* [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's @ 2019-09-03 14:57 Damian Nikodem 2019-09-03 16:56 ` Ni, Ray 2019-09-03 17:31 ` Laszlo Ersek 0 siblings, 2 replies; 5+ messages in thread From: Damian Nikodem @ 2019-09-03 14:57 UTC (permalink / raw) To: devel Cc: Damian Nikodem, Eric Dong, Ray Ni, Benjamin You, Laszlo Ersek, Krzysztof Rusocki Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's acquisition attempt of the Busy spinlock (non-blocking mode). UserProc is the user's procedure to execute on an AP. UserProcCompletion is the user procedure's completion spinlock. All other variables are from EDK2. BSP AP ===================================================================================== APHandler () WaitForSemaphore (Run) << initial state >> AcquireSpinLock (UserProcCompletion) SmmStartupThisAp (Procedure) AcquireSpinLockOrFail (Busy) ReleaseSemaphore (Run) UserProc () DoStuff() DoSomeOtherStuff () AcquireSpinLockOrFail (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) ^^ waiting in a loop for user procedure's completion == these fail ReleaseSpinLock (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) ^^ this succeeds ReleaseSpinLock (UserProcCompletion) << return control to the caller and reenter the flow >>> AcquireSpinLock (UserProcCompletion) SmmStartupThisAp (Procedure) AcquireSpinLockOrFail (Busy) ^^ this wins the race with AP's ReleaseSpinLock and fails; ReleaseSpinLock (Busy) return EFI_INVALID_PARAMETER; To remedy, if AcquireSpinLockOrFail (of the Busy spinlock) fails, perform regular AcquireSpinLock -- this eliminates the race condition. Signed-off-by: Damian Nikodem <damian.nikodem@intel.com> Cc: Eric Dong <eric.dong@intel.com> Cc: Ray Ni <ray.ni@intel.com> Cc: Benjamin You <benjamin.you@intel.com> Cc: Laszlo Ersek <lersek@redhat.com> Cc: Krzysztof Rusocki <krzysztof.rusocki@intel.com> diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c index d8d2b6f444..206e196a76 100644 --- a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c @@ -1239,8 +1239,16 @@ InternalSmmStartupThisAp ( AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); } else { if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) { - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData->CpuData[%d].Busy\n", CpuIndex)); - return EFI_NOT_READY; + DEBUG ((DEBUG_INFO, "BSP[%d] finds AP[%d] busy at proc 0x%llX (param 0x%llX), ", + mSmmMpSyncData->BspIndex, + CpuIndex, + *mSmmMpSyncData->CpuData[CpuIndex].Procedure, + (VOID*)mSmmMpSyncData->CpuData[CpuIndex].Parameter)); + DEBUG ((DEBUG_INFO, "new proc 0x%llX (param 0x%llX). Waiting for the previous AP procedure to complete...\n", + Procedure, + ProcArguments)); + + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); } *Token = (MM_COMPLETION) CreateToken (); -------------------------------------------------------------------- Intel Technology Poland sp. z o.o. ul. Slowackiego 173 | 80-298 Gdansk | Sad Rejonowy Gdansk Polnoc | VII Wydzial Gospodarczy Krajowego Rejestru Sadowego - KRS 101882 | NIP 957-07-52-316 | Kapital zakladowy 200.000 PLN. Ta wiadomosc wraz z zalacznikami jest przeznaczona dla okreslonego adresata i moze zawierac informacje poufne. W razie przypadkowego otrzymania tej wiadomosci, prosimy o powiadomienie nadawcy oraz trwale jej usuniecie; jakiekolwiek przegladanie lub rozpowszechnianie jest zabronione. This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by others is strictly prohibited. ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's 2019-09-03 14:57 [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's Damian Nikodem @ 2019-09-03 16:56 ` Ni, Ray 2019-09-04 5:37 ` Dong, Eric 2019-09-04 12:14 ` Rusocki, Krzysztof 2019-09-03 17:31 ` Laszlo Ersek 1 sibling, 2 replies; 5+ messages in thread From: Ni, Ray @ 2019-09-03 16:56 UTC (permalink / raw) To: Nikodem, Damian, devel@edk2.groups.io Cc: Dong, Eric, You, Benjamin, Laszlo Ersek, Rusocki, Krzysztof 1. can we directly call AcquireSpinLock()? *OrFail() can be removed IMO. 2. It's a patch to change the behavior of SmmStartupThisAP(). So that to reduce the potential bugs in caller's code. Patch title is a bit mis-leading. > -----Original Message----- > From: Nikodem, Damian > Sent: Tuesday, September 3, 2019 7:58 AM > To: devel@edk2.groups.io > Cc: Nikodem, Damian <damian.nikodem@intel.com>; Dong, Eric <eric.dong@intel.com>; Ni, Ray <ray.ni@intel.com>; You, > Benjamin <benjamin.you@intel.com>; Laszlo Ersek <lersek@redhat.com>; Rusocki, Krzysztof <krzysztof.rusocki@intel.com> > Subject: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user- > triggered SmmStartupThisAP's > > Race condition between APHandler's release of Busy spinlock and > user-triggered SmmStartupThisAP's acquisition attempt of the Busy spinlock (non-blocking mode). > > UserProc is the user's procedure to execute on an AP. > UserProcCompletion is the user procedure's completion spinlock. > All other variables are from EDK2. > > BSP AP > ===================================================================================== > APHandler () > WaitForSemaphore (Run) > > << initial state >> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ReleaseSemaphore (Run) > UserProc () > DoStuff() DoSomeOtherStuff () > > AcquireSpinLockOrFail (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) > > ^^ waiting in a loop for user procedure's > completion == these fail > ReleaseSpinLock (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) > > ^^ this succeeds > > ReleaseSpinLock (UserProcCompletion) > > << return control to the caller and > reenter the flow >>> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ^^ this wins the race with AP's > ReleaseSpinLock and fails; > ReleaseSpinLock (Busy) > return EFI_INVALID_PARAMETER; > > To remedy, if AcquireSpinLockOrFail (of the Busy spinlock) fails, perform regular AcquireSpinLock -- this eliminates the race > condition. > > Signed-off-by: Damian Nikodem <damian.nikodem@intel.com> > Cc: Eric Dong <eric.dong@intel.com> > Cc: Ray Ni <ray.ni@intel.com> > Cc: Benjamin You <benjamin.you@intel.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Krzysztof Rusocki <krzysztof.rusocki@intel.com> > > diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > index d8d2b6f444..206e196a76 100644 > --- a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > @@ -1239,8 +1239,16 @@ InternalSmmStartupThisAp ( > AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } else { > if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) { > - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData->CpuData[%d].Busy\n", CpuIndex)); > - return EFI_NOT_READY; > + DEBUG ((DEBUG_INFO, "BSP[%d] finds AP[%d] busy at proc 0x%llX (param 0x%llX), ", > + mSmmMpSyncData->BspIndex, > + CpuIndex, > + *mSmmMpSyncData->CpuData[CpuIndex].Procedure, > + (VOID*)mSmmMpSyncData->CpuData[CpuIndex].Parameter)); > + DEBUG ((DEBUG_INFO, "new proc 0x%llX (param 0x%llX). Waiting for the previous AP procedure to complete...