drm/i915: Allocate a status page for each timeline

Allocate a page for use as a status page by a group of timelines, as we
only need a dword of storage for each (rounded up to the cacheline for
safety) we can pack multiple timelines into the same page. Each timeline
will then be able to track its own HW seqno.

v2: Reuse the common per-engine HWSP for the solitary ringbuffer
timeline, so that we do not have to emit (using per-gen specialised
vfuncs) the breadcrumb into the distinct timeline HWSP and instead can
keep on using the common MI_STORE_DWORD_INDEX. However, to maintain the
sleight-of-hand for the global/per-context seqno switchover, we will
store both temporarily (and so use a custom offset for the shared timeline
HWSP until the switch over).

v3: Keep things simple and allocate a page for each timeline, page
sharing comes next.

v4: I was caught repeating the same MI_STORE_DWORD_IMM over and over
again in selftests.

v5: And caught red handed copying create timeline + check.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190128181812.22804-3-chris@chris-wilson.co.uk

1 files changed, 325 insertions, 1 deletions

diff --git a/drivers/gpu/drm/i915/selftests/i915_timeline.c b/drivers/gpu/drm/i915/selftests/i915_timeline.c
index 19f1c6a5c8fb..1585b614510d 100644
--- a/drivers/gpu/drm/i915/selftests/i915_timeline.c
+++ b/drivers/gpu/drm/i915/selftests/i915_timeline.c
@@ -7,6 +7,7 @@
 #include "../i915_selftest.h"
 #include "i915_random.h"
 
+#include "igt_flush_test.h"
 #include "mock_gem_device.h"
 #include "mock_timeline.h"
 
@@ -256,7 +257,7 @@ static int bench_sync(void *arg)
 	return 0;
 }
 
-int i915_gem_timeline_mock_selftests(void)
+int i915_timeline_mock_selftests(void)
 {
 	static const struct i915_subtest tests[] = {
 		SUBTEST(igt_sync),
@@ -265,3 +266,326 @@ int i915_gem_timeline_mock_selftests(void)
 
