[PATCH i-g-t 1/3] lib/dummyload: Add pollable spin batch

* [PATCH i-g-t 1/3] lib/dummyload: Add pollable spin batch
@ 2018-03-22 17:24 ` Tvrtko Ursulin
  0 siblings, 0 replies; 21+ messages in thread
From: Tvrtko Ursulin @ 2018-03-22 17:24 UTC (permalink / raw)
  To: igt-dev; +Cc: Intel-gfx

From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Pollable spin batch exports a spin->running pointer which can be checked
by dereferencing it to see if the spinner is actually executing on the
GPU.

This is useful for tests which want to make sure they do not proceed with
their next step whilst the spinner is potentially only being processed by
the driver and not actually executing.

Pollable spinner can be created with igt_spin_batch_new_poll or
__igt_spin_batch_new_poll, after which igt_spin_busywait_until_running can
be used to busy wait until it is executing.

v2:
 * Move READ_ONCE to igt_core.
 * Add igt_spin_busywait_until_running. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 benchmarks/gem_wsim.c |   2 -
 lib/igt_core.h        |   2 +
 lib/igt_dummyload.c   | 192 ++++++++++++++++++++++++++++++++++++++++----------
 lib/igt_dummyload.h   |  17 +++++
 lib/igt_gt.c          |   2 +-
 lib/ioctl_wrappers.c  |   2 +-
 lib/ioctl_wrappers.h  |   1 +
 7 files changed, 175 insertions(+), 43 deletions(-)

diff --git a/benchmarks/gem_wsim.c b/benchmarks/gem_wsim.c
index c15dc365ea95..57dec7b5de34 100644
--- a/benchmarks/gem_wsim.c
+++ b/benchmarks/gem_wsim.c
@@ -980,8 +980,6 @@ current_seqno(struct workload *wrk, enum intel_engine_id engine)
 		return wrk->seqno[engine];
 }
 
-#define READ_ONCE(x) (*(volatile typeof(x) *)(&(x)))
-
 static uint32_t
 read_status_page(struct workload *wrk, unsigned int idx)
 {
diff --git a/lib/igt_core.h b/lib/igt_core.h
index 66523a208c31..57e97f2bb5f4 100644
--- a/lib/igt_core.h
+++ b/lib/igt_core.h
@@ -950,4 +950,6 @@ void igt_kmsg(const char *format, ...);
 #define KMSG_INFO	"<6>[IGT] "
 #define KMSG_DEBUG	"<7>[IGT] "
 
+#define READ_ONCE(x) (*(volatile typeof(x) *)(&(x)))
+
 #endif /* IGT_CORE_H */
diff --git a/lib/igt_dummyload.c b/lib/igt_dummyload.c
index dbc92e8f2951..98ab7ac2c6e9 100644
--- a/lib/igt_dummyload.c
+++ b/lib/igt_dummyload.c
@@ -74,35 +74,48 @@ fill_reloc(struct drm_i915_gem_relocation_entry *reloc,
 	reloc->write_domain = write_domains;
 }
 
-static int emit_recursive_batch(igt_spin_t *spin,
-				int fd, uint32_t ctx, unsigned engine,
-				uint32_t dep, bool out_fence)
+#define OUT_FENCE	(1 << 0)
+#define POLL_RUN	(1 << 1)
+
+static int
+emit_recursive_batch(igt_spin_t *spin, int fd, uint32_t ctx, unsigned engine,
+		     uint32_t dep, unsigned int flags)
 {
 #define SCRATCH 0
 #define BATCH 1
 	const int gen = intel_gen(intel_get_drm_devid(fd));
-	struct drm_i915_gem_exec_object2 obj[2];
-	struct drm_i915_gem_relocation_entry relocs[2];
-	struct drm_i915_gem_execbuffer2 execbuf;
+	struct drm_i915_gem_relocation_entry relocs[2], *r;
+	struct drm_i915_gem_execbuffer2 *execbuf;
+	struct drm_i915_gem_exec_object2 *obj;
 	unsigned int engines[16];
 	unsigned int nengine;
 	int fence_fd = -1;
-	uint32_t *batch;
+	uint32_t *batch, *batch_start;
 	int i;
 
 	nengine = 0;
 	if (engine == ALL_ENGINES) {
-		for_each_engine(fd, engine)
-			if (engine)
+		for_each_engine(fd, engine) {
+			if (engine) {
+			if (flags & POLL_RUN)
+				igt_require(!(flags & POLL_RUN) ||
+					    gem_can_store_dword(fd, engine));
+
 				engines[nengine++] = engine;
+			}
+		}
 	} else {
 		gem_require_ring(fd, engine);
+		igt_require(!(flags & POLL_RUN) ||
+			    gem_can_store_dword(fd, engine));
 		engines[nengine++] = engine;
 	}
 	igt_require(nengine);
 
