clear-pkgs-linux-iot-lts2018/0596-drm-i915-gvt-inject-er...

165 lines
5.9 KiB
Diff

From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Min He <min.he@intel.com>
Date: Fri, 14 Sep 2018 16:10:18 +0800
Subject: [PATCH] drm/i915/gvt: inject error interrupt to DomU when GPU hang
When GVT finds a request from DomU causes GPU hang, it will trigger
an error interrupt to guest, so that DomU can trigger a virtual GPU
reset.
Change-Id: I49d9339e99ebfdbe9b158ba311655ab356562bae
Signed-off-by: Min He <min.he@intel.com>
Signed-off-by: Satyeshwar Singh <satyeshwar.singh@intel.com>
Reviewed-on:
Reviewed-by: Jiang, Fei <fei.jiang@intel.com>
Reviewed-by: Dong, Eddie <eddie.dong@intel.com>
Tested-by: Dong, Eddie <eddie.dong@intel.com>
---
drivers/gpu/drm/i915/gvt/interrupt.c | 7 +++++++
drivers/gpu/drm/i915/gvt/interrupt.h | 2 ++
drivers/gpu/drm/i915/gvt/scheduler.c | 27 +++++++++++++++++++++++++++
drivers/gpu/drm/i915/gvt/scheduler.h | 1 +
4 files changed, 37 insertions(+)
diff --git a/drivers/gpu/drm/i915/gvt/interrupt.c b/drivers/gpu/drm/i915/gvt/interrupt.c
index d749d46..06ce906 100644
--- a/drivers/gpu/drm/i915/gvt/interrupt.c
+++ b/drivers/gpu/drm/i915/gvt/interrupt.c
@@ -69,6 +69,7 @@ static const char * const irq_name[INTEL_GVT_EVENT_MAX] = {
[VCS_PAGE_DIRECTORY_FAULT] = "Video page directory faults",
[VCS_AS_CONTEXT_SWITCH] = "Video AS Context Switch Interrupt",
[VCS2_MI_USER_INTERRUPT] = "VCS2 Video CS MI USER INTERRUPT",
+ [VCS2_CMD_STREAMER_ERR] = "VCS2 Video CS error interrupt",
[VCS2_MI_FLUSH_DW] = "VCS2 Video MI FLUSH DW notify",
[VCS2_AS_CONTEXT_SWITCH] = "VCS2 Context Switch Interrupt",
@@ -524,21 +525,26 @@ static void gen8_init_irq(
/* GEN8 interrupt GT0 events */
SET_BIT_INFO(irq, 0, RCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 3, RCS_CMD_STREAMER_ERR, INTEL_GVT_IRQ_INFO_GT0);
SET_BIT_INFO(irq, 4, RCS_PIPE_CONTROL, INTEL_GVT_IRQ_INFO_GT0);
SET_BIT_INFO(irq, 8, RCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
SET_BIT_INFO(irq, 16, BCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT0);
+ SET_BIT_INFO(irq, 19, BCS_CMD_STREAMER_ERR, INTEL_GVT_IRQ_INFO_GT0);
SET_BIT_INFO(irq, 20, BCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT0);
SET_BIT_INFO(irq, 24, BCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT0);
/* GEN8 interrupt GT1 events */
SET_BIT_INFO(irq, 0, VCS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 3, VCS_CMD_STREAMER_ERR, INTEL_GVT_IRQ_INFO_GT1);
SET_BIT_INFO(irq, 4, VCS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT1);
SET_BIT_INFO(irq, 8, VCS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT1);
if (HAS_BSD2(gvt->dev_priv)) {
SET_BIT_INFO(irq, 16, VCS2_MI_USER_INTERRUPT,
INTEL_GVT_IRQ_INFO_GT1);
+ SET_BIT_INFO(irq, 19, VCS2_CMD_STREAMER_ERR,
+ INTEL_GVT_IRQ_INFO_GT1);
SET_BIT_INFO(irq, 20, VCS2_MI_FLUSH_DW,
INTEL_GVT_IRQ_INFO_GT1);
SET_BIT_INFO(irq, 24, VCS2_AS_CONTEXT_SWITCH,
