From: "Ankur Arora" <ankur.a.arora@oracle.com>
To: Laszlo Ersek <lersek@redhat.com>, devel@edk2.groups.io
Cc: imammedo@redhat.com, boris.ostrovsky@oracle.com,
Jian J Wang <jian.j.wang@intel.com>,
Michael D Kinney <michael.d.kinney@intel.com>,
Liming Gao <gaoliming@byosoft.com.cn>,
Zhiguang Liu <zhiguang.liu@intel.com>,
Eric Dong <eric.dong@intel.com>, Ray Ni <ray.ni@intel.com>,
Rahul Kumar <rahul1.kumar@intel.com>,
Aaron Young <aaron.young@oracle.com>
Subject: Re: [edk2-devel] [PATCH v2 05/10] MdePkg: add MmRegisterShutdownInterface()
Date: Thu, 7 Jan 2021 13:19:29 -0800 [thread overview]
Message-ID: <6893343b-bfb0-2b2b-4954-72a2ad838c48@oracle.com> (raw)
In-Reply-To: <f9d6b0e6-fb51-4d39-4834-887dad50f2c2@redhat.com>
On 2021-01-07 12:48 p.m., Laszlo Ersek wrote:
> On 01/07/21 20:55, Ankur Arora wrote:
>> Add MmRegisterShutdownInterface(), which is used to register a callback,
>> that gets used to do the final ejection as part of CPU hot-unplug.
>>
>> Cc: Jian J Wang <jian.j.wang@intel.com>
>> Cc: Michael D Kinney <michael.d.kinney@intel.com>
>> Cc: Liming Gao <gaoliming@byosoft.com.cn>
>> Cc: Zhiguang Liu <zhiguang.liu@intel.com>
>> Cc: Eric Dong <eric.dong@intel.com>
>> Cc: Ray Ni <ray.ni@intel.com>
>> Cc: Laszlo Ersek <lersek@redhat.com>
>> Cc: Rahul Kumar <rahul1.kumar@intel.com>
>> Cc: Igor Mammedov <imammedo@redhat.com>
>> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>> Cc: Aaron Young <aaron.young@oracle.com>
>> Ref: https://bugzilla.tianocore.org/show_bug.cgi?id=3132
>> Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
>> ---
>>
>> Not sure this is the right way to register a callback to be called
>> from SmmCpuFeaturesRendezvousExit(). Happy to hear suggestions for
>> a better way to accomplish this.
>
> No, it's not.
>
> SmmCpuFeaturesRendezvousExit() is an interface in the
> "UefiCpuPkg/Include/Library/SmmCpuFeaturesLib.h" class. Each platform is
> supposed to create its own instance of this library class. For example,
> OVMF's instance is at:
>
> OvmfPkg/Library/SmmCpuFeaturesLib
>
> You can modify the SmmCpuFeaturesRendezvousExit() function
> implementation in that library instance.
>
> I don't think you can easily modify the function *declaration* though,
> as that would break a whole lot of platforms that exist outside of edk2.
>
> Additionally, I don't think you can modify EFI_MM_SYSTEM_TABLE. That's a
> structure defined in the Platform Init specification. You'd first need
> to standardize (via the Platform Init Working Group of the UEFI Forum)
> the proposed change, and a compatible way to upgrade would have to be
> found. An alternative would be the "code first" procedure, which exists
> so that code can be contributed ahead of changing the published
> standards. But I'm unsure (I don't remember) how that procedure works in
> practice.
Thanks, yeah I'm definitely not looking to do that. I did look through
the spec but not enough to get any kind of familiarity with it.
>
> Either way, I would advise against this approach. We should limit
> ourselves to modifying only the contents of
> SmmCpuFeaturesRendezvousExit() in OvmfPkg/Library/SmmCpuFeaturesLib.
That was my initial approach. The problem that ran into was that which
CPUs to eject depends on state held in OvmfPkg/CpuHotPlugSmm (see patch 8)
which, I could not figure out a way to make that available given that the
libraries are linked statically.
The other way I can think of doing this would be to just reprobe this state
from QEMU. The problem with that is that then there's zero shared state
between the removal and the ejection.
>
> You should basically migrate the contents of CpuUnplugExitWork() (in
> patch#8) into SmmCpuFeaturesRendezvousExit(), without changing the
> prototype of SmmCpuFeaturesRendezvousExit(). This might require you to
> calculate "IsBSP" on the spot, possibly with the help of some global
> data that you set up in advance.
The reason for trying to capture IsBSP was that mSmmMpSyncData->BspIndex
has gone at this point. If, as you say, there are significant problems
with changing the prototype, then I can figure out a way to avoid that,
especially given that we get the BSP by reading an APIC MSR anyway.
>
> *However*, regarding "IsBSP" itself -- on the QEMU platform, the BSP is
> neither unpluggable, nor switchable with any one of the APs. A removal
> for the BSP will never be attempted, so it's fine to add much simpler
> code that recognizes such an attempt (as a sanity check), and hangs hard
> on it. If that would lead to grave complications, then don't bother with
> it; just assume (or enforce elsewhere) