drm/i915/gvt: Use vgpu_lock to protect per vgpu access

The patch set splits out 2 small locks from the original big gvt lock:
  - vgpu_lock protects per-vGPU data and logic, especially the vGPU
    trap emulation path.
  - sched_lock protects gvt scheudler structure, context schedule logic
    and vGPU's schedule data.

Use vgpu_lock to replace the gvt big lock. By doing this, the
mmio read/write trap path, vgpu virtual event emulation and other
vgpu related process, would be protected under per vgpu_lock.

v9:
  - Change commit author since the patches are improved a lot compared
    with original version.
    Original author: Pei Zhang <pei.zhang@intel.com>
  - Rebase to latest gvt-staging.
v8:
  - Correct coding and comment style.
  - Rebase to latest gvt-staging.
v7:
  - Remove gtt_lock since already proteced by gvt_lock and vgpu_lock.
  - Fix a typo in intel_gvt_deactivate_vgpu, unlock the wrong lock.
v6:
  - Rebase to latest gvt-staging.
v5:
  - Rebase to latest gvt-staging.
  - intel_vgpu_page_track_handler should use vgpu_lock.
v4:
  - Rebase to latest gvt-staging.
  - Protect vgpu->active access with vgpu_lock.
  - Do not wait gpu idle in vgpu_lock.
v3: update to latest code base
v2: add gvt->lock in function gvt_check_vblank_emulation

Performance comparison on Kabylake platform.
  - Configuration:
    Host: Ubuntu 16.04.
    Guest 1 & 2: Ubuntu 16.04.

glmark2 score comparison:
  - Configuration:
    Host: glxgears.
    Guests: glmark2.
+--------------------------------+-----------------+
| Setup                          | glmark2 score   |
+--------------------------------+-----------------+
| unified lock, iommu=on         | 58~62 (avg. 60) |
+--------------------------------+-----------------+
| unified lock, iommu=igfx_off   | 57~61 (avg. 59) |
+--------------------------------+-----------------+
| per-logic lock, iommu=on       | 60~68 (avg. 64) |
+--------------------------------+-----------------+
| per-logic lock, iommu=igfx_off | 61~67 (avg. 64) |
+--------------------------------+-----------------+

lock_stat comparison:
  - Configuration:
    Stop lock stat immediately after boot up.
    Boot 2 VM Guests.
    Run glmark2 in guests.
    Start perf lock_stat for 20 seconds and stop again.
  - Legend: c - contentions; w - waittime-avg
+------------+-----------------+-----------+---------------+------------+
|            | gvt_lock        |sched_lock | vgpu_lock     | gtt_lock   |
+ lock type; +-----------------+-----------+---------------+------------+
| iommu set  | c     | w       | c  | w    | c    | w      | c   | w    |
+------------+-------+---------+----+------+------+--------+-----+------+
| unified;   | 20697 | 839     |N/A | N/A  | N/A  | N/A    | N/A | N/A  |
| on         |       |         |    |      |      |        |     |      |
+------------+-------+---------+----+------+------+--------+-----+------+
| unified;   | 21838 | 658.15  |N/A | N/A  | N/A  | N/A    | N/A | N/A  |
| igfx_off   |       |         |    |      |      |        |     |      |
+------------+-------+---------+----+------+------+--------+-----+------+
| per-logic; | 1553  | 1599.96 |9458|429.97| 5846 | 274.33 | 0   | 0.00 |
| on         |       |         |    |      |      |        |     |      |
+------------+-------+---------+----+------+------+--------+-----+------+
| per-logic; | 1911  | 1678.32 |8335|445.16| 5451 | 244.80 | 0   | 0.00 |
| igfx_off   |       |         |    |      |      |        |     |      |
+------------+-------+---------+----+------+------+--------+-----+------+

Signed-off-by: Pei Zhang <pei.zhang@intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

1 files changed, 27 insertions, 26 deletions

diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c
index bf75300c1ec1..889d10f8ee96 100644
--- a/drivers/gpu/drm/i915/gvt/vgpu.c
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -226,22 +226,20 @@ void intel_gvt_activate_vgpu(struct intel_vgpu *vgpu)
  */
 void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu)
 {
-	struct intel_gvt *gvt = vgpu->gvt;
-
-	mutex_lock(&gvt->lock);
+	mutex_lock(&vgpu->vgpu_lock);
 
 	vgpu->active = false;
 
 	if (atomic_read(&vgpu->submission.running_workload_num)) {
-		mutex_unlock(&gvt->lock);
+		mutex_unlock(&vgpu->vgpu_lock);
 		intel_gvt_wait_vgpu_idle(vgpu);
-		mutex_lock(&gvt->lock);
+		mutex_lock(&vgpu->vgpu_lock);
 	}
 
 	intel_vgpu_stop_schedule(vgpu);
 	intel_vgpu_dmabuf_cleanup(vgpu);
 
-	mutex_unlock(&gvt->lock);
+	mutex_unlock(&vgpu->vgpu_lock);
 }
 
 /**
@@ -255,14 +253,11 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 {
 	struct intel_gvt *gvt = vgpu->gvt;
 
-	mutex_lock(&gvt->lock);
+	mutex_lock(&vgpu->vgpu_lock);
 
