diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gvt/scheduler.c')
| -rw-r--r-- | drivers/gpu/drm/i915/gvt/scheduler.c | 142 |
1 files changed, 102 insertions, 40 deletions
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 2144fb46d0e1..6c79d16b381e 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -84,8 +84,8 @@ static void sr_oa_regs(struct intel_vgpu_workload *workload, u32 *reg_state, bool save) { struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv; - u32 ctx_oactxctrl = dev_priv->perf.oa.ctx_oactxctrl_offset; - u32 ctx_flexeu0 = dev_priv->perf.oa.ctx_flexeu0_offset; + u32 ctx_oactxctrl = dev_priv->perf.ctx_oactxctrl_offset; + u32 ctx_flexeu0 = dev_priv->perf.ctx_flexeu0_offset; int i = 0; u32 flex_mmio[] = { i915_mmio_reg_offset(EU_PERF_CNTL0), @@ -291,9 +291,6 @@ shadow_context_descriptor_update(struct intel_context *ce, * Update bits 0-11 of the context descriptor which includes flags * like GEN8_CTX_* cached in desc_template */ - desc &= U64_MAX << 12; - desc |= ce->gem_context->desc_template & ((1ULL << 12) - 1); - desc &= ~(0x3 << GEN8_CTX_ADDRESSING_MODE_SHIFT); desc |= workload->ctx_desc.addressing_mode << GEN8_CTX_ADDRESSING_MODE_SHIFT; @@ -364,16 +361,13 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) wa_ctx->indirect_ctx.shadow_va = NULL; } -static int set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, - struct i915_gem_context *ctx) +static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, + struct i915_gem_context *ctx) { struct intel_vgpu_mm *mm = workload->shadow_mm; struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm); int i = 0; - if (mm->type != INTEL_GVT_MM_PPGTT || !mm->ppgtt_mm.shadowed) - return -EINVAL; - if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { px_dma(ppgtt->pd) = mm->ppgtt_mm.shadow_pdps[0]; } else { @@ -384,8 +378,6 @@ static int set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload, px_dma(pd) = mm->ppgtt_mm.shadow_pdps[i]; } } - - return 0; } static int @@ -576,6 +568,16 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) return 0; } +static void update_vreg_in_ctx(struct intel_vgpu_workload *workload) +{ + struct intel_vgpu *vgpu = workload->vgpu; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + u32 ring_base; + + ring_base = dev_priv->engine[workload->ring_id]->mmio_base; + vgpu_vreg_t(vgpu, RING_START(ring_base)) = workload->rb_start; +} + static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; @@ -614,6 +616,8 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) static int prepare_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; + int ring = workload->ring_id; int ret = 0; ret = intel_vgpu_pin_mm(workload->shadow_mm); @@ -622,8 +626,16 @@ static int prepare_workload(struct intel_vgpu_workload *workload) return ret; } + if (workload->shadow_mm->type != INTEL_GVT_MM_PPGTT || + !workload->shadow_mm->ppgtt_mm.shadowed) { + gvt_vgpu_err("workload shadow ppgtt isn't ready\n"); + return -EINVAL; + } + update_shadow_pdps(workload); + set_context_ppgtt_from_shadow(workload, s->shadow[ring]->gem_context); + ret = intel_vgpu_sync_oos_pages(workload->vgpu); if (ret) { gvt_vgpu_err("fail to vgpu sync oos pages\n"); @@ -674,7 +686,6 @@ static int dispatch_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; - struct intel_vgpu_submission *s = &vgpu->submission; struct i915_request *rq; int ring_id = workload->ring_id; int ret; @@ -685,13 +696,6 @@ static int dispatch_workload(struct intel_vgpu_workload *workload) mutex_lock(&vgpu->vgpu_lock); mutex_lock(&dev_priv->drm.struct_mutex); - ret = set_context_ppgtt_from_shadow(workload, - s->shadow[ring_id]->gem_context); - if (ret < 0) { - gvt_vgpu_err("workload shadow ppgtt isn't ready\n"); - goto err_req; - } - ret = intel_gvt_workload_req_alloc(workload); if (ret) goto err_req; @@ -990,6 +994,7 @@ static int workload_thread(void *priv) int ret; bool need_force_wake = (INTEL_GEN(gvt->dev_priv) >= 9); DEFINE_WAIT_FUNC(wait, woken_wake_function); + struct intel_runtime_pm *rpm = &gvt->dev_priv->runtime_pm; kfree(p); @@ -1013,12 +1018,21 @@ static int workload_thread(void *priv) workload->ring_id, workload, workload->vgpu->id); + intel_runtime_pm_get(rpm); + gvt_dbg_sched("ring id %d will dispatch workload %p\n", workload->ring_id, workload); if (need_force_wake) intel_uncore_forcewake_get(&gvt->dev_priv->uncore, FORCEWAKE_ALL); + /* + * Update the vReg of the vGPU which submitted this + * workload. The vGPU may use these registers for checking + * the context state. The value comes from GPU commands + * in this workload. + */ + update_vreg_in_ctx(workload); ret = dispatch_workload(workload); @@ -1042,6 +1056,7 @@ complete: intel_uncore_forcewake_put(&gvt->dev_priv->uncore, FORCEWAKE_ALL); + intel_runtime_pm_put_unchecked(rpm); if (ret && (vgpu_is_vm_unhealthy(ret))) enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); } @@ -1156,7 +1171,7 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) intel_vgpu_select_submission_ops(vgpu, ALL_ENGINES, 0); - i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(s->shadow[0]->gem_context->vm)); + i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(s->shadow[0]->vm)); for_each_engine(engine, vgpu->gvt->dev_priv, id) intel_context_unpin(s->shadow[id]); @@ -1214,30 +1229,43 @@ i915_context_ppgtt_root_save(struct intel_vgpu_submission *s, */ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) { + struct drm_i915_private *i915 = vgpu->gvt->dev_priv; struct intel_vgpu_submission *s = &vgpu->submission; struct intel_engine_cs *engine; struct i915_gem_context *ctx; enum intel_engine_id i; int ret; - ctx = i915_gem_context_create_gvt(&vgpu->gvt->dev_priv->drm); - if (IS_ERR(ctx)) - return PTR_ERR(ctx); + mutex_lock(&i915->drm.struct_mutex); + + ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX); + if (IS_ERR(ctx)) { + ret = PTR_ERR(ctx); + goto out_unlock; + } + + i915_gem_context_set_force_single_submission(ctx); i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm)); - for_each_engine(engine, vgpu->gvt->dev_priv, i) { + for_each_engine(engine, i915, i) { struct intel_context *ce; INIT_LIST_HEAD(&s->workload_q_head[i]); s->shadow[i] = ERR_PTR(-EINVAL); - ce = i915_gem_context_get_engine(ctx, i); + ce = intel_context_create(ctx, engine); if (IS_ERR(ce)) { ret = PTR_ERR(ce); goto out_shadow_ctx; } + if (!USES_GUC_SUBMISSION(i915)) { /* Max ring buffer size */ + const unsigned int ring_size = 512 * SZ_4K; + + ce->ring = __intel_context_ring_size(ring_size); + } + ret = intel_context_pin(ce); intel_context_put(ce); if (ret) @@ -1264,17 +1292,21 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES); i915_gem_context_put(ctx); + mutex_unlock(&i915->drm.struct_mutex); return 0; out_shadow_ctx: i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm)); - for_each_engine(engine, vgpu->gvt->dev_priv, i) { + for_each_engine(engine, i915, i) { if (IS_ERR(s->shadow[i])) break; intel_context_unpin(s->shadow[i]); + intel_context_put(s->shadow[i]); } i915_gem_context_put(ctx); +out_unlock: + mutex_unlock(&i915->drm.struct_mutex); return ret; } @@ -1423,9 +1455,6 @@ static int prepare_mm(struct intel_vgpu_workload *workload) #define same_context(a, b) (((a)->context_id == (b)->context_id) && \ ((a)->lrca == (b)->lrca)) -#define get_last_workload(q) \ - (list_empty(q) ? NULL : container_of(q->prev, \ - struct intel_vgpu_workload, list)) /** * intel_vgpu_create_workload - create a vGPU workload * @vgpu: a vGPU @@ -1445,7 +1474,7 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, { 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 *last_workload = NULL; struct intel_vgpu_workload *workload = NULL; struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; u64 ring_context_gpa; @@ -1471,15 +1500,20 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, head &= RB_HEAD_OFF_MASK; tail &= RB_TAIL_OFF_MASK; - if (last_workload && same_context(&last_workload->ctx_desc, desc)) { - gvt_dbg_el("ring id %d cur workload == last\n", ring_id); - gvt_dbg_el("ctx head %x real head %lx\n", head, - last_workload->rb_tail); - /* - * cannot use guest context head pointer here, - * as it might not be updated at this time - */ - head = last_workload->rb_tail; + list_for_each_entry_reverse(last_workload, q, list) { + + if (same_context(&last_workload->ctx_desc, desc)) { + gvt_dbg_el("ring id %d cur workload == last\n", + ring_id); + gvt_dbg_el("ctx head %x real head %lx\n", head, + last_workload->rb_tail); + /* + * cannot use guest context head pointer here, + * as it might not be updated at this time + */ + head = last_workload->rb_tail; + break; + } } gvt_dbg_el("ring id %d begin a new workload\n", ring_id); @@ -1492,6 +1526,12 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4); + if (!intel_gvt_ggtt_validate_range(vgpu, start, + _RING_CTL_BUF_SIZE(ctl))) { + gvt_vgpu_err("context contain invalid rb at: 0x%x\n", start); + return ERR_PTR(-EINVAL); + } + workload = alloc_workload(vgpu); if (IS_ERR(workload)) return workload; @@ -1516,9 +1556,31 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, workload->wa_ctx.indirect_ctx.size = (indirect_ctx & INDIRECT_CTX_SIZE_MASK) * CACHELINE_BYTES; + + if (workload->wa_ctx.indirect_ctx.size != 0) { + if (!intel_gvt_ggtt_validate_range(vgpu, + workload->wa_ctx.indirect_ctx.guest_gma, + workload->wa_ctx.indirect_ctx.size)) { + gvt_vgpu_err("invalid wa_ctx at: 0x%lx\n", + workload->wa_ctx.indirect_ctx.guest_gma); + kmem_cache_free(s->workloads, workload); + return ERR_PTR(-EINVAL); + } + } + workload->wa_ctx.per_ctx.guest_gma = per_ctx & PER_CTX_ADDR_MASK; workload->wa_ctx.per_ctx.valid = per_ctx & 1; + if (workload->wa_ctx.per_ctx.valid) { + if (!intel_gvt_ggtt_validate_range(vgpu, + workload->wa_ctx.per_ctx.guest_gma, + CACHELINE_BYTES)) { + gvt_vgpu_err("invalid per_ctx at: 0x%lx\n", + workload->wa_ctx.per_ctx.guest_gma); + kmem_cache_free(s->workloads, workload); + return ERR_PTR(-EINVAL); + } + } } gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n", |