-	memset(&execbuf, 0, sizeof(execbuf));
-	memset(obj, 0, sizeof(obj));
+	memset(&spin->execbuf, 0, sizeof(spin->execbuf));
+	execbuf = &spin->execbuf;
+	memset(spin->obj, 0, sizeof(spin->obj));
+	obj = spin->obj;
 	memset(relocs, 0, sizeof(relocs));
 
 	obj[BATCH].handle = gem_create(fd, BATCH_SIZE);
@@ -113,19 +126,66 @@ static int emit_recursive_batch(igt_spin_t *spin,
 				       	BATCH_SIZE, PROT_WRITE);
 	gem_set_domain(fd, obj[BATCH].handle,
 			I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT);
-	execbuf.buffer_count++;
+	execbuf->buffer_count++;
+	batch_start = batch;
 
 	if (dep) {
+		igt_assert(!(flags & POLL_RUN));
+
 		/* dummy write to dependency */
 		obj[SCRATCH].handle = dep;
 		fill_reloc(&relocs[obj[BATCH].relocation_count++],
 			   dep, 1020,
 			   I915_GEM_DOMAIN_RENDER,
 			   I915_GEM_DOMAIN_RENDER);
-		execbuf.buffer_count++;
+		execbuf->buffer_count++;
+	} else if (flags & POLL_RUN) {
+		unsigned int offset;
+
+		igt_assert(!dep);
+
+		if (gen == 4 || gen == 5)
+			execbuf->flags |= I915_EXEC_SECURE;
+
+		spin->poll_handle = gem_create(fd, 4096);
+
+		if (__gem_set_caching(fd, spin->poll_handle,
+				      I915_CACHING_CACHED) == 0)
+			spin->running = __gem_mmap__cpu(fd, spin->poll_handle,
+							0, 4096,
+							PROT_READ | PROT_WRITE);
+		else
+			spin->running = __gem_mmap__wc(fd, spin->poll_handle,
+						       0, 4096,
+						       PROT_READ | PROT_WRITE);
+		igt_assert(spin->running);
+		igt_assert_eq(*spin->running, 0);
+
+		*batch++ = MI_STORE_DWORD_IMM | (gen < 6 ? 1 << 22 : 0);
+
+		if (gen >= 8) {
+			offset = 1;
+			*batch++ = 0;
+			*batch++ = 0;
+		} else if (gen >= 4) {
+			offset = 2;
+			*batch++ = 0;
+			*batch++ = 0;
+		} else {
+			offset = 1;
+			batch[-1]--;
+			*batch++ = 0;
+		}
+
+		*batch++ = 1;
+
+		obj[SCRATCH].handle = spin->poll_handle;
+		fill_reloc(&relocs[obj[BATCH].relocation_count++],
+			   spin->poll_handle, offset, 0, 0);
+		execbuf->buffer_count++;
 	}
 
-	spin->batch = batch;
+	spin->batch = batch = batch_start + 64 / sizeof(*batch);
 	spin->handle = obj[BATCH].handle;
 
 	/* Allow ourselves to be preempted */
@@ -145,40 +205,42 @@ static int emit_recursive_batch(igt_spin_t *spin,
 	batch += 1000;
 
 	/* recurse */
-	fill_reloc(&relocs[obj[BATCH].relocation_count],
-		   obj[BATCH].handle, (batch - spin->batch) + 1,
-		   I915_GEM_DOMAIN_COMMAND, 0);
+	r = &relocs[obj[BATCH].relocation_count++];
+	r->target_handle = obj[BATCH].handle;
+	r->offset = (batch + 1 - batch_start) * sizeof(*batch);
+	r->read_domains = I915_GEM_DOMAIN_COMMAND;
+	r->delta = 64;
 	if (gen >= 8) {
 		*batch++ = MI_BATCH_BUFFER_START | 1 << 8 | 1;
-		*batch++ = 0;
+		*batch++ = r->delta;
 		*batch++ = 0;
 	} else if (gen >= 6) {
 		*batch++ = MI_BATCH_BUFFER_START | 1 << 8;
-		*batch++ = 0;
+		*batch++ = r->delta;
 	} else {
 		*batch++ = MI_BATCH_BUFFER_START | 2 << 6;
-		*batch = 0;
-		if (gen < 4) {
-			*batch |= 1;
-			relocs[obj[BATCH].relocation_count].delta = 1;
-		}
+		if (gen < 4)
+			r->delta |= 1;
+		*batch = r->delta;
 		batch++;
 	}
-	obj[BATCH].relocation_count++;
 	obj[BATCH].relocs_ptr = to_user_pointer(relocs);
 
-	execbuf.buffers_ptr = to_user_pointer(obj + (2 - execbuf.buffer_count));
-	execbuf.rsvd1 = ctx;
+	execbuf->buffers_ptr = to_user_pointer(obj +
+					       (2 - execbuf->buffer_count));
+	execbuf->rsvd1 = ctx;
 
-	if (out_fence)
-		execbuf.flags |= I915_EXEC_FENCE_OUT;
+	if (flags & OUT_FENCE)
+		execbuf->flags |= I915_EXEC_FENCE_OUT;
 
 	for (i = 0; i < nengine; i++) {
-		execbuf.flags &= ~ENGINE_MASK;
-		execbuf.flags |= engines[i];
-		gem_execbuf_wr(fd, &execbuf);
-		if (out_fence) {
-			int _fd = execbuf.rsvd2 >> 32;
+		execbuf->flags &= ~ENGINE_MASK;
+		execbuf->flags |= engines[i];
+
+		gem_execbuf_wr(fd, execbuf);
+
+		if (flags & OUT_FENCE) {
+			int _fd = execbuf->rsvd2 >> 32;
 
 			igt_assert(_fd >= 0);
 			if (fence_fd == -1) {
@@ -194,12 +256,20 @@ static int emit_recursive_batch(igt_spin_t *spin,
 		}
 	}
 
+	/* Make it easier for callers to resubmit. */
+
+	obj[BATCH].relocation_count = 0;
+	obj[BATCH].relocs_ptr = 0;
+
+	obj[SCRATCH].flags = EXEC_OBJECT_PINNED;
+	obj[BATCH].flags = EXEC_OBJECT_PINNED;
+
 	return fence_fd;
 }
 
 static igt_spin_t *
 ___igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep,
-		      int out_fence)
+		      unsigned int flags)
 {
 	igt_spin_t *spin;
 
@@ -207,7 +277,7 @@ ___igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep,
 	igt_assert(spin);
 
 	spin->out_fence = emit_recursive_batch(spin, fd, ctx, engine, dep,
-					       out_fence);
+					       flags);
 
 	pthread_mutex_lock(&list_lock);
 	igt_list_add(&spin->link, &spin_list);
@@ -219,7 +289,7 @@ ___igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep,
 igt_spin_t *
 __igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep)
 {
-	return ___igt_spin_batch_new(fd, ctx, engine, dep, false);
+	return ___igt_spin_batch_new(fd, ctx, engine, dep, 0);
 }
 
 /**
@@ -253,7 +323,7 @@ igt_spin_batch_new(int fd, uint32_t ctx, unsigned engine, uint32_t dep)
 igt_spin_t *
 __igt_spin_batch_new_fence(int fd, uint32_t ctx, unsigned engine)
 {
-	return ___igt_spin_batch_new(fd, ctx, engine, 0, true);
+	return ___igt_spin_batch_new(fd, ctx, engine, 0, OUT_FENCE);
 }
 
 /**
@@ -286,6 +356,42 @@ igt_spin_batch_new_fence(int fd, uint32_t ctx, unsigned engine)
 	return spin;
 }
 
+igt_spin_t *
+__igt_spin_batch_new_poll(int fd, uint32_t ctx, unsigned engine)
+{
+	return ___igt_spin_batch_new(fd, ctx, engine, 0, POLL_RUN);
+}
+
+/**
+ * igt_spin_batch_new_poll:
+ * @fd: open i915 drm file descriptor
+ * @engine: Ring to execute batch OR'd with execbuf flags. If value is less
+ *          than 0, execute on all available rings.
+ *
+ * Start a recursive batch on a ring. Immediately returns a #igt_spin_t that
+ * contains the batch's handle that can be waited upon. The returned structure
+ * must be passed to igt_spin_batch_free() for post-processing.
+ *
+ * igt_spin_t->running will containt a pointer which target will change from
+ * zero to one once the spinner actually starts executing on the GPU.
+ *
+ * Returns:
+ * Structure with helper internal state for igt_spin_batch_free().
+ */
+igt_spin_t *
+igt_spin_batch_new_poll(int fd, uint32_t ctx, unsigned engine)
+{
+	igt_spin_t *spin;
+
+	igt_require_gem(fd);
+	igt_require(gem_mmap__has_wc(fd));
+
+	spin = __igt_spin_batch_new_poll(fd, ctx, engine);
+	igt_assert(gem_bo_busy(fd, spin->handle));
+
+	return spin;
+}
+
 static void notify(union sigval arg)
 {
 	igt_spin_t *spin = arg.sival_ptr;
@@ -340,6 +446,8 @@ void igt_spin_batch_end(igt_spin_t *spin)
 	if (!spin)
 		return;
 