@@ -547,6 +553,7 @@ static void gen8_init_irq(
/* GEN8 interrupt GT3 events */
SET_BIT_INFO(irq, 0, VECS_MI_USER_INTERRUPT, INTEL_GVT_IRQ_INFO_GT3);
+ SET_BIT_INFO(irq, 3, VECS_CMD_STREAMER_ERR, INTEL_GVT_IRQ_INFO_GT3);
SET_BIT_INFO(irq, 4, VECS_MI_FLUSH_DW, INTEL_GVT_IRQ_INFO_GT3);
SET_BIT_INFO(irq, 8, VECS_AS_CONTEXT_SWITCH, INTEL_GVT_IRQ_INFO_GT3);
diff --git a/drivers/gpu/drm/i915/gvt/interrupt.h b/drivers/gpu/drm/i915/gvt/interrupt.h
index f7d7ade..6ec761a 100644
--- a/drivers/gpu/drm/i915/gvt/interrupt.h
+++ b/drivers/gpu/drm/i915/gvt/interrupt.h
@@ -53,6 +53,7 @@ enum intel_gvt_event_type {
VCS_AS_CONTEXT_SWITCH,
VCS2_MI_USER_INTERRUPT,
+ VCS2_CMD_STREAMER_ERR,
VCS2_MI_FLUSH_DW,
VCS2_AS_CONTEXT_SWITCH,
@@ -64,6 +65,7 @@ enum intel_gvt_event_type {
BCS_AS_CONTEXT_SWITCH,
VECS_MI_USER_INTERRUPT,
+ VECS_CMD_STREAMER_ERR,
VECS_MI_FLUSH_DW,
VECS_AS_CONTEXT_SWITCH,
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index b0e5428..95b65f7 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -682,6 +682,7 @@ static int dispatch_workload(struct intel_vgpu_workload *workload)
ret = prepare_workload(workload);
+ workload->guilty_count = atomic_read(&workload->req->gem_context->guilty_count);
out:
if (ret)
workload->status = ret;
@@ -898,6 +899,9 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
list_del_init(&workload->list);
+ if (workload->status == -EIO)
+ intel_vgpu_reset_submission(vgpu, 1 << ring_id);
+
if (!workload->status) {
release_shadow_batch_buffer(workload);
if(gvt_shadow_wa_ctx)
@@ -933,6 +937,18 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id)
mutex_unlock(&vgpu->vgpu_lock);
}
+static void inject_error_cs_irq(struct intel_vgpu *vgpu, int ring_id)
+{
+ enum intel_gvt_event_type events[] = {
+ RCS_CMD_STREAMER_ERR,
+ BCS_CMD_STREAMER_ERR,
+ VCS_CMD_STREAMER_ERR,
+ VCS2_CMD_STREAMER_ERR,
+ VECS_CMD_STREAMER_ERR,
+ };
+ intel_vgpu_trigger_virtual_event(vgpu, events[ring_id]);
+}
+
struct workload_thread_param {
struct intel_gvt *gvt;
int ring_id;
@@ -1002,6 +1018,17 @@ static int workload_thread(void *priv)
if (lret >= 0 && workload->status == -EINPROGRESS)
workload->status = 0;
+ /*
+ * increased guilty_count means that this request triggerred
+ * a GPU reset, so we need to notify the guest about the
+ * hang.
+ */
+ if (workload->guilty_count <
+ atomic_read(&workload->req->gem_context->guilty_count)) {
+ workload->status = -EIO;
+ inject_error_cs_irq(workload->vgpu, ring_id);
+ }
+
complete:
gvt_dbg_sched("will complete workload %p, status: %d\n",
workload, workload->status);
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h
index ca5529d..043c2ff 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.h
+++ b/drivers/gpu/drm/i915/gvt/scheduler.h
@@ -84,6 +84,7 @@ struct intel_vgpu_workload {
/* if this workload has been dispatched to i915? */
bool dispatched;
int status;
+ unsigned int guilty_count;
struct intel_vgpu_mm *shadow_mm;
--
https://clearlinux.org