 	WARN(vgpu->active, "vGPU is still active!\n");
 
 	intel_gvt_debugfs_remove_vgpu(vgpu);
-	idr_remove(&gvt->vgpu_idr, vgpu->id);
-	if (idr_is_empty(&gvt->vgpu_idr))
-		intel_gvt_clean_irq(gvt);
 	intel_vgpu_clean_sched_policy(vgpu);
 	intel_vgpu_clean_submission(vgpu);
 	intel_vgpu_clean_display(vgpu);
@@ -272,10 +267,16 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu)
 	intel_vgpu_free_resource(vgpu);
 	intel_vgpu_clean_mmio(vgpu);
 	intel_vgpu_dmabuf_cleanup(vgpu);
-	vfree(vgpu);
+	mutex_unlock(&vgpu->vgpu_lock);
 
+	mutex_lock(&gvt->lock);
+	idr_remove(&gvt->vgpu_idr, vgpu->id);
+	if (idr_is_empty(&gvt->vgpu_idr))
+		intel_gvt_clean_irq(gvt);
 	intel_gvt_update_vgpu_types(gvt);
 	mutex_unlock(&gvt->lock);
+
+	vfree(vgpu);
 }
 
 #define IDLE_VGPU_IDR 0
@@ -301,6 +302,7 @@ struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt)
 
 	vgpu->id = IDLE_VGPU_IDR;
 	vgpu->gvt = gvt;
+	mutex_init(&vgpu->vgpu_lock);
 
 	for (i = 0; i < I915_NUM_ENGINES; i++)
 		INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]);
@@ -327,7 +329,10 @@ out_free_vgpu:
  */
 void intel_gvt_destroy_idle_vgpu(struct intel_vgpu *vgpu)
 {
+	mutex_lock(&vgpu->vgpu_lock);
 	intel_vgpu_clean_sched_policy(vgpu);
+	mutex_unlock(&vgpu->vgpu_lock);
+
 	vfree(vgpu);
 }
 
@@ -345,8 +350,6 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	if (!vgpu)
 		return ERR_PTR(-ENOMEM);
 
-	mutex_lock(&gvt->lock);
-
 	ret = idr_alloc(&gvt->vgpu_idr, vgpu, IDLE_VGPU_IDR + 1, GVT_MAX_VGPU,
 		GFP_KERNEL);
 	if (ret < 0)
@@ -356,6 +359,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	vgpu->handle = param->handle;
 	vgpu->gvt = gvt;
 	vgpu->sched_ctl.weight = param->weight;
+	mutex_init(&vgpu->vgpu_lock);
 	INIT_LIST_HEAD(&vgpu->dmabuf_obj_list_head);
 	INIT_RADIX_TREE(&vgpu->page_track_tree, GFP_KERNEL);
 	idr_init(&vgpu->object_idr);
@@ -403,8 +407,6 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	if (ret)
 		goto out_clean_sched_policy;
 
-	mutex_unlock(&gvt->lock);
-
 	return vgpu;
 
 out_clean_sched_policy:
@@ -427,7 +429,6 @@ out_clean_idr:
 	idr_remove(&gvt->vgpu_idr, vgpu->id);
 out_free_vgpu:
 	vfree(vgpu);
-	mutex_unlock(&gvt->lock);
 	return ERR_PTR(ret);
 }
 
@@ -459,12 +460,12 @@ struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
 	param.low_gm_sz = BYTES_TO_MB(param.low_gm_sz);
 	param.high_gm_sz = BYTES_TO_MB(param.high_gm_sz);
 
+	mutex_lock(&gvt->lock);
 	vgpu = __intel_gvt_create_vgpu(gvt, &param);
-	if (IS_ERR(vgpu))
-		return vgpu;
-
-	/* calculate left instance change for types */
-	intel_gvt_update_vgpu_types(gvt);
+	if (!IS_ERR(vgpu))
+		/* calculate left instance change for types */
+		intel_gvt_update_vgpu_types(gvt);
+	mutex_unlock(&gvt->lock);
 
 	return vgpu;
 }
@@ -476,7 +477,7 @@ struct intel_vgpu *intel_gvt_create_vgpu(struct intel_gvt *gvt,
  * @engine_mask: engines to reset for GT reset
  *
  * This function is called when user wants to reset a virtual GPU through
- * device model reset or GT reset. The caller should hold the gvt lock.
+ * device model reset or GT reset. The caller should hold the vgpu lock.
  *
  * vGPU Device Model Level Reset (DMLR) simulates the PCI level reset to reset
  * the whole vGPU to default state as when it is created. This vGPU function
@@ -516,9 +517,9 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
 	 * scheduler when the reset is triggered by current vgpu.
 	 */
 	if (scheduler->current_vgpu == NULL) {
-		mutex_unlock(&gvt->lock);
+		mutex_unlock(&vgpu->vgpu_lock);
 		intel_gvt_wait_vgpu_idle(vgpu);
-		mutex_lock(&gvt->lock);
+		mutex_lock(&vgpu->vgpu_lock);
 	}
 
 	intel_vgpu_reset_submission(vgpu, resetting_eng);
@@ -558,7 +559,7 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr,
  */
 void intel_gvt_reset_vgpu(struct intel_vgpu *vgpu)
 {
-	mutex_lock(&vgpu->gvt->lock);
+	mutex_lock(&vgpu->vgpu_lock);
 	intel_gvt_reset_vgpu_locked(vgpu, true, 0);
-	mutex_unlock(&vgpu->gvt->lock);
+	mutex_unlock(&vgpu->vgpu_lock);
 }