From 1406a14b0ed977fc18f43398b391e4bb5d744174 Mon Sep 17 00:00:00 2001
From: Zhi Wang <zhi.a.wang@intel.com>
Date: Sun, 10 Sep 2017 21:15:18 +0800
Subject: [PATCH] drm/i915/gvt: Introduce intel_vgpu_submission
Git-commit: 1406a14b0ed977fc18f43398b391e4bb5d744174
Patch-mainline: v4.16-rc1
References: FATE#322643 bsc#1055900
Introduce intel_vgpu_submission to hold all members related to submission
in struct intel_vgpu before.
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Acked-by: Takashi Iwai <tiwai@suse.de>
---
drivers/gpu/drm/i915/gvt/execlist.c | 30 +++++++++------
drivers/gpu/drm/i915/gvt/gvt.h | 17 +++++---
drivers/gpu/drm/i915/gvt/handlers.c | 2 -
drivers/gpu/drm/i915/gvt/kvmgt.c | 2 -
drivers/gpu/drm/i915/gvt/render.c | 9 ++--
drivers/gpu/drm/i915/gvt/scheduler.c | 68 +++++++++++++++++++----------------
drivers/gpu/drm/i915/gvt/scheduler.h | 2 -
drivers/gpu/drm/i915/gvt/vgpu.c | 4 +-
8 files changed, 77 insertions(+), 57 deletions(-)
--- a/drivers/gpu/drm/i915/gvt/execlist.c
+++ b/drivers/gpu/drm/i915/gvt/execlist.c
@@ -362,7 +362,7 @@ static void free_workload(struct intel_v
{
intel_vgpu_unpin_mm(workload->shadow_mm);
intel_gvt_mm_unreference(workload->shadow_mm);
- kmem_cache_free(workload->vgpu->workloads, workload);
+ kmem_cache_free(workload->vgpu->submission.workloads, workload);
}
#define get_desc_from_elsp_dwords(ed, i) \
@@ -401,7 +401,8 @@ static int update_wa_ctx_2_shadow_ctx(st
struct intel_vgpu_workload,
wa_ctx);
int ring_id = workload->ring_id;
- struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+ struct intel_vgpu_submission *s = &workload->vgpu->submission;
+ struct i915_gem_context *shadow_ctx = s->shadow_ctx;
struct drm_i915_gem_object *ctx_obj =
shadow_ctx->engine[ring_id].state->obj;
struct execlist_ring_context *shadow_ring_context;
@@ -474,6 +475,7 @@ static void release_shadow_batch_buffer(
static int prepare_execlist_workload(struct intel_vgpu_workload *workload)
{
struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
struct execlist_ctx_descriptor_format ctx[2];
int ring_id = workload->ring_id;
int ret;
@@ -520,7 +522,7 @@ static int prepare_execlist_workload(str
ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);
- ret = emulate_execlist_schedule_in(&vgpu->execlist[ring_id], ctx);
+ ret = emulate_execlist_schedule_in(&s->execlist[ring_id], ctx);
if (!ret)
goto out;
else
@@ -539,7 +541,8 @@ static int complete_execlist_workload(st
{
struct intel_vgpu *vgpu = workload->vgpu;
int ring_id = workload->ring_id;
- struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
struct intel_vgpu_workload *next_workload;
struct list_head *next = workload_q_head(vgpu, ring_id)->next;
bool lite_restore = false;
@@ -658,6 +661,7 @@ static int submit_context(struct intel_v
struct execlist_ctx_descriptor_format *desc,
bool emulate_schedule_in)
{
+ struct intel_vgpu_submission *s = &vgpu->submission;
struct list_head *q = workload_q_head(vgpu, ring_id);
struct intel_vgpu_workload *last_workload = get_last_workload(q);
struct intel_vgpu_workload *workload = NULL;
@@ -695,7 +699,7 @@ static int submit_context(struct intel_v
gvt_dbg_el("ring id %d begin a new workload\n", ring_id);
- workload = kmem_cache_zalloc(vgpu->workloads, GFP_KERNEL);
+ workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL);
if (!workload)
return -ENOMEM;
@@ -744,7 +748,7 @@ static int submit_context(struct intel_v
}
if (emulate_schedule_in)
- workload->elsp_dwords = vgpu->execlist[ring_id].elsp_dwords;
+ workload->elsp_dwords = s->execlist[ring_id].elsp_dwords;
gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n",
workload, ring_id, head, tail, start, ctl);
@@ -754,7 +758,7 @@ static int submit_context(struct intel_v
ret = prepare_mm(workload);
if (ret) {
- kmem_cache_free(vgpu->workloads, workload);
+ kmem_cache_free(s->workloads, workload);
return ret;
}
