From f2085c8e950d536c7982182b0c9c015804011dd2 Mon Sep 17 00:00:00 2001
From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue, 27 Aug 2019 17:17:25 +0100
Subject: drm/i915/selftests: Remove accidental serialization between gpu_fill
Git-commit: f2085c8e950d536c7982182b0c9c015804011dd2
Patch-mainline: v5.5-rc1
References: bsc#1152489
Upon object creation for live_gem_contexts, we fill the object with
known scratch and flush it out of the CPU cache. Before performing the
GPU fill, we don't need to flush it again and so avoid serialising with
previous fills.
However, we do need some throttling on the internal interfaces if we do
not want to run out of memory!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190827161726.3640-1-chris@chris-wilson.co.uk
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
---
.../drm/i915/gem/selftests/i915_gem_context.c | 83 ++++++++++++++++---
1 file changed, 72 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index 37a177e37665..63116c4fa8ba 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -180,12 +180,6 @@ static int gpu_fill(struct intel_context *ce,
if (IS_ERR(vma))
return PTR_ERR(vma);
- i915_gem_object_lock(obj);
- err = i915_gem_object_set_to_gtt_domain(obj, true);
- i915_gem_object_unlock(obj);
- if (err)
- return err;
-
err = i915_vma_pin(vma, 0, 0, PIN_HIGH | PIN_USER);
if (err)
return err;
@@ -343,6 +337,45 @@ static unsigned long max_dwords(struct drm_i915_gem_object *obj)
return npages / DW_PER_PAGE;
}
+static void throttle_release(struct i915_request **q, int count)
+{
+ int i;
+
+ for (i = 0; i < count; i++) {
+ if (IS_ERR_OR_NULL(q[i]))
+ continue;
+
+ i915_request_put(fetch_and_zero(&q[i]));
+ }
+}
+
+static int throttle(struct intel_context *ce,
+ struct i915_request **q, int count)
+{
+ int i;
+
+ if (!IS_ERR_OR_NULL(q[0])) {
+ if (i915_request_wait(q[0],
+ I915_WAIT_INTERRUPTIBLE,
+ MAX_SCHEDULE_TIMEOUT) < 0)
+ return -EINTR;
+
+ i915_request_put(q[0]);
+ }
+
+ for (i = 0; i < count - 1; i++)
+ q[i] = q[i + 1];
+
+ q[i] = intel_context_create_request(ce);
+ if (IS_ERR(q[i]))
+ return PTR_ERR(q[i]);
+
+ i915_request_get(q[i]);
+ i915_request_add(q[i]);
+
+ return 0;
+}
+
static int igt_ctx_exec(void *arg)
{
struct drm_i915_private *i915 = arg;
@@ -362,6 +395,7 @@ static int igt_ctx_exec(void *arg)
for_each_engine(engine, i915, id) {
struct drm_i915_gem_object *obj = NULL;
unsigned long ncontexts, ndwords, dw;
+ struct i915_request *tq[5] = {};
struct igt_live_test t;
struct drm_file *file;
IGT_TIMEOUT(end_time);
@@ -409,13 +443,18 @@ static int igt_ctx_exec(void *arg)
}
err = gpu_fill(ce, obj, dw);
- intel_context_put(ce);
-
if (err) {
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name, ctx->hw_id,
yesno(!!ctx->vm), err);
+ intel_context_put(ce);
+ goto out_unlock;
+ }
+
+ err = throttle(ce, tq, ARRAY_SIZE(tq));
+ if (err) {
+ intel_context_put(ce);
goto out_unlock;
}
@@ -426,6 +465,8 @@ static int igt_ctx_exec(void *arg)
ndwords++;
ncontexts++;
+
+ intel_context_put(ce);
}
pr_info("Submitted %lu contexts to %s, filling %lu dwords\n",
@@ -444,6 +485,7 @@ static int igt_ctx_exec(void *arg)
}
out_unlock:
+ throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
mutex_unlock(&i915->drm.struct_mutex);
@@ -461,6 +503,7 @@ static int igt_ctx_exec(void *arg)
static int igt_shared_ctx_exec(void *arg)
{
struct drm_i915_private *i915 = arg;
+ struct i915_request *tq[5] = {};
struct i915_gem_context *parent;
struct intel_engine_cs *engine;
enum intel_engine_id id;
@@ -535,14 +578,20 @@ static int igt_shared_ctx_exec(void *arg)
}
err = gpu_fill(ce, obj, dw);
- intel_context_put(ce);
- kernel_context_close(ctx);
-
if (err) {
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name, ctx->hw_id,
yesno(!!ctx->vm), err);
+ intel_context_put(ce);
+ kernel_context_close(ctx);
+ goto out_test;
+ }
+
+ err = throttle(ce, tq, ARRAY_SIZE(tq));
+ if (err) {
+ intel_context_put(ce);
+ kernel_context_close(ctx);
goto out_test;
}
@@ -553,6 +602,9 @@ static int igt_shared_ctx_exec(void *arg)
ndwords++;
ncontexts++;
+
+ intel_context_put(ce);
+ kernel_context_close(ctx);
}
pr_info("Submitted %lu contexts to %s, filling %lu dwords\n",
ncontexts, engine->name, ndwords);
@@ -574,6 +626,7 @@ static int igt_shared_ctx_exec(void *arg)
mutex_lock(&i915->drm.struct_mutex);
}
out_test:
+ throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
out_unlock:
@@ -1050,6 +1103,7 @@ static int igt_ctx_readonly(void *arg)
{
struct drm_i915_private *i915 = arg;
struct drm_i915_gem_object *obj = NULL;
+ struct i915_request *tq[5] = {};
struct i915_address_space *vm;
struct i915_gem_context *ctx;
unsigned long idx, ndwords, dw;
@@ -1121,6 +1175,12 @@ static int igt_ctx_readonly(void *arg)
goto out_unlock;
}
+ err = throttle(ce, tq, ARRAY_SIZE(tq));
+ if (err) {
+ i915_gem_context_unlock_engines(ctx);
+ goto out_unlock;
+ }
+
if (++dw == max_dwords(obj)) {
obj = NULL;
dw = 0;
@@ -1151,6 +1211,7 @@ static int igt_ctx_readonly(void *arg)
}
out_unlock:
+ throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
mutex_unlock(&i915->drm.struct_mutex);
--
2.28.0