 	return i915_subtests(tests, NULL);
 }
+
+static int emit_ggtt_store_dw(struct i915_request *rq, u32 addr, u32 value)
+{
+	u32 *cs;
+
+	cs = intel_ring_begin(rq, 4);
+	if (IS_ERR(cs))
+		return PTR_ERR(cs);
+
+	if (INTEL_GEN(rq->i915) >= 8) {
+		*cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
+		*cs++ = addr;
+		*cs++ = 0;
+		*cs++ = value;
+	} else if (INTEL_GEN(rq->i915) >= 4) {
+		*cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
+		*cs++ = 0;
+		*cs++ = addr;
+		*cs++ = value;
+	} else {
+		*cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL;
+		*cs++ = addr;
+		*cs++ = value;
+		*cs++ = MI_NOOP;
+	}
+
+	intel_ring_advance(rq, cs);
+
+	return 0;
+}
+
+static u32 hwsp_address(const struct i915_timeline *tl)
+{
+	return i915_ggtt_offset(tl->hwsp_ggtt) + tl->hwsp_offset;
+}
+
+static struct i915_request *
+tl_write(struct i915_timeline *tl, struct intel_engine_cs *engine, u32 value)
+{
+	struct i915_request *rq;
+	int err;
+
+	lockdep_assert_held(&tl->i915->drm.struct_mutex); /* lazy rq refs */
+
+	err = i915_timeline_pin(tl);
+	if (err) {
+		rq = ERR_PTR(err);
+		goto out;
+	}
+
+	rq = i915_request_alloc(engine, engine->i915->kernel_context);
+	if (IS_ERR(rq))
+		goto out_unpin;
+
+	err = emit_ggtt_store_dw(rq, hwsp_address(tl), value);
+	i915_request_add(rq);
+	if (err)
+		rq = ERR_PTR(err);
+
+out_unpin:
+	i915_timeline_unpin(tl);
+out:
+	if (IS_ERR(rq))
+		pr_err("Failed to write to timeline!\n");
+	return rq;
+}
+
+static struct i915_timeline *
+checked_i915_timeline_create(struct drm_i915_private *i915)
+{
+	struct i915_timeline *tl;
+
+	tl = i915_timeline_create(i915, "live", NULL);
+	if (IS_ERR(tl))
+		return tl;
+
+	if (*tl->hwsp_seqno != tl->seqno) {
+		pr_err("Timeline created with incorrect breadcrumb, found %x, expected %x\n",
+		       *tl->hwsp_seqno, tl->seqno);
+		i915_timeline_put(tl);
+		return ERR_PTR(-EINVAL);
+	}
+
+	return tl;
+}
+
+static int live_hwsp_engine(void *arg)
+{
+#define NUM_TIMELINES 4096
+	struct drm_i915_private *i915 = arg;
+	struct i915_timeline **timelines;
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	intel_wakeref_t wakeref;
+	unsigned long count, n;
+	int err = 0;
+
+	/*
+	 * Create a bunch of timelines and check we can write
+	 * independently to each of their breadcrumb slots.
+	 */
+
+	timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES,
+				   sizeof(*timelines),
+				   GFP_KERNEL);
+	if (!timelines)
+		return -ENOMEM;
+
+	mutex_lock(&i915->drm.struct_mutex);
+	wakeref = intel_runtime_pm_get(i915);
+
+	count = 0;
+	for_each_engine(engine, i915, id) {
+		if (!intel_engine_can_store_dword(engine))
+			continue;
+
+		for (n = 0; n < NUM_TIMELINES; n++) {
+			struct i915_timeline *tl;
+			struct i915_request *rq;
+
+			tl = checked_i915_timeline_create(i915);
+			if (IS_ERR(tl)) {
+				err = PTR_ERR(tl);
+				goto out;
+			}
+
+			rq = tl_write(tl, engine, count);
+			if (IS_ERR(rq)) {
+				i915_timeline_put(tl);
+				err = PTR_ERR(rq);
+				goto out;
+			}
+
+			timelines[count++] = tl;
+		}
+	}
+
+out:
+	if (igt_flush_test(i915, I915_WAIT_LOCKED))
+		err = -EIO;
+
+	for (n = 0; n < count; n++) {
+		struct i915_timeline *tl = timelines[n];
+
+		if (!err && *tl->hwsp_seqno != n) {
+			pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n",
+			       n, *tl->hwsp_seqno);
+			err = -EINVAL;
+		}
+		i915_timeline_put(tl);
+	}
+
+	intel_runtime_pm_put(i915, wakeref);
+	mutex_unlock(&i915->drm.struct_mutex);
+
+	kvfree(timelines);
+
+	return err;
+#undef NUM_TIMELINES
+}
+
+static int live_hwsp_alternate(void *arg)
+{
+#define NUM_TIMELINES 4096
+	struct drm_i915_private *i915 = arg;
+	struct i915_timeline **timelines;
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	intel_wakeref_t wakeref;
+	unsigned long count, n;
+	int err = 0;
+
+	/*
+	 * Create a bunch of timelines and check we can write
+	 * independently to each of their breadcrumb slots with adjacent
+	 * engines.
+	 */
+
+	timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES,
+				   sizeof(*timelines),
+				   GFP_KERNEL);
+	if (!timelines)
+		return -ENOMEM;
+
+	mutex_lock(&i915->drm.struct_mutex);
+	wakeref = intel_runtime_pm_get(i915);
+
+	count = 0;
+	for (n = 0; n < NUM_TIMELINES; n++) {
+		for_each_engine(engine, i915, id) {
+			struct i915_timeline *tl;
+			struct i915_request *rq;
+
+			if (!intel_engine_can_store_dword(engine))
+				continue;
+
+			tl = checked_i915_timeline_create(i915);
+			if (IS_ERR(tl)) {
+				err = PTR_ERR(tl);
+				goto out;
+			}
+
+			rq = tl_write(tl, engine, count);
+			if (IS_ERR(rq)) {
+				i915_timeline_put(tl);
+				err = PTR_ERR(rq);
+				goto out;
+			}
+
+			timelines[count++] = tl;
+		}
+	}
+
+out:
+	if (igt_flush_test(i915, I915_WAIT_LOCKED))
+		err = -EIO;
+
+	for (n = 0; n < count; n++) {
+		struct i915_timeline *tl = timelines[n];
+
+		if (!err && *tl->hwsp_seqno != n) {
+			pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n",
+			       n, *tl->hwsp_seqno);
+			err = -EINVAL;
+		}
+		i915_timeline_put(tl);
+	}
+
+	intel_runtime_pm_put(i915, wakeref);
+	mutex_unlock(&i915->drm.struct_mutex);
+
+	kvfree(timelines);
+
+	return err;
+#undef NUM_TIMELINES
+}
+
+static int live_hwsp_recycle(void *arg)
+{
+	struct drm_i915_private *i915 = arg;
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	intel_wakeref_t wakeref;
+	unsigned long count;
+	int err = 0;
+
+	/*
+	 * Check seqno writes into one timeline at a time. We expect to
+	 * recycle the breadcrumb slot between iterations and neither
+	 * want to confuse ourselves or the GPU.
+	 */
+
+	mutex_lock(&i915->drm.struct_mutex);
+	wakeref = intel_runtime_pm_get(i915);
+
+	count = 0;
+	for_each_engine(engine, i915, id) {
+		IGT_TIMEOUT(end_time);
+
+		if (!intel_engine_can_store_dword(engine))
+			continue;
+
+		do {
+			struct i915_timeline *tl;
+			struct i915_request *rq;
+
+			tl = checked_i915_timeline_create(i915);
+			if (IS_ERR(tl)) {
+				err = PTR_ERR(tl);
+				goto out;
+			}
+
+			rq = tl_write(tl, engine, count);
+			if (IS_ERR(rq)) {
+				i915_timeline_put(tl);
+				err = PTR_ERR(rq);
+				goto out;
+			}
+
+			if (i915_request_wait(rq,
+					      I915_WAIT_LOCKED,
+					      HZ / 5) < 0) {
+				pr_err("Wait for timeline writes timed out!\n");
+				i915_timeline_put(tl);
+				err = -EIO;
+				goto out;
+			}
+
+			if (*tl->hwsp_seqno != count) {
+				pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n",
+				       count, *tl->hwsp_seqno);
+				err = -EINVAL;
+			}
+
+			i915_timeline_put(tl);
+			count++;
+
+			if (err)
+				goto out;
+
+			i915_timelines_park(i915); /* Encourage recycling! */
+		} while (!__igt_timeout(end_time, NULL));
+	}
+
+out:
+	if (igt_flush_test(i915, I915_WAIT_LOCKED))
+		err = -EIO;
+	intel_runtime_pm_put(i915, wakeref);
+	mutex_unlock(&i915->drm.struct_mutex);
+
+	return err;
+}
+
+int i915_timeline_live_selftests(struct drm_i915_private *i915)
+{
+	static const struct i915_subtest tests[] = {
+		SUBTEST(live_hwsp_recycle),
+		SUBTEST(live_hwsp_engine),
+		SUBTEST(live_hwsp_alternate),
+	};
+
+	return i915_subtests(tests, i915);
+}