@@ -775,7 +779,8 @@ static int submit_context(struct intel_v
int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
{
- struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
struct execlist_ctx_descriptor_format *desc[2];
int i, ret;
@@ -817,7 +822,8 @@ inv_desc:
static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
{
- struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id];
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
struct execlist_context_status_pointer_format ctx_status_ptr;
u32 ctx_status_ptr_reg;
@@ -839,6 +845,7 @@ static void init_vgpu_execlist(struct in
static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask)
{
+ struct intel_vgpu_submission *s = &vgpu->submission;
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
struct intel_engine_cs *engine;
struct intel_vgpu_workload *pos, *n;
@@ -847,12 +854,11 @@ static void clean_workloads(struct intel
/* free the unsubmited workloads in the queues. */
for_each_engine_masked(engine, dev_priv, engine_mask, tmp) {
list_for_each_entry_safe(pos, n,
- &vgpu->workload_q_head[engine->id], list) {
+ &s->workload_q_head[engine->id], list) {
list_del_init(&pos->list);
free_workload(pos);
}
-
- clear_bit(engine->id, vgpu->shadow_ctx_desc_updated);
+ clear_bit(engine->id, s->shadow_ctx_desc_updated);
}
}
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -142,6 +142,15 @@ struct vgpu_sched_ctl {
int weight;
};
+struct intel_vgpu_submission {
+ struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
+ struct list_head workload_q_head[I915_NUM_ENGINES];
+ struct kmem_cache *workloads;
+ atomic_t running_workload_num;
+ struct i915_gem_context *shadow_ctx;
+ DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+};
+
struct intel_vgpu {
struct intel_gvt *gvt;
int id;
@@ -161,16 +170,12 @@ struct intel_vgpu {
struct intel_vgpu_gtt gtt;
struct intel_vgpu_opregion opregion;
struct intel_vgpu_display display;
- struct intel_vgpu_execlist execlist[I915_NUM_ENGINES];
- struct list_head workload_q_head[I915_NUM_ENGINES];
- struct kmem_cache *workloads;
- atomic_t running_workload_num;
+ struct intel_vgpu_submission submission;
/* 1/2K for each reserve ring buffer */
void *reserve_ring_buffer_va[I915_NUM_ENGINES];
int reserve_ring_buffer_size[I915_NUM_ENGINES];
DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
- struct i915_gem_context *shadow_ctx;
- DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES);
+
#if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
struct {
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -1444,7 +1444,7 @@ static int elsp_mmio_write(struct intel_
if (WARN_ON(ring_id < 0 || ring_id > I915_NUM_ENGINES - 1))
return -EINVAL;
- execlist = &vgpu->execlist[ring_id];
+ execlist = &vgpu->submission.execlist[ring_id];
execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data;
if (execlist->elsp_dwords.index == 3) {
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -1193,7 +1193,7 @@ hw_id_show(struct device *dev, struct de
struct intel_vgpu *vgpu = (struct intel_vgpu *)
mdev_get_drvdata(mdev);
return sprintf(buf, "%u\n",
- vgpu->shadow_ctx->hw_id);
+ vgpu->submission.shadow_ctx->hw_id);
}
return sprintf(buf, "\n");
}
--- a/drivers/gpu/drm/i915/gvt/render.c
+++ b/drivers/gpu/drm/i915/gvt/render.c
@@ -261,14 +261,15 @@ static void restore_mocs(struct intel_vg
static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id)
{
struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
- struct render_mmio *mmio;
- u32 v;
- int i, array_size;
- u32 *reg_state = vgpu->shadow_ctx->engine[ring_id].lrc_reg_state;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ u32 *reg_state = s->shadow_ctx->engine[ring_id].lrc_reg_state;
u32 ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL];
u32 inhibit_mask =
_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT);
i915_reg_t last_reg = _MMIO(0);
+ struct render_mmio *mmio;