\n", > + Procedure, > + ProcArguments)); > + > + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } > > *Token = (MM_COMPLETION) CreateToken (); ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's 2019-09-03 16:56 ` Ni, Ray @ 2019-09-04 5:37 ` Dong, Eric 2019-09-04 12:14 ` Rusocki, Krzysztof 1 sibling, 0 replies; 5+ messages in thread From: Dong, Eric @ 2019-09-04 5:37 UTC (permalink / raw) To: Ni, Ray, Nikodem, Damian, devel@edk2.groups.io Cc: You, Benjamin, Laszlo Ersek, Rusocki, Krzysztof Agree with Ray, no need to call AcquireSpinLockOrFail anymore. I think final code like change like below: - if (Token == NULL) { - AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); - } else { - if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) { - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData->CpuData[%d].Busy\n", CpuIndex)); - return EFI_NOT_READY; - } + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); + if (Token != NULL) { Thanks, Eric > -----Original Message----- > From: Ni, Ray > Sent: Wednesday, September 4, 2019 12:56 AM > To: Nikodem, Damian <damian.nikodem@intel.com>; devel@edk2.groups.io > Cc: Dong, Eric <eric.dong@intel.com>; You, Benjamin > <benjamin.you@intel.com>; Laszlo Ersek <lersek@redhat.com>; Rusocki, > Krzysztof <krzysztof.rusocki@intel.com> > Subject: RE: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between > APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's > > 1. can we directly call AcquireSpinLock()? *OrFail() can be removed IMO. > 2. It's a patch to change the behavior of SmmStartupThisAP(). So that to > reduce the potential bugs in caller's code. Patch title is a bit mis-leading. > > > -----Original Message----- > > From: Nikodem, Damian > > Sent: Tuesday, September 3, 2019 7:58 AM > > To: devel@edk2.groups.io > > Cc: Nikodem, Damian <damian.nikodem@intel.com>; Dong, Eric > > <eric.dong@intel.com>; Ni, Ray <ray.ni@intel.com>; You, Benjamin > > <benjamin.you@intel.com>; Laszlo Ersek <lersek@redhat.com>; Rusocki, > > Krzysztof <krzysztof.rusocki@intel.com> > > Subject: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between > > APHandler's release of Busy spinlock and user- triggered > > SmmStartupThisAP's > > > > Race condition between APHandler's release of Busy spinlock and > > user-triggered SmmStartupThisAP's acquisition attempt of the Busy spinlock > (non-blocking mode). > > > > UserProc is the user's procedure to execute on an AP. > > UserProcCompletion is the user procedure's completion spinlock. > > All other variables are from EDK2. > > > > BSP AP > > > =============================================================== > ====================== > > > APHandler () > > > WaitForSemaphore (Run) > > > > << initial > state >> > > > > AcquireSpinLock (UserProcCompletion) > > SmmStartupThisAp (Procedure) > > AcquireSpinLockOrFail (Busy) > > ReleaseSemaphore (Run) > > > UserProc () > > DoStuff() DoSomeOtherStuff () > > > > AcquireSpinLockOrFail (UserProcCompletion) AcquireSpinLockOrFail > > (UserProcCompletion) > > > > ^^ waiting in a loop for user procedure's > > completion == these fail > > ReleaseSpinLock (UserProcCompletion) AcquireSpinLockOrFail > (UserProcCompletion) > > > > ^^ this succeeds > > > > ReleaseSpinLock (UserProcCompletion) > > > > << return control to the caller and > > reenter the flow >>> > > > > AcquireSpinLock (UserProcCompletion) > > SmmStartupThisAp (Procedure) > > AcquireSpinLockOrFail (Busy) > > ^^ this wins the race with AP's > > ReleaseSpinLock and fails; > > > ReleaseSpinLock (Busy) > > return EFI_INVALID_PARAMETER; > > > > To remedy, if AcquireSpinLockOrFail (of the Busy spinlock) fails, > > perform regular AcquireSpinLock -- this eliminates the race condition. > > > > Signed-off-by: Damian Nikodem <damian.nikodem@intel.com> > > Cc: Eric Dong <eric.dong@intel.com> > > Cc: Ray Ni <ray.ni@intel.com> > > Cc: Benjamin You <benjamin.you@intel.com> > > Cc: Laszlo Ersek <lersek@redhat.com> > > Cc: Krzysztof Rusocki <krzysztof.rusocki@intel.com> > > > > diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > > b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > > index d8d2b6f444..206e196a76 100644 > > --- a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > > +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > > @@ -1239,8 +1239,16 @@ InternalSmmStartupThisAp ( > > AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > > } else { > > if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) > { > > - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData- > >CpuData[%d].Busy\n", CpuIndex)); > > - return EFI_NOT_READY; > > + DEBUG ((DEBUG_INFO, "BSP[%d] finds AP[%d] busy at proc 0x%llX > (param 0x%llX), ", > > + mSmmMpSyncData->BspIndex, > > + CpuIndex, > > + *mSmmMpSyncData->CpuData[CpuIndex].Procedure, > > + (VOID*)mSmmMpSyncData->CpuData[CpuIndex].Parameter)); > > + DEBUG ((DEBUG_INFO, "new proc 0x%llX (param 0x%llX). Waiting for > the previous AP procedure to complete...\n", > > + Procedure, > > + ProcArguments)); > > + > > + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > > } > > > > *Token = (MM_COMPLETION) CreateToken (); ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's 2019-09-03 16:56 ` Ni, Ray 2019-09-04 5:37 ` Dong, Eric @ 2019-09-04 12:14 ` Rusocki, Krzysztof 1 sibling, 0 replies; 5+ messages in thread From: Rusocki, Krzysztof @ 2019-09-04 12:14 UTC (permalink / raw) To: Ni, Ray, Nikodem, Damian, devel@edk2.groups.io Cc: Dong, Eric, You, Benjamin, Laszlo Ersek From: Ni, Ray Sent: Tuesday, September 3, 2019 6:56 PM > 1. can we directly call AcquireSpinLock()? *OrFail() can be removed IMO. > 2. It's a patch to change the behavior of SmmStartupThisAP(). So that to reduce the potential bugs in caller's code. Patch title is a bit mis-leading. The way I see it, it's currently a mix of both: a) catching issues w/caller's code (user calling SmmStartupThisAp while AP executing previous procedure) and b) tackling the deficiency of the API because the caller has no 100% confident way of determining if non-blocking AP execution has completed or not [!!] I agree it's not the implementation responsibility to hand-hold the user (at least not beyond returning the error code). That said, I don't see an issue dropping the debug message and just using single call to AcquireSpinLock. However, I do want to emphasize, the primary intent of the change is to deal with point (b) rather than (a) :-) Thanks, Krzysztof > -----Original Message----- > From: Nikodem, Damian > Sent: Tuesday, September 3, 2019 7:58 AM > To: devel@edk2.groups.io > Cc: Nikodem, Damian <damian.nikodem@intel.com>; Dong, Eric > <eric.dong@intel.com>; Ni, Ray <ray.ni@intel.com>; You, Benjamin > <benjamin.you@intel.com>; Laszlo Ersek <lersek@redhat.com>; Rusocki, > Krzysztof <krzysztof.rusocki@intel.com> > Subject: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between > APHandler's release of Busy spinlock and user- triggered > SmmStartupThisAP's > > Race condition between APHandler's release of Busy spinlock and > user-triggered SmmStartupThisAP's acquisition attempt of the Busy spinlock (non-blocking mode). > > UserProc is the user's procedure to execute on an AP. > UserProcCompletion is the user procedure's completion spinlock. > All other variables are from EDK2. > > BSP AP > ===================================================================================== > APHandler () > WaitForSemaphore (Run) > > << initial state >> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ReleaseSemaphore (Run) > UserProc () > DoStuff() DoSomeOtherStuff () > > AcquireSpinLockOrFail (UserProcCompletion) AcquireSpinLockOrFail > (UserProcCompletion) > > ^^ waiting in a loop for user procedure's > completion == these fail > ReleaseSpinLock (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) > > ^^ this succeeds > > ReleaseSpinLock (UserProcCompletion) > > << return control to the caller and > reenter the flow >>> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ^^ this wins the race with AP's > ReleaseSpinLock and fails; > ReleaseSpinLock (Busy) > return EFI_INVALID_PARAMETER; > > To remedy, if AcquireSpinLockOrFail (of the Busy spinlock) fails, > perform regular AcquireSpinLock -- this eliminates the race condition. > > Signed-off-by: Damian Nikodem <damian.nikodem@intel.com> > Cc: Eric Dong <eric.dong@intel.com> > Cc: Ray Ni <ray.ni@intel.com> > Cc: Benjamin You <benjamin.you@intel.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Krzysztof Rusocki <krzysztof.rusocki@intel.com> > > diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > index d8d2b6f444..206e196a76 100644 > --- a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > @@ -1239,8 +1239,16 @@ InternalSmmStartupThisAp ( > AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } else { > if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) { > - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData->CpuData[%d].Busy\n", CpuIndex)); > - return EFI_NOT_READY; > + DEBUG ((DEBUG_INFO, "BSP[%d] finds AP[%d] busy at proc 0x%llX (param 0x%llX), ", > + mSmmMpSyncData->BspIndex, > + CpuIndex, > + *mSmmMpSyncData->CpuData[CpuIndex].Procedure, > + (VOID*)mSmmMpSyncData->CpuData[CpuIndex].Parameter)); > + DEBUG ((DEBUG_INFO, "new proc 0x%llX (param 0x%llX). Waiting for the previous AP procedure to complete...\n", > + Procedure, > + ProcArguments)); > + > + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } > > *Token = (MM_COMPLETION) CreateToken (); -------------------------------------------------------------------- Intel Technology Poland sp. z o.o. ul. Slowackiego 173 | 80-298 Gdansk | Sad Rejonowy Gdansk Polnoc | VII Wydzial Gospodarczy Krajowego Rejestru Sadowego - KRS 101882 | NIP 957-07-52-316 | Kapital zakladowy 200.000 PLN. Ta wiadomosc wraz z zalacznikami jest przeznaczona dla okreslonego adresata i moze zawierac informacje poufne. W razie przypadkowego otrzymania tej wiadomosci, prosimy o powiadomienie nadawcy oraz trwale jej usuniecie; jakiekolwiek przegladanie lub rozpowszechnianie jest zabronione. This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by others is strictly prohibited. ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's 2019-09-03 14:57 [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's Damian Nikodem 2019-09-03 16:56 ` Ni, Ray @ 2019-09-03 17:31 ` Laszlo Ersek 1 sibling, 0 replies; 5+ messages in thread From: Laszlo Ersek @ 2019-09-03 17:31 UTC (permalink / raw) To: Damian Nikodem, devel; +Cc: Eric Dong, Ray Ni, Benjamin You, Krzysztof Rusocki On 09/03/19 16:57, Damian Nikodem wrote: > Race condition between APHandler's release of Busy spinlock and > user-triggered SmmStartupThisAP's acquisition attempt of the Busy spinlock (non-blocking mode). > > UserProc is the user's procedure to execute on an AP. > UserProcCompletion is the user procedure's completion spinlock. > All other variables are from EDK2. > > BSP AP > ===================================================================================== > APHandler () > WaitForSemaphore (Run) > > << initial state >> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ReleaseSemaphore (Run) > UserProc () > DoStuff() DoSomeOtherStuff () > > AcquireSpinLockOrFail (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) > > ^^ waiting in a loop for user procedure's > completion == these fail > ReleaseSpinLock (UserProcCompletion) AcquireSpinLockOrFail (UserProcCompletion) > > ^^ this succeeds > > ReleaseSpinLock (UserProcCompletion) > > << return control to the caller and > reenter the flow >>> > > AcquireSpinLock (UserProcCompletion) > SmmStartupThisAp (Procedure) > AcquireSpinLockOrFail (Busy) > ^^ this wins the race with AP's > ReleaseSpinLock and fails; > ReleaseSpinLock (Busy) > return EFI_INVALID_PARAMETER; Sorry, I can't make any sense of this sequence diagram. It seems to have fallen apart due to formatting or other email issues. For example, if we have "AP" and "BSP" columns, I would expect every function name to show up in either. But APHandler() is to the right of *both* columns. Please clean up the commit message: - subject line should be no longer than 74 chars - continuous text paragraphs should be properly filled, and wrapped at 74 chars - the diagram can extend more widely, but it should be a diagram please. (I'm not as familiar with this code as other UefiCpuPkg reviewers, so I absolutely depend on the commit message to guide me.) Thanks Laszlo > > To remedy, if AcquireSpinLockOrFail (of the Busy spinlock) fails, perform regular AcquireSpinLock -- this eliminates the race condition. > > Signed-off-by: Damian Nikodem <damian.nikodem@intel.com> > Cc: Eric Dong <eric.dong@intel.com> > Cc: Ray Ni <ray.ni@intel.com> > Cc: Benjamin You <benjamin.you@intel.com> > Cc: Laszlo Ersek <lersek@redhat.com> > Cc: Krzysztof Rusocki <krzysztof.rusocki@intel.com> > > diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > index d8d2b6f444..206e196a76 100644 > --- a/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/MpService.c > @@ -1239,8 +1239,16 @@ InternalSmmStartupThisAp ( > AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } else { > if (!AcquireSpinLockOrFail (mSmmMpSyncData->CpuData[CpuIndex].Busy)) { > - DEBUG((DEBUG_ERROR, "Can't acquire mSmmMpSyncData->CpuData[%d].Busy\n", CpuIndex)); > - return EFI_NOT_READY; > + DEBUG ((DEBUG_INFO, "BSP[%d] finds AP[%d] busy at proc 0x%llX (param 0x%llX), ", > + mSmmMpSyncData->BspIndex, > + CpuIndex, > + *mSmmMpSyncData->CpuData[CpuIndex].Procedure, > + (VOID*)mSmmMpSyncData->CpuData[CpuIndex].Parameter)); > + DEBUG ((DEBUG_INFO, "new proc 0x%llX (param 0x%llX). Waiting for the previous AP procedure to complete...\n", > + Procedure, > + ProcArguments)); > + > + AcquireSpinLock (mSmmMpSyncData->CpuData[CpuIndex].Busy); > } > > *Token = (MM_COMPLETION) CreateToken (); > -------------------------------------------------------------------- > > Intel Technology Poland sp. z o.o. > ul. Slowackiego 173 | 80-298 Gdansk | Sad Rejonowy Gdansk Polnoc | VII Wydzial Gospodarczy Krajowego Rejestru Sadowego - KRS 101882 | NIP 957-07-52-316 | Kapital zakladowy 200.000 PLN. > > Ta wiadomosc wraz z zalacznikami jest przeznaczona dla okreslonego adresata i moze zawierac informacje poufne. W razie przypadkowego otrzymania tej wiadomosci, prosimy o powiadomienie nadawcy oraz trwale jej usuniecie; jakiekolwiek > przegladanie lub rozpowszechnianie jest zabronione. > This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). If you are not the intended recipient, please contact the sender and delete all copies; any review or distribution by > others is strictly prohibited. > ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2019-09-04 12:14 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2019-09-03 14:57 [PATCH] UefiCpuPkg/PiSmmCpuDxeSmm: Race condition between APHandler's release of Busy spinlock and user-triggered SmmStartupThisAP's Damian Nikodem 2019-09-03 16:56 ` Ni, Ray 2019-09-04 5:37 ` Dong, Eric 2019-09-04 12:14 ` Rusocki, Krzysztof 2019-09-03 17:31 ` Laszlo Ersek
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox