"To track whether a request has started on HW, we can emit a breadcrumb at the beginning of the request and check its timeline's HWSP to see if the breadcrumb has advanced past the start of this request." It means all the request which timeline's has_init_breadcrumb is true, then the emit_init_breadcrumb process must have before emitting the real commands, otherwise, the scheduler might get a wrong state of this request during reset. If the request is exactly the guilty one, the scheduler won't terminate it with the wrong state. To avoid this, do emit_init_breadcrumb for all the requests from gvt.
v2: cc to stable kernel
Fixes: 8547444137ec ("drm/i915: Identify active requests") Cc: stable@vger.kernel.org Signed-off-by: Weinan weinan.z.li@intel.com --- drivers/gpu/drm/i915/gvt/scheduler.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 7c99bbc3e2b8..ccd71152c9bc 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -298,12 +298,31 @@ static int copy_workload_to_ring_buffer(struct intel_vgpu_workload *workload) struct i915_request *req = workload->req; void *shadow_ring_buffer_va; u32 *cs; + int err;
if ((IS_KABYLAKE(req->i915) || IS_BROXTON(req->i915) || IS_COFFEELAKE(req->i915)) && is_inhibit_context(req->hw_context)) intel_vgpu_restore_inhibit_context(vgpu, req);
+ /* + * To track whether a request has started on HW, we can emit a + * breadcrumb at the beginning of the request and check its + * timeline's HWSP to see if the breadcrumb has advanced past the + * start of this request. Actually, the request must have the + * init_breadcrumb if its timeline set has_init_bread_crumb, or the + * scheduler might get a wrong state of it during reset. Since the + * requests from gvt always set the has_init_breadcrumb flag, here + * need to do the emit_init_breadcrumb for all the requests. + */ + if (req->engine->emit_init_breadcrumb) { + err = req->engine->emit_init_breadcrumb(req); + if (err) { + gvt_vgpu_err("fail to emit init breadcrumb\n"); + return err; + } + } + /* allocate shadow ring buffer */ cs = intel_ring_begin(workload->req, workload->rb_len / sizeof(u32)); if (IS_ERR(cs)) {
On 2019.05.10 15:57:20 +0800, Weinan wrote:
"To track whether a request has started on HW, we can emit a breadcrumb at the beginning of the request and check its timeline's HWSP to see if the breadcrumb has advanced past the start of this request." It means all the request which timeline's has_init_breadcrumb is true, then the emit_init_breadcrumb process must have before emitting the real commands, otherwise, the scheduler might get a wrong state of this request during reset. If the request is exactly the guilty one, the scheduler won't terminate it with the wrong state. To avoid this, do emit_init_breadcrumb for all the requests from gvt.
v2: cc to stable kernel
Fixes: 8547444137ec ("drm/i915: Identify active requests") Cc: stable@vger.kernel.org Signed-off-by: Weinan weinan.z.li@intel.com
Thanks for the fix!
Acked-by: Zhenyu Wang zhenyuw@linux.intel.com
drivers/gpu/drm/i915/gvt/scheduler.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 7c99bbc3e2b8..ccd71152c9bc 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -298,12 +298,31 @@ static int copy_workload_to_ring_buffer(struct intel_vgpu_workload *workload) struct i915_request *req = workload->req; void *shadow_ring_buffer_va; u32 *cs;
- int err;
if ((IS_KABYLAKE(req->i915) || IS_BROXTON(req->i915) || IS_COFFEELAKE(req->i915)) && is_inhibit_context(req->hw_context)) intel_vgpu_restore_inhibit_context(vgpu, req);
- /*
* To track whether a request has started on HW, we can emit a* breadcrumb at the beginning of the request and check its* timeline's HWSP to see if the breadcrumb has advanced past the* start of this request. Actually, the request must have the* init_breadcrumb if its timeline set has_init_bread_crumb, or the* scheduler might get a wrong state of it during reset. Since the* requests from gvt always set the has_init_breadcrumb flag, here* need to do the emit_init_breadcrumb for all the requests.*/- if (req->engine->emit_init_breadcrumb) {
err = req->engine->emit_init_breadcrumb(req);if (err) {gvt_vgpu_err("fail to emit init breadcrumb\n");return err;}- }
- /* allocate shadow ring buffer */ cs = intel_ring_begin(workload->req, workload->rb_len / sizeof(u32)); if (IS_ERR(cs)) {
-- 2.17.1
intel-gvt-dev mailing list intel-gvt-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
linux-stable-mirror@lists.linaro.org
-
Weinan -
Zhenyu Wang