Re: [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib: Avoid AP calls PeiService.

["Dong, Eric" <eric.dong@intel.com>]

Hi all,

We got some feedback about this BZ. Someone think this timeout is valuable for the debug purpose, and oppose to remove it. 
https://bugzilla.tianocore.org/show_bug.cgi?id=1419

So I'm back to here and want to still use this change. I not use "update PcdSpinLockTimeout to 0 in platform" solution because I think core driver depends on platform policy is not a good design. 

Do you guys have any other concern?

Thanks,
Eric

> -----Original Message-----
> From: Brian J. Johnson [mailto:brian.johnson@hpe.com]
> Sent: Friday, December 21, 2018 12:36 AM
> To: Yao, Jiewen <jiewen.yao@intel.com>; Ni, Ruiyu <ruiyu.ni@intel.com>;
> Dong, Eric <eric.dong@intel.com>; edk2-devel@lists.01.org
> Cc: Laszlo Ersek <lersek@redhat.com>
> Subject: Re: [edk2] [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib: Avoid
> AP calls PeiService.
> 
> Agreed.  We've seen issues on real platforms with timed-out spinlocks in DXE
> causing calls to GetPerformanceCounter and DebugAssert.  (DXE has the
> same code, with the same issues.)
> 
> Note that it's possible to set PcdSpinLockTimeout=0 to work around the issue
> on a particular platform, or in a particular module.  But if you have to do that
> for every module which uses APs, and hence could contend for a spinlock, it
> kind of defeats the point....  We're better off removing the timeout code.
> 
> Thanks,
> Brian
> 
> On 12/19/18 8:08 PM, Yao, Jiewen wrote:
> > Yes, I agree, if we don't have any real case.
> >
> >
> >> -----Original Message-----
> >> From: Ni, Ruiyu
> >> Sent: Thursday, December 20, 2018 10:07 AM
> >> To: Dong, Eric <eric.dong@intel.com>; Yao, Jiewen
> >> <jiewen.yao@intel.com>; edk2-devel@lists.01.org
> >> Cc: Laszlo Ersek <lersek@redhat.com>
> >> Subject: RE: [edk2] [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib:
> >> Avoid AP calls PeiService.
> >>
> >> Can you just change the AcquireSpinLock() behavior to remove the
> >> Timeout PCD consumption?
> >>
> >> I haven't seen a real case that the timed acquisition of spin lock is needed.
> >>
> >>
> >> Thanks/Ray
> >>
> >>> -----Original Message-----
> >>> From: Dong, Eric <eric.dong@intel.com>
> >>> Sent: Thursday, December 20, 2018 9:23 AM
> >>> To: Yao, Jiewen <jiewen.yao@intel.com>; edk2-devel@lists.01.org
> >>> Cc: Ni, Ruiyu <ruiyu.ni@intel.com>; Laszlo Ersek <lersek@redhat.com>
> >>> Subject: RE: [edk2] [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib:
> >>> Avoid AP calls PeiService.
> >>>
> >>>
> >>> Agreed, Maybe it's time to add a new API like
> >>> AcquireSpinLockWithoutTimeOut?
> >>>
> >>> Thanks,
> >>> Eric
> >>>> -----Original Message-----
> >>>> From: Yao, Jiewen
> >>>> Sent: Thursday, December 20, 2018 9:19 AM
> >>>> To: Dong, Eric <eric.dong@intel.com>; edk2-devel@lists.01.org
> >>>> Cc: Ni, Ruiyu <ruiyu.ni@intel.com>; Laszlo Ersek
> >>>> <lersek@redhat.com>
> >>>> Subject: RE: [edk2] [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib:
> >>>> Avoid AP calls PeiService.
> >>>>
> >>>> Hi
> >>>> If we think below code is generic, can we have an API for that?
> >>>>
> >>>> +      //
> >>>> +      // Wait for the AP to release the MSR spin lock.
> >>>> +      //
> >>>> +      while (!AcquireSpinLockOrFail (&CpuFlags->ConsoleLogLock)) {
> >>>> +        CpuPause ();
> >>>> +      }
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>> -----Original Message-----
> >>>>> From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On
> >>>>> Behalf Of Eric Dong
> >>>>> Sent: Thursday, December 20, 2018 9:16 AM
> >>>>> To: edk2-devel@lists.01.org
> >>>>> Cc: Ni, Ruiyu <ruiyu.ni@intel.com>; Laszlo Ersek
> >>>>> <lersek@redhat.com>
> >>>>> Subject: [edk2] [Patch 1/3] UefiCpuPkg/RegisterCpuFeaturesLib:
> >>>>> Avoid AP calls PeiService.
> >>>>>
> >>>>> In AcquireSpinLock function, it calls GetPerformanceCounter which
> >>>>> final calls PeiService service. This patch avoid to call
> >>>>> AcquireSpinLock function.
> >>>>>
> >>>>> BZ: https://bugzilla.tianocore.org/show_bug.cgi?id=1411
> >>>>>
> >>>>> Cc: Ruiyu Ni <ruiyu.ni@intel.com>
> >>>>> Cc: Laszlo Ersek <lersek@redhat.com>
> >>>>> Contributed-under: TianoCore Contribution Agreement 1.1
> >>>>> Signed-off-by: Eric Dong <eric.dong@intel.com>
> >>>>> ---
> >>>>>
> >>>>> UefiCpuPkg/Library/RegisterCpuFeaturesLib/CpuFeaturesInitialize.c
> >>>>> |
> >>>>> 7
> >>>>> ++++++-
> >>>>>   1 file changed, 6 insertions(+), 1 deletion(-)
> >>>>>
> >>>>> diff --git
> >>>>> a/UefiCpuPkg/Library/RegisterCpuFeaturesLib/CpuFeaturesInitialize.
> >>>>> c
> >>>>> b/UefiCpuPkg/Library/RegisterCpuFeaturesLib/CpuFeaturesInitialize.
> >>>>> c index 624ddee055..a64326239f 100644
> >>>>> ---
> >>>>> a/UefiCpuPkg/Library/RegisterCpuFeaturesLib/CpuFeaturesInitialize.
> >>>>> c
> >>>>> +++
> >> b/UefiCpuPkg/Library/RegisterCpuFeaturesLib/CpuFeaturesInitialize.
> >>>>> +++ c
> >>>>> @@ -832,7 +832,12 @@ ProgramProcessorRegister (
> >>>>>       RegisterTableEntry = &RegisterTableEntryHead[Index];
> >>>>>
> >>>>>       DEBUG_CODE_BEGIN ();
> >>>>> -      AcquireSpinLock (&CpuFlags->ConsoleLogLock);
> >>>>> +      //
> >>>>> +      // Wait for the AP to release the MSR spin lock.
> >>>>> +      //
> >>>>> +      while (!AcquireSpinLockOrFail (&CpuFlags->ConsoleLogLock)) {
> >>>>> +        CpuPause ();
> >>>>> +      }
> >>>>>         ThreadIndex = ApLocation->Package *
> >> CpuStatus->MaxCoreCount *
> >>>>> CpuStatus->MaxThreadCount +
> >>>>>                 ApLocation->Core * CpuStatus->MaxThreadCount +
> >>>>>                 ApLocation->Thread;
> >>>>> --
> >>>>> 2.15.0.windows.1
> >>>>>
> >>>>> _______________________________________________
> >>>>> edk2-devel mailing list
> >>>>> edk2-devel@lists.01.org
> >>>>> https://lists.01.org/mailman/listinfo/edk2-devel
> > _______________________________________________
> > edk2-devel mailing list
> > edk2-devel@lists.01.org
> > https://lists.01.org/mailman/listinfo/edk2-devel
> >
> 
> 
> --
> Brian J. Johnson
> Enterprise X86 Lab
> 
> Hewlett Packard Enterprise
> 
> brian.johnson@hpe.com