[Intel-gfx] [PATCH 04/17] drm/i915: Move engine reset prepare/finish to backends
Chris Wilson
chris at chris-wilson.co.uk
Fri Mar 30 10:39:02 UTC 2018
In preparation to more carefully handling incomplete preemption during
reset by execlists, we move the existing code wholesale to the backends
under a couple of new reset vfuncs.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: MichaĆ Winiarski <michal.winiarski at intel.com>
CC: Michel Thierry <michel.thierry at intel.com>
Cc: Jeff McGee <jeff.mcgee at intel.com>
Reviewed-by: Jeff McGee <jeff.mcgee at intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 47 +++-----------------------
drivers/gpu/drm/i915/intel_lrc.c | 59 +++++++++++++++++++++++++++++++--
drivers/gpu/drm/i915/intel_ringbuffer.c | 23 +++++++++++--
drivers/gpu/drm/i915/intel_ringbuffer.h | 9 +++--
4 files changed, 88 insertions(+), 50 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 9650a7b10c5f..038867c96809 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2917,7 +2917,7 @@ static bool engine_stalled(struct intel_engine_cs *engine)
struct i915_request *
i915_gem_reset_prepare_engine(struct intel_engine_cs *engine)
{
- struct i915_request *request = NULL;
+ struct i915_request *request;
/*
* During the reset sequence, we must prevent the engine from
@@ -2940,40 +2940,7 @@ i915_gem_reset_prepare_engine(struct intel_engine_cs *engine)
*/
kthread_park(engine->breadcrumbs.signaler);
- /*
- * Prevent request submission to the hardware until we have
- * completed the reset in i915_gem_reset_finish(). If a request
- * is completed by one engine, it may then queue a request
- * to a second via its execlists->tasklet *just* as we are
- * calling engine->init_hw() and also writing the ELSP.
- * Turning off the execlists->tasklet until the reset is over
- * prevents the race.
- *
- * Note that this needs to be a single atomic operation on the
- * tasklet (flush existing tasks, prevent new tasks) to prevent
- * a race between reset and set-wedged. It is not, so we do the best
- * we can atm and make sure we don't lock the machine up in the more
- * common case of recursively being called from set-wedged from inside
- * i915_reset.
- */
- if (!atomic_read(&engine->execlists.tasklet.count))
- tasklet_kill(&engine->execlists.tasklet);
- tasklet_disable(&engine->execlists.tasklet);
-
- /*
- * We're using worker to queue preemption requests from the tasklet in
- * GuC submission mode.
- * Even though tasklet was disabled, we may still have a worker queued.
- * Let's make sure that all workers scheduled before disabling the
- * tasklet are completed before continuing with the reset.
- */
- if (engine->i915->guc.preempt_wq)
- flush_workqueue(engine->i915->guc.preempt_wq);
-
- if (engine->irq_seqno_barrier)
- engine->irq_seqno_barrier(engine);
-
- request = i915_gem_find_active_request(engine);
+ request = engine->reset.prepare(engine);
if (request && request->fence.error == -EIO)
request = ERR_PTR(-EIO); /* Previous reset failed! */
@@ -3114,13 +3081,8 @@ void i915_gem_reset_engine(struct intel_engine_cs *engine,
if (request)
request = i915_gem_reset_request(engine, request);
- if (request) {
- DRM_DEBUG_DRIVER("resetting %s to restart from tail of request 0x%x\n",
- engine->name, request->global_seqno);
- }
-
/* Setup the CS to resume from the breadcrumb of the hung request */
- engine->reset_hw(engine, request);
+ engine->reset.reset(engine, request);
}
void i915_gem_reset(struct drm_i915_private *dev_priv)
@@ -3172,7 +3134,8 @@ void i915_gem_reset(struct drm_i915_private *dev_priv)
void i915_gem_reset_finish_engine(struct intel_engine_cs *engine)
{
- tasklet_enable(&engine->execlists.tasklet);
+ engine->reset.finish(engine);
+
kthread_unpark(engine->breadcrumbs.signaler);
intel_uncore_forcewake_put(engine->i915, FORCEWAKE_ALL);
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 51e930323626..42f9af625e88 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -1732,8 +1732,48 @@ static int gen9_init_render_ring(struct intel_engine_cs *engine)
return init_workarounds_ring(engine);
}
-static void reset_common_ring(struct intel_engine_cs *engine,
- struct i915_request *request)
+static struct i915_request *
+execlists_reset_prepare(struct intel_engine_cs *engine)
+{
+ struct intel_engine_execlists * const execlists = &engine->execlists;
+
+ GEM_TRACE("%s\n", engine->name);
+
+ /*
+ * Prevent request submission to the hardware until we have
+ * completed the reset in i915_gem_reset_finish(). If a request
+ * is completed by one engine, it may then queue a request
+ * to a second via its execlists->tasklet *just* as we are
+ * calling engine->init_hw() and also writing the ELSP.
+ * Turning off the execlists->tasklet until the reset is over
+ * prevents the race.
+ *
+ * Note that this needs to be a single atomic operation on the
+ * tasklet (flush existing tasks, prevent new tasks) to prevent
+ * a race between reset and set-wedged. It is not, so we do the best
+ * we can atm and make sure we don't lock the machine up in the more
+ * common case of recursively being called from set-wedged from inside
+ * i915_reset.
+ */
+ if (!atomic_read(&execlists->tasklet.count))
+ tasklet_kill(&execlists->tasklet);
+ tasklet_disable(&execlists->tasklet);
+
+ /*
+ * We're using worker to queue preemption requests from the tasklet in
+ * GuC submission mode.
+ * Even though tasklet was disabled, we may still have a worker queued.
+ * Let's make sure that all workers scheduled before disabling the
+ * tasklet are completed before continuing with the reset.
+ */
+ if (engine->i915->guc.preempt_wq)
+ flush_workqueue(engine->i915->guc.preempt_wq);
+
+ return i915_gem_find_active_request(engine);
+}
+
+static void execlists_reset(struct intel_engine_cs *engine,
+ struct i915_request *request)
{
struct intel_engine_execlists * const execlists = &engine->execlists;
struct intel_context *ce;
@@ -1742,6 +1782,9 @@ static void reset_common_ring(struct intel_engine_cs *engine,
GEM_TRACE("%s seqno=%x\n",
engine->name, request ? request->global_seqno : 0);
+ /* The submission tasklet must be disabled, engine->reset.prepare(). */
+ GEM_BUG_ON(!atomic_read(&execlists->tasklet.count));
+
/* See execlists_cancel_requests() for the irq/spinlock split. */
local_irq_save(flags);
@@ -1802,6 +1845,13 @@ static void reset_common_ring(struct intel_engine_cs *engine,
unwind_wa_tail(request);
}
+static void execlists_reset_finish(struct intel_engine_cs *engine)
+{
+ tasklet_enable(&engine->execlists.tasklet);
+
+ GEM_TRACE("%s\n", engine->name);
+}
+
static int intel_logical_ring_emit_pdps(struct i915_request *rq)
{
struct i915_hw_ppgtt *ppgtt = rq->ctx->ppgtt;
@@ -2126,7 +2176,10 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine)
{
/* Default vfuncs which can be overriden by each engine. */
engine->init_hw = gen8_init_common_ring;
- engine->reset_hw = reset_common_ring;
+
+ engine->reset.prepare = execlists_reset_prepare;
+ engine->reset.reset = execlists_reset;
+ engine->reset.finish = execlists_reset_finish;
engine->context_pin = execlists_context_pin;
engine->context_unpin = execlists_context_unpin;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 04d9d9a946a7..5dadbc435c0e 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -530,9 +530,20 @@ static int init_ring_common(struct intel_engine_cs *engine)
return ret;
}
-static void reset_ring_common(struct intel_engine_cs *engine,
- struct i915_request *request)
+static struct i915_request *reset_prepare(struct intel_engine_cs *engine)
{
+ if (engine->irq_seqno_barrier)
+ engine->irq_seqno_barrier(engine);
+
+ return i915_gem_find_active_request(engine);
+}
+
+static void reset_ring(struct intel_engine_cs *engine,
+ struct i915_request *request)
+{
+ GEM_TRACE("%s seqno=%x\n",
+ engine->name, request ? request->global_seqno : 0);
+
/*
* RC6 must be prevented until the reset is complete and the engine
* reinitialised. If it occurs in the middle of this sequence, the
@@ -595,6 +606,10 @@ static void reset_ring_common(struct intel_engine_cs *engine,
}
}
+static void reset_finish(struct intel_engine_cs *engine)
+{
+}
+
static int intel_rcs_ctx_init(struct i915_request *rq)
{
int ret;
@@ -1987,7 +2002,9 @@ static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv,
intel_ring_init_semaphores(dev_priv, engine);
engine->init_hw = init_ring_common;
- engine->reset_hw = reset_ring_common;
+ engine->reset.prepare = reset_prepare;
+ engine->reset.reset = reset_ring;
+ engine->reset.finish = reset_finish;
engine->context_pin = intel_ring_context_pin;
engine->context_unpin = intel_ring_context_unpin;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 2e20627e254b..15d624925594 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -418,8 +418,13 @@ struct intel_engine_cs {
void (*irq_disable)(struct intel_engine_cs *engine);
int (*init_hw)(struct intel_engine_cs *engine);
- void (*reset_hw)(struct intel_engine_cs *engine,
- struct i915_request *rq);
+
+ struct {
+ struct i915_request *(*prepare)(struct intel_engine_cs *engine);
+ void (*reset)(struct intel_engine_cs *engine,
+ struct i915_request *rq);
+ void (*finish)(struct intel_engine_cs *engine);
+ } reset;
void (*park)(struct intel_engine_cs *engine);
void (*unpark)(struct intel_engine_cs *engine);
--
2.16.3
More information about the Intel-gfx
mailing list