+ u32 v;
+ int i, array_size;
if (IS_SKYLAKE(vgpu->gvt->dev_priv)
|| IS_KABYLAKE(vgpu->gvt->dev_priv)) {
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -57,7 +57,7 @@ static int populate_shadow_context(struc
struct intel_vgpu *vgpu = workload->vgpu;
struct intel_gvt *gvt = vgpu->gvt;
int ring_id = workload->ring_id;
- struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
+ struct i915_gem_context *shadow_ctx = vgpu->submission.shadow_ctx;
struct drm_i915_gem_object *ctx_obj =
shadow_ctx->engine[ring_id].state->obj;
struct execlist_ring_context *shadow_ring_context;
@@ -267,11 +267,12 @@ void release_shadow_wa_ctx(struct intel_
*/
int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload)
{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+ struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
int ring_id = workload->ring_id;
- struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
- struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
struct intel_engine_cs *engine = dev_priv->engine[ring_id];
- struct intel_vgpu *vgpu = workload->vgpu;
struct intel_ring *ring;
int ret;
@@ -284,7 +285,7 @@ int intel_gvt_scan_and_shadow_workload(s
shadow_ctx->desc_template |= workload->ctx_desc.addressing_mode <<
GEN8_CTX_ADDRESSING_MODE_SHIFT;
- if (!test_and_set_bit(ring_id, vgpu->shadow_ctx_desc_updated))
+ if (!test_and_set_bit(ring_id, s->shadow_ctx_desc_updated))
shadow_context_descriptor_update(shadow_ctx,
dev_priv->engine[ring_id]);
@@ -334,7 +335,8 @@ int intel_gvt_generate_request(struct in
struct intel_engine_cs *engine = dev_priv->engine[ring_id];
struct drm_i915_gem_request *rq;
struct intel_vgpu *vgpu = workload->vgpu;
- struct i915_gem_context *shadow_ctx = vgpu->shadow_ctx;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_gem_context *shadow_ctx = s->shadow_ctx;
int ret;
rq = i915_gem_request_alloc(dev_priv->engine[ring_id], shadow_ctx);
@@ -360,9 +362,11 @@ err_unpin:
static int dispatch_workload(struct intel_vgpu_workload *workload)
{
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_gem_context *shadow_ctx = s->shadow_ctx;
+ struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
int ring_id = workload->ring_id;
- struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
- struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv;
struct intel_engine_cs *engine = dev_priv->engine[ring_id];
int ret = 0;
@@ -448,7 +452,7 @@ static struct intel_vgpu_workload *pick_
gvt_dbg_sched("ring id %d pick new workload %p\n", ring_id, workload);
- atomic_inc(&workload->vgpu->running_workload_num);
+ atomic_inc(&workload->vgpu->submission.running_workload_num);
out:
mutex_unlock(&gvt->lock);
return workload;
@@ -458,8 +462,9 @@ static void update_guest_context(struct
{
struct intel_vgpu *vgpu = workload->vgpu;
struct intel_gvt *gvt = vgpu->gvt;
+ struct intel_vgpu_submission *s = &vgpu->submission;
+ struct i915_gem_context *shadow_ctx = s->shadow_ctx;
int ring_id = workload->ring_id;
- struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx;
struct drm_i915_gem_object *ctx_obj =
shadow_ctx->engine[ring_id].state->obj;
struct execlist_ring_context *shadow_ring_context;
@@ -525,15 +530,14 @@ static void update_guest_context(struct
static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
{
struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
- struct intel_vgpu_workload *workload;
- struct intel_vgpu *vgpu;
+ struct intel_vgpu_workload *workload =
+ scheduler->current_workload[ring_id];
+ struct intel_vgpu *vgpu = workload->vgpu;
+ struct intel_vgpu_submission *s = &vgpu->submission;
int event;
mutex_lock(&gvt->lock);
- workload = scheduler->current_workload[ring_id];
- vgpu = workload->vgpu;
-
/* For the workload w/ request, needs to wait for the context
* switch to make sure request is completed.
* For the workload w/o request, directly complete the workload.
@@ -570,7 +574,7 @@ static void complete_current_workload(st
}
mutex_lock(&dev_priv->drm.struct_mutex);
/* unpin shadow ctx as the shadow_ctx update is done */
- engine->context_unpin(engine, workload->vgpu->shadow_ctx);
+ engine->context_unpin(engine, s->shadow_ctx);
mutex_unlock(&dev_priv->drm.struct_mutex);
}
@@ -582,7 +586,7 @@ static void complete_current_workload(st
list_del_init(&workload->list);
workload->complete(workload);
- atomic_dec(&vgpu->running_workload_num);
+ atomic_dec(&s->running_workload_num);
wake_up(&scheduler->workload_complete_wq);
if (gvt->scheduler.need_reschedule)
@@ -671,14 +675,15 @@ complete:
void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu)
{
+ struct intel_vgpu_submission *s = &vgpu->submission;
struct intel_gvt *gvt = vgpu->gvt;
struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
- if (atomic_read(&vgpu->running_workload_num)) {
+ if (atomic_read(&s->running_workload_num)) {
gvt_dbg_sched("wait vgpu idle\n");
wait_event(scheduler->workload_complete_wq,
- !atomic_read(&vgpu->running_workload_num));
+ !atomic_read(&s->running_workload_num));
}
}
@@ -752,8 +757,10 @@ err:
*/
void intel_vgpu_clean_submission(struct intel_vgpu *vgpu)
{
- i915_gem_context_put(vgpu->shadow_ctx);
- kmem_cache_destroy(vgpu->workloads);
+ struct intel_vgpu_submission *s = &vgpu->submission;
+
+ i915_gem_context_put(s->shadow_ctx);
+ kmem_cache_destroy(s->workloads);
}
/**
@@ -768,35 +775,36 @@ void intel_vgpu_clean_submission(struct
*/
int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
{
+ struct intel_vgpu_submission *s = &vgpu->submission;
enum intel_engine_id i;
struct intel_engine_cs *engine;
int ret;
- vgpu->shadow_ctx = i915_gem_context_create_gvt(
+ s->shadow_ctx = i915_gem_context_create_gvt(
&vgpu->gvt->dev_priv->drm);
- if (IS_ERR(vgpu->shadow_ctx))
- return PTR_ERR(vgpu->shadow_ctx);
+ if (IS_ERR(s->shadow_ctx))
+ return PTR_ERR(s->shadow_ctx);
- bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES);
+ bitmap_zero(s->shadow_ctx_desc_updated, I915_NUM_ENGINES);
- vgpu->workloads = kmem_cache_create("gvt-g_vgpu_workload",
+ s->workloads = kmem_cache_create("gvt-g_vgpu_workload",
sizeof(struct intel_vgpu_workload), 0,
SLAB_HWCACHE_ALIGN,
NULL);
- if (!vgpu->workloads) {
+ if (!s->workloads) {
ret = -ENOMEM;
goto out_shadow_ctx;
}
for_each_engine(engine, vgpu->gvt->dev_priv, i)
- INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
+ INIT_LIST_HEAD(&s->workload_q_head[i]);
- atomic_set(&vgpu->running_workload_num, 0);
+ atomic_set(&s->running_workload_num, 0);
return 0;
out_shadow_ctx:
- i915_gem_context_put(vgpu->shadow_ctx);
+ i915_gem_context_put(s->shadow_ctx);
return ret;
}
--- a/drivers/gpu/drm/i915/gvt/scheduler.h
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -122,7 +122,7 @@ struct intel_shadow_bb_entry {
};
#define workload_q_head(vgpu, ring_id) \
- (&(vgpu->workload_q_head[ring_id]))
+ (&(vgpu->submission.workload_q_head[ring_id]))
#define queue_workload(workload) do { \
list_add_tail(&workload->list, \
--- a/drivers/gpu/drm/i915/gvt/vgpu.c
+++ b/drivers/gpu/drm/i915/gvt/vgpu.c
@@ -226,7 +226,7 @@ void intel_gvt_deactivate_vgpu(struct in
vgpu->active = false;
- if (atomic_read(&vgpu->running_workload_num)) {
+ if (atomic_read(&vgpu->submission.running_workload_num)) {
mutex_unlock(&gvt->lock);
intel_gvt_wait_vgpu_idle(vgpu);
mutex_lock(&gvt->lock);
@@ -293,7 +293,7 @@ struct intel_vgpu *intel_gvt_create_idle
vgpu->gvt = gvt;
for (i = 0; i < I915_NUM_ENGINES; i++)
- INIT_LIST_HEAD(&vgpu->workload_q_head[i]);
+ INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]);
ret = intel_vgpu_init_sched_policy(vgpu);
if (ret)