drm/i915/gvt: Do not check shadowed when adding ppgtt to GVT GEM context

Submitted by Colin Xu on July 4, 2019, 12:09 a.m.

Details

Message ID 20190704000929.3359-1-colin.xu@intel.com
State New
Headers show
Series "drm/i915/gvt: Do not check shadowed when adding ppgtt to GVT GEM context" ( rev: 1 ) in Intel GVT devel

Not browsing as part of any series.

Commit Message

Colin Xu July 4, 2019, 12:09 a.m.
Windows guest can't run after force-TDR with host log:
...
gvt: vgpu 1: workload shadow ppgtt isn't ready
gvt: vgpu 1: fail to dispatch workload, skip
...

The error is raised by set_context_ppgtt_from_shadow(), when it checks
and found the shadow_mm isn't marked as shadowed.

In work thread before each submission, a shadow_mm is set to shadowed in:
shadow_ppgtt_mm()
<-intel_vgpu_pin_mm()
<-prepare_workload()
<-dispatch_workload()
<-workload_thread()
However checking whether or not shadow_mm is shadowed is prior to it:
set_context_ppgtt_from_shadow()
<-dispatch_workload()
<-workload_thread()

In normal case, create workload will check the existence of shadow_mm,
if not it will create a new one and marked as shadowed. If already exist
it will reuse the old one. Since shadow_mm is reused, checking of shadowed
in set_context_ppgtt_from_shadow() actually always see the state set in
creation, but not the state set in intel_vgpu_pin_mm().

When force-TDR, all engines are reset, since it's not dmlr level, all
ppgtt_mm are invalidated but not destroyed. Invalidation will mark all
reused shadow_mm as not shadowed but still keeps in ppgtt_mm_list_head.
If workload submission phase those shadow_mm are reused with shadowed
not set, then set_context_ppgtt_from_shadow() will report error.

No need to check shadowed state in set_context_ppgtt_from_shadow().

Fixes: 4f15665ccbba (drm/i915: Add ppgtt to GVT GEM context)

Signed-off-by: Colin Xu <colin.xu@intel.com>
---
 drivers/gpu/drm/i915/gvt/scheduler.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 196b4155a309..a5db57926962 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -371,7 +371,7 @@  static int set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
 	struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
 	int i = 0;
 
-	if (mm->type != INTEL_GVT_MM_PPGTT || !mm->ppgtt_mm.shadowed)
+	if (mm->type != INTEL_GVT_MM_PPGTT)
 		return -EINVAL;
 
 	if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {

Comments



On 2019-07-04 10:55, Zhenyu Wang wrote:
> On 2019.07.04 08:09:29 +0800, Colin Xu wrote:
>> Windows guest can't run after force-TDR with host log:
>> ...
>> gvt: vgpu 1: workload shadow ppgtt isn't ready
>> gvt: vgpu 1: fail to dispatch workload, skip
>> ...
>>
>> The error is raised by set_context_ppgtt_from_shadow(), when it checks
>> and found the shadow_mm isn't marked as shadowed.
>>
>> In work thread before each submission, a shadow_mm is set to shadowed in:
>> shadow_ppgtt_mm()
>> <-intel_vgpu_pin_mm()
>> <-prepare_workload()
>> <-dispatch_workload()
>> <-workload_thread()
>> However checking whether or not shadow_mm is shadowed is prior to it:
>> set_context_ppgtt_from_shadow()
>> <-dispatch_workload()
>> <-workload_thread()
>>
>> In normal case, create workload will check the existence of shadow_mm,
>> if not it will create a new one and marked as shadowed. If already exist
>> it will reuse the old one. Since shadow_mm is reused, checking of shadowed
>> in set_context_ppgtt_from_shadow() actually always see the state set in
>> creation, but not the state set in intel_vgpu_pin_mm().
>>
>> When force-TDR, all engines are reset, since it's not dmlr level, all
>> ppgtt_mm are invalidated but not destroyed. Invalidation will mark all
>> reused shadow_mm as not shadowed but still keeps in ppgtt_mm_list_head.
>> If workload submission phase those shadow_mm are reused with shadowed
>> not set, then set_context_ppgtt_from_shadow() will report error.
>>
>> No need to check shadowed state in set_context_ppgtt_from_shadow().
> Thanks for tracing this! I think this is a side effect from 1e18d5e6731d674fee0bb4b66f5ea61e504452a3,
> which I moved this setting out of scan function.

I think even before 1e18d5e6731d674fee0bb4b66f5ea61e504452a it still has problem
since intel_gvt_scan_and_shadow_workload() is before prepare_workload().

>
> One potential issue with this is shadow_pdps at this time might not have valid value yet,
> how about we move shadow ctx ppgtt setting after we really prepared mm for workload and
> before i915 submission? As this setting now is mostly to please i915, we have already set
> context's ppgtt by ourself.
>
If prepare_workload() failed, should we still set context ppgtt for the shadow?

>> Fixes: 4f15665ccbba (drm/i915: Add ppgtt to GVT GEM context)
>>
>> Signed-off-by: Colin Xu <colin.xu@intel.com>
>> ---
>>   drivers/gpu/drm/i915/gvt/scheduler.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
>> index 196b4155a309..a5db57926962 100644
>> --- a/drivers/gpu/drm/i915/gvt/scheduler.c
>> +++ b/drivers/gpu/drm/i915/gvt/scheduler.c
>> @@ -371,7 +371,7 @@ static int set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
>>   	struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
>>   	int i = 0;
>>   
>> -	if (mm->type != INTEL_GVT_MM_PPGTT || !mm->ppgtt_mm.shadowed)
>> +	if (mm->type != INTEL_GVT_MM_PPGTT)
>>   		return -EINVAL;
>>   
>>   	if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
>> -- 
>> 2.22.0
>>
>> _______________________________________________
>> intel-gvt-dev mailing list
>> intel-gvt-dev@lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev