- bool lazy_coherency)
+static inline bool i915_gem_request_started(const struct drm_i915_gem_request *req)
{
- if (!lazy_coherency && req->engine->irq_seqno_barrier)
- req->engine->irq_seqno_barrier(req->engine);
- return i915_seqno_passed(req->engine->get_seqno(req->engine),
+ return i915_seqno_passed(intel_engine_get_seqno(req->engine),
req->previous_seqno);
}
-static inline bool i915_gem_request_completed(struct drm_i915_gem_request *req,
- bool lazy_coherency)
+static inline bool i915_gem_request_completed(const struct drm_i915_gem_request *req)
{
- if (!lazy_coherency && req->engine->irq_seqno_barrier)
- req->engine->irq_seqno_barrier(req->engine);
- return i915_seqno_passed(req->engine->get_seqno(req->engine),
+ return i915_seqno_passed(intel_engine_get_seqno(req->engine),
req->seqno);
}
+bool __i915_spin_request(const struct drm_i915_gem_request *request,
+ int state, unsigned long timeout_us);
+static inline bool i915_spin_request(const struct drm_i915_gem_request *request,
+ int state, unsigned long timeout_us)
+{
+ return (i915_gem_request_started(request) &&
+ __i915_spin_request(request, state, timeout_us));
+}
+
int __must_check i915_gem_get_seqno(struct drm_i915_private *dev_priv, u32 *seqno);
int __must_check i915_gem_set_seqno(struct drm_device *dev, u32 seqno);
struct drm_i915_gem_request *
i915_gem_find_active_request(struct intel_engine_cs *engine);
-bool i915_gem_retire_requests(struct drm_i915_private *dev_priv);
+void i915_gem_retire_requests(struct drm_i915_private *dev_priv);
void i915_gem_retire_requests_ring(struct intel_engine_cs *engine);
static inline u32 i915_reset_counter(struct i915_gpu_error *error)
return ((i915_reset_counter(error) & ~I915_WEDGED) + 1) / 2;
}
-static inline bool i915_stop_ring_allow_ban(struct drm_i915_private *dev_priv)
-{
- return dev_priv->gpu_error.stop_rings == 0 ||
- dev_priv->gpu_error.stop_rings & I915_STOP_RING_ALLOW_BAN;
-}
-
-static inline bool i915_stop_ring_allow_warn(struct drm_i915_private *dev_priv)
-{
- return dev_priv->gpu_error.stop_rings == 0 ||
- dev_priv->gpu_error.stop_rings & I915_STOP_RING_ALLOW_WARN;
-}
-
void i915_gem_reset(struct drm_device *dev);
bool i915_gem_clflush_object(struct drm_i915_gem_object *obj, bool force);
int __must_check i915_gem_init(struct drm_device *dev);
int __must_check i915_gem_init_hw(struct drm_device *dev);
void i915_gem_init_swizzling(struct drm_device *dev);
void i915_gem_cleanup_engines(struct drm_device *dev);
-int __must_check i915_gpu_idle(struct drm_device *dev);
+int __must_check i915_gem_wait_for_idle(struct drm_i915_private *dev_priv);
int __must_check i915_gem_suspend(struct drm_device *dev);
void __i915_add_request(struct drm_i915_gem_request *req,
struct drm_i915_gem_object *batch_obj,
extern int intel_opregion_get_panel_type(struct drm_i915_private *dev_priv);
#else
static inline int intel_opregion_setup(struct drm_i915_private *dev) { return 0; }
-static inline void intel_opregion_init(struct drm_i915_private *dev) { }
-static inline void intel_opregion_fini(struct drm_i915_private *dev) { }
+static inline void intel_opregion_register(struct drm_i915_private *dev_priv) { }
+static inline void intel_opregion_unregister(struct drm_i915_private *dev_priv) { }
static inline void intel_opregion_asle_intr(struct drm_i915_private *dev_priv)
{
}
*/
#define I915_READ_FW(reg__) __raw_i915_read32(dev_priv, (reg__))
#define I915_WRITE_FW(reg__, val__) __raw_i915_write32(dev_priv, (reg__), (val__))
+#define I915_WRITE64_FW(reg__, val__) __raw_i915_write64(dev_priv, (reg__), (val__))
#define POSTING_READ_FW(reg__) (void)I915_READ_FW(reg__)
/* "Broadcast RGB" property */
schedule_timeout_uninterruptible(remaining_jiffies);
}
}
-
-static inline void i915_trace_irq_get(struct intel_engine_cs *engine,
- struct drm_i915_gem_request *req)
+static inline bool __i915_request_irq_complete(struct drm_i915_gem_request *req)
{
- if (engine->trace_irq_req == NULL && engine->irq_get(engine))
- i915_gem_request_assign(&engine->trace_irq_req, req);
+ struct intel_engine_cs *engine = req->engine;
+
+ /* Before we do the heavier coherent read of the seqno,
+ * check the value (hopefully) in the CPU cacheline.
+ */
+ if (i915_gem_request_completed(req))
+ return true;
+
+ /* Ensure our read of the seqno is coherent so that we
+ * do not "miss an interrupt" (i.e. if this is the last
+ * request and the seqno write from the GPU is not visible
+ * by the time the interrupt fires, we will see that the
+ * request is incomplete and go back to sleep awaiting
+ * another interrupt that will never come.)
+ *
+ * Strictly, we only need to do this once after an interrupt,
+ * but it is easier and safer to do it every time the waiter
+ * is woken.
+ */
+ if (engine->irq_seqno_barrier &&
+ cmpxchg_relaxed(&engine->irq_posted, 1, 0)) {
+ /* The ordering of irq_posted versus applying the barrier
+ * is crucial. The clearing of the current irq_posted must
+ * be visible before we perform the barrier operation,
+ * such that if a subsequent interrupt arrives, irq_posted
+ * is reasserted and our task rewoken (which causes us to
+ * do another __i915_request_irq_complete() immediately
+ * and reapply the barrier). Conversely, if the clear
+ * occurs after the barrier, then an interrupt that arrived
+ * whilst we waited on the barrier would not trigger a
+ * barrier on the next pass, and the read may not see the
+ * seqno update.
+ */
+ engine->irq_seqno_barrier(engine);
+ if (i915_gem_request_completed(req))
+ return true;
+ }
+
+ /* We need to check whether any gpu reset happened in between
+ * the request being submitted and now. If a reset has occurred,
+ * the seqno will have been advance past ours and our request
+ * is complete. If we are in the process of handling a reset,
+ * the request is effectively complete as the rendering will
+ * be discarded, but we need to return in order to drop the
+ * struct_mutex.
+ */
+ if (i915_reset_in_progress(&req->i915->gpu_error))
+ return true;
+
+ return false;
}
#endif
projects / cascardo / linux.git / blobdiff