[Intel-gfx] [PATCH] drm/i915/selftests: Exercise potential false lite-restore

[Tvrtko Ursulin]

On 01/10/2019 10:51, Chris Wilson wrote:
> If execlists's lite-restore is based on the common GEM context tag
> rather than the per-intel_context LRCA, then a context switch between
> two intel_contexts on the same engine derived from the same GEM context
> will perform a lite-restore instead of a full context switch. We can
> exploit this by poisoning the ringbuffer of the first context and trying
> to trick a simple RING_TAIL update (i.e. lite-restore)
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> ---
> Then switch back to ce[0] for fun.
> ---
>   drivers/gpu/drm/i915/gt/selftest_lrc.c | 132 +++++++++++++++++++++++++
>   1 file changed, 132 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> index 93f2fcdc49bf..b90b970a44b9 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
> @@ -79,6 +79,137 @@ static int live_sanitycheck(void *arg)
>   	return err;
>   }
>   
> +static int live_unlite_restore(void *arg)
> +{
> +	struct drm_i915_private *i915 = arg;
> +	struct intel_engine_cs *engine;
> +	struct i915_gem_context *ctx;
> +	enum intel_engine_id id;
> +	intel_wakeref_t wakeref;
> +	struct igt_spinner spin;
> +	int err = -ENOMEM;
> +
> +	/*
> +	 * Check that we can correctly context switch between 2 instances
> +	 * on the same engine from the same parent context.
> +	 */
> +
> +	mutex_lock(&i915->drm.struct_mutex);
> +	wakeref = intel_runtime_pm_get(&i915->runtime_pm);
> +
> +	if (igt_spinner_init(&spin, &i915->gt))
> +		goto err_unlock;
> +
> +	ctx = kernel_context(i915);
> +	if (!ctx)
> +		goto err_spin;
> +
> +	for_each_engine(engine, i915, id) {
> +		struct intel_context *ce[2] = {};
> +		struct i915_request *rq[3];
> +		struct igt_live_test t;
> +		int n;
> +
> +		if (!intel_engine_can_store_dword(engine))
> +			continue;
> +
> +		if (igt_live_test_begin(&t, i915, __func__, engine->name)) {
> +			err = -EIO;
> +			break;
> +		}
> +
> +		for (n = 0; n < ARRAY_SIZE(ce); n++) {
> +			ce[n] = intel_context_create(ctx, engine);
> +			if (IS_ERR(ce[n])) {
> +				err = PTR_ERR(ce[n]);
> +				goto err_ce;
> +			}
> +
> +			err = intel_context_pin(ce[n]);
> +			if (err)
> +				goto err_ce;

If pinning fails err_ce path will underflow the unpin. Perhaps you need 
to only store in ce[] when both steps have passed and keep it in a local 
until then.

> +
> +			/*
> +			 * Setup the pair of contexts such that if we
> +			 * lite-restore using the RING_TAIL from ce[1] it
> +			 * will execute garbage from ce[0]->ring.
> +			 */
> +			memset(ce[n]->ring->vaddr,
> +			       POISON_INUSE,
> +			       ce[n]->ring->vma->size);
> +		}
> +		intel_ring_reset(ce[1]->ring, ce[1]->ring->vma->size / 2);
> +		__execlists_update_reg_state(ce[1], engine);
> +
> +		rq[0] = igt_spinner_create_request(&spin, ce[0], MI_ARB_CHECK);
> +		if (IS_ERR(rq[0])) {
> +			err = PTR_ERR(rq[0]);
> +			goto err_ce;
> +		}
> +
> +		GEM_BUG_ON(rq[0]->tail > ce[1]->ring->emit);
> +		i915_request_get(rq[0]);
> +		i915_request_add(rq[0]);
> +
> +		if (!igt_wait_for_spinner(&spin, rq[0])) {
> +			i915_request_put(rq[0]);
> +			goto err_ce;
> +		}
> +
> +		rq[1] = i915_request_create(ce[1]);
> +		if (IS_ERR(rq[1])) {
> +			err = PTR_ERR(rq[1]);
> +			i915_request_put(rq[0]);
> +			goto err_ce;
> +		}
> +		GEM_BUG_ON(rq[1]->tail <= rq[0]->tail);
> +
> +		/* Ensure we do a completion switch from ce[0] to ce[1] */
> +		i915_request_await_dma_fence(rq[1], &rq[0]->fence);

What do you mean by completion switch? You are setting up a dependency 
so rq[1] (and rq[2]) won't be put into the elsp until spinner is ended 
so it may not even be a context switch. Wouldn't you actually need the 
opposite? I was expecting you would let the spinner run, make sure rq[1] 
is in elsp and then count on time slicing to trigger a context switch.

Regards,

Tvrtko

> +		i915_request_put(rq[0]);
> +
> +		i915_request_get(rq[1]);
> +		i915_request_add(rq[1]);
> +
> +		/* And switch back to ce[0] for good measure */
> +		rq[2] = i915_request_create(ce[0]);
> +		if (IS_ERR(rq[2])) {
> +			err = PTR_ERR(rq[2]);
> +			i915_request_put(rq[1]);
> +			goto err_ce;
> +		}
> +		GEM_BUG_ON(rq[2]->tail > rq[1]->tail);
> +
> +		i915_request_await_dma_fence(rq[2], &rq[1]->fence);
> +		i915_request_put(rq[1]);
> +
> +		i915_request_add(rq[2]);
> +
> +err_ce:
> +		igt_spinner_end(&spin);
> +		for (n = 0; n < ARRAY_SIZE(ce); n++) {
> +			if (IS_ERR_OR_NULL(ce[n]))
> +				break;
> +
> +			intel_context_unpin(ce[n]);
> +			intel_context_put(ce[n]);
> +		}
> +
> +		if (igt_live_test_end(&t))
> +			err = -EIO;
> +		if (err)
> +			break;
> +	}
> +
> +	kernel_context_close(ctx);
> +err_spin:
> +	igt_spinner_fini(&spin);
> +err_unlock:
> +	intel_runtime_pm_put(&i915->runtime_pm, wakeref);
> +	mutex_unlock(&i915->drm.struct_mutex);
> +	return err;
> +}
> +
>   static int
>   emit_semaphore_chain(struct i915_request *rq, struct i915_vma *vma, int idx)
>   {
> @@ -2178,6 +2309,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915)
>   {
>   	static const struct i915_subtest tests[] = {
>   		SUBTEST(live_sanitycheck),
> +		SUBTEST(live_unlite_restore),
>   		SUBTEST(live_timeslice_preempt),
>   		SUBTEST(live_busywait_preempt),
>   		SUBTEST(live_preempt),
>