+	igt_assert(*spin->batch == MI_ARB_CHK ||
+		   *spin->batch == MI_BATCH_BUFFER_END);
 	*spin->batch = MI_BATCH_BUFFER_END;
 	__sync_synchronize();
 }
@@ -365,7 +473,13 @@ void igt_spin_batch_free(int fd, igt_spin_t *spin)
 		timer_delete(spin->timer);
 
 	igt_spin_batch_end(spin);
-	gem_munmap(spin->batch, BATCH_SIZE);
+	gem_munmap((void *)((unsigned long)spin->batch & (~4095UL)),
+		   BATCH_SIZE);
+
+	if (spin->running) {
+		gem_munmap(spin->running, 4096);
+		gem_close(fd, spin->poll_handle);
+	}
 
 	gem_close(fd, spin->handle);
 
diff --git a/lib/igt_dummyload.h b/lib/igt_dummyload.h
index 4103e4ab9e36..a8ec213fe8f3 100644
--- a/lib/igt_dummyload.h
+++ b/lib/igt_dummyload.h
@@ -36,6 +36,10 @@ typedef struct igt_spin {
 	struct igt_list link;
 	uint32_t *batch;
 	int out_fence;
+	struct drm_i915_gem_exec_object2 obj[2];
+	struct drm_i915_gem_execbuffer2 execbuf;
+	uint32_t poll_handle;
+	bool *running;
 } igt_spin_t;
 
 igt_spin_t *__igt_spin_batch_new(int fd,
@@ -55,10 +59,23 @@ igt_spin_t *igt_spin_batch_new_fence(int fd,
 				     uint32_t ctx,
 				     unsigned engine);
 
+igt_spin_t *__igt_spin_batch_new_poll(int fd,
+				       uint32_t ctx,
+				       unsigned engine);
+igt_spin_t *igt_spin_batch_new_poll(int fd,
+				    uint32_t ctx,
+				    unsigned engine);
+
 void igt_spin_batch_set_timeout(igt_spin_t *spin, int64_t ns);
 void igt_spin_batch_end(igt_spin_t *spin);
 void igt_spin_batch_free(int fd, igt_spin_t *spin);
 
+static inline void igt_spin_busywait_until_running(igt_spin_t *spin)
+{
+	while (!READ_ONCE(*spin->running))
+		;
+}
+
 void igt_terminate_spin_batches(void);
 
 enum igt_cork_type {
diff --git a/lib/igt_gt.c b/lib/igt_gt.c
index 01aebc670862..4569fd36bd85 100644
--- a/lib/igt_gt.c
+++ b/lib/igt_gt.c
@@ -609,7 +609,7 @@ bool gem_can_store_dword(int fd, unsigned int engine)
 	if (gen == 3 && (info->is_grantsdale || info->is_alviso))
 		return false; /* only supports physical addresses */
 
-	if (gen == 6 && (engine & ~(3<<13)) == I915_EXEC_BSD)
+	if (gen == 6 && ((engine & 0x3f) == I915_EXEC_BSD))
 		return false; /* kills the machine! */
 
 	if (info->is_broadwater)
diff --git a/lib/ioctl_wrappers.c b/lib/ioctl_wrappers.c
index 8748cfcfc04f..4e1a08bf06b4 100644
--- a/lib/ioctl_wrappers.c
+++ b/lib/ioctl_wrappers.c
@@ -198,7 +198,7 @@ void gem_set_tiling(int fd, uint32_t handle, uint32_t tiling, uint32_t stride)
 	igt_assert(__gem_set_tiling(fd, handle, tiling, stride) == 0);
 }
 
-static int __gem_set_caching(int fd, uint32_t handle, uint32_t caching)
+int __gem_set_caching(int fd, uint32_t handle, uint32_t caching)
 {
 	struct drm_i915_gem_caching arg;
 	int err;
diff --git a/lib/ioctl_wrappers.h b/lib/ioctl_wrappers.h
index 13fbe3c103c0..b966f72c90a8 100644
--- a/lib/ioctl_wrappers.h
+++ b/lib/ioctl_wrappers.h
@@ -61,6 +61,7 @@ bool gem_get_tiling(int fd, uint32_t handle, uint32_t *tiling, uint32_t *swizzle
 void gem_set_tiling(int fd, uint32_t handle, uint32_t tiling, uint32_t stride);
 int __gem_set_tiling(int fd, uint32_t handle, uint32_t tiling, uint32_t stride);
 
+int __gem_set_caching(int fd, uint32_t handle, uint32_t caching);
 void gem_set_caching(int fd, uint32_t handle, uint32_t caching);
 uint32_t gem_get_caching(int fd, uint32_t handle);
 uint32_t gem_flink(int fd, uint32_t handle);
-- 
2.14.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 21+ messages in thread