[PATCH 3/4] drm/i915: Record the default hw state after reset upon load
Chris Wilson
chris at chris-wilson.co.uk
Wed Nov 1 13:18:05 UTC 2017
Take a copy of the HW state after a reset upon module loading by
executing a context switch from a blank context to the kernel context,
thus saving the default hw state over the blank context image.
We can then use the default hw state to initialise any future context,
ensuring that each starts with the default view of hw state.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem.c | 57 +++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/i915_gem_context.c | 10 +++---
drivers/gpu/drm/i915/i915_gem_context.h | 3 ++
drivers/gpu/drm/i915/intel_engine_cs.c | 3 ++
drivers/gpu/drm/i915/intel_lrc.c | 27 +++++++++++++---
drivers/gpu/drm/i915/intel_ringbuffer.c | 37 +++++++++++++++++++--
drivers/gpu/drm/i915/intel_ringbuffer.h | 1 +
7 files changed, 125 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 73a4e5685890..58fa650141d3 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4974,6 +4974,59 @@ bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value)
return true;
}
+static int __intel_engines_record_defaults(struct drm_i915_private *i915)
+{
+ struct i915_gem_context *ctx;
+ struct intel_engine_cs *engine;
+ enum intel_engine_id id;
+ int err;
+
+ /*
+ * As we reset the gpu during very early sanitisation, the current
+ * register state on the GPU should reflect its defaults values.
+ * We load a context onto the hw (with restore-inhibit), then switch
+ * over to a second context to save that default register state. We
+ * can then prime every new context with that state so they all start
+ * from defaults.
+ */
+
+ ctx = i915_gem_context_create_kernel(i915, 0);
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
+
+ for_each_engine(engine, i915, id) {
+ struct drm_i915_gem_request *rq;
+
+ rq = i915_gem_request_alloc(engine, ctx);
+ if (IS_ERR(rq)) {
+ err = PTR_ERR(rq);
+ goto out_ctx;
+ }
+ i915_add_request(rq);
+ }
+
+ err = i915_gem_switch_to_kernel_context(i915);
+ if (err)
+ goto out_ctx;
+
+ err = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+ if (err)
+ goto out_ctx;
+
+ for_each_engine(engine, i915, id) {
+ if (!ctx->engine[id].state)
+ continue;
+
+ engine->default_state =
+ i915_gem_object_get(ctx->engine[id].state->obj);
+ }
+
+out_ctx:
+ i915_gem_context_set_closed(ctx);
+ i915_gem_context_put(ctx);
+ return err;
+}
+
int i915_gem_init(struct drm_i915_private *dev_priv)
{
int ret;
@@ -5029,6 +5082,10 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
intel_init_gt_powersave(dev_priv);
+ ret = __intel_engines_record_defaults(dev_priv);
+ if (ret)
+ goto out_unlock;
+
out_unlock:
if (ret == -EIO) {
/* Allow engine initialisation to fail by marking the GPU as
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 10affb35ac56..495524fb2065 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -343,7 +343,7 @@ static void __destroy_hw_context(struct i915_gem_context *ctx,
* context state of the GPU for applications that don't utilize HW contexts, as
* well as an idle case.
*/
-static struct i915_gem_context *
+struct i915_gem_context *
i915_gem_create_context(struct drm_i915_private *dev_priv,
struct drm_i915_file_private *file_priv)
{
@@ -418,8 +418,8 @@ i915_gem_context_create_gvt(struct drm_device *dev)
return ctx;
}
-static struct i915_gem_context *
-create_kernel_context(struct drm_i915_private *i915, int prio)
+struct i915_gem_context *
+i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
{
struct i915_gem_context *ctx;
@@ -473,7 +473,7 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
ida_init(&dev_priv->contexts.hw_ida);
/* lowest priority; idle task */
- ctx = create_kernel_context(dev_priv, I915_PRIORITY_MIN);
+ ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN);
if (IS_ERR(ctx)) {
DRM_ERROR("Failed to create default global context\n");
err = PTR_ERR(ctx);
@@ -487,7 +487,7 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
dev_priv->kernel_context = ctx;
/* highest priority; preempting task */
- ctx = create_kernel_context(dev_priv, INT_MAX);
+ ctx = i915_gem_context_create_kernel(dev_priv, INT_MAX);
if (IS_ERR(ctx)) {
DRM_ERROR("Failed to create default preempt context\n");
err = PTR_ERR(ctx);
diff --git a/drivers/gpu/drm/i915/i915_gem_context.h b/drivers/gpu/drm/i915/i915_gem_context.h
index 44688e22a5c2..fb482ec1351b 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/i915_gem_context.h
@@ -292,6 +292,9 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data,
int i915_gem_context_reset_stats_ioctl(struct drm_device *dev, void *data,
struct drm_file *file);
+struct i915_gem_context *
+i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio);
+
static inline struct i915_gem_context *
i915_gem_context_get(struct i915_gem_context *ctx)
{
diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index f31f2d6384c3..ad80955684b0 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -679,6 +679,9 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine)
intel_engine_cleanup_cmd_parser(engine);
i915_gem_batch_pool_fini(&engine->batch_pool);
+ if (engine->default_state)
+ i915_gem_object_put(engine->default_state);
+
if (HAS_LOGICAL_RING_PREEMPTION(engine->i915))
engine->context_unpin(engine, engine->i915->preempt_context);
engine->context_unpin(engine, engine->i915->kernel_context);
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 6840ec8db037..001e6f7ef595 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -2100,7 +2100,6 @@ static void execlists_init_reg_state(u32 *regs,
CTX_REG(regs, CTX_CONTEXT_CONTROL, RING_CONTEXT_CONTROL(engine),
_MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH |
- CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT |
(HAS_RESOURCE_STREAMER(dev_priv) ?
CTX_CTRL_RS_CTX_ENABLE : 0)));
CTX_REG(regs, CTX_RING_HEAD, RING_HEAD(base), 0);
@@ -2177,6 +2176,7 @@ populate_lr_context(struct i915_gem_context *ctx,
struct intel_ring *ring)
{
void *vaddr;
+ u32 *regs;
int ret;
ret = i915_gem_object_set_to_cpu_domain(ctx_obj, true);
@@ -2193,11 +2193,28 @@ populate_lr_context(struct i915_gem_context *ctx,
}
ctx_obj->mm.dirty = true;
+ if (engine->default_state) {
+ void *defaults;
+
+ defaults = i915_gem_object_pin_map(engine->default_state,
+ I915_MAP_WB);
+ if (IS_ERR(defaults))
+ return PTR_ERR(defaults);
+
+ memcpy(vaddr + LRC_HEADER_PAGES * PAGE_SIZE,
+ defaults + LRC_HEADER_PAGES * PAGE_SIZE,
+ engine->context_size);
+ i915_gem_object_unpin_map(engine->default_state);
+ }
+
/* The second page of the context object contains some fields which must
* be set up prior to the first execution. */
-
- execlists_init_reg_state(vaddr + LRC_STATE_PN * PAGE_SIZE,
- ctx, engine, ring);
+ regs = vaddr + LRC_STATE_PN * PAGE_SIZE;
+ execlists_init_reg_state(regs, ctx, engine, ring);
+ if (!engine->default_state) {
+ regs[CTX_CONTEXT_CONTROL+1] |=
+ _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
+ }
i915_gem_object_unpin_map(ctx_obj);
@@ -2250,7 +2267,7 @@ static int execlists_context_deferred_alloc(struct i915_gem_context *ctx,
ce->ring = ring;
ce->state = vma;
- ce->initialised |= engine->init_context == NULL;
+ ce->initialised |= engine->default_state || !engine->init_context;
return 0;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 47fadf8da84e..f5cedd8037df 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1383,11 +1383,34 @@ alloc_context_vma(struct intel_engine_cs *engine)
struct drm_i915_private *i915 = engine->i915;
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
+ int err;
obj = i915_gem_object_create(i915, engine->context_size);
if (IS_ERR(obj))
return ERR_CAST(obj);
+ if (engine->default_state) {
+ void *defaults, *vaddr;
+
+ vaddr = i915_gem_object_pin_map(obj, I915_MAP_WB);
+ if (IS_ERR(vaddr)) {
+ err = PTR_ERR(vaddr);
+ goto err_obj;
+ }
+
+ defaults = i915_gem_object_pin_map(engine->default_state,
+ I915_MAP_WB);
+ if (IS_ERR(defaults)) {
+ err = PTR_ERR(defaults);
+ goto err_map;
+ }
+
+ memcpy(vaddr, defaults, engine->context_size);
+
+ i915_gem_object_unpin_map(engine->default_state);
+ i915_gem_object_unpin_map(obj);
+ }
+
/*
* Try to make the context utilize L3 as well as LLC.
*
@@ -1409,10 +1432,18 @@ alloc_context_vma(struct intel_engine_cs *engine)
}
vma = i915_vma_instance(obj, &i915->ggtt.base, NULL);
- if (IS_ERR(vma))
- i915_gem_object_put(obj);
+ if (IS_ERR(vma)) {
+ err = PTR_ERR(vma);
+ goto err_obj;
+ }
return vma;
+
+err_map:
+ i915_gem_object_unpin_map(obj);
+err_obj:
+ i915_gem_object_put(obj);
+ return ERR_PTR(err);
}
static struct intel_ring *
@@ -1456,7 +1487,7 @@ intel_ring_context_pin(struct intel_engine_cs *engine,
* as during eviction we cannot allocate and pin the renderstate in
* order to initialise the context.
*/
- if (i915_gem_context_is_kernel(ctx))
+ if (i915_gem_context_is_kernel(ctx) || engine->default_state)
ce->initialised = true;
i915_gem_context_get(ctx);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 69ad875fd011..1d752b9a3065 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -303,6 +303,7 @@ struct intel_engine_cs {
struct intel_ring *buffer;
struct intel_timeline *timeline;
+ struct drm_i915_gem_object *default_state;
struct intel_render_state *render_state;
atomic_t irq_count;
--
2.15.0.rc2
More information about the Intel-gfx-trybot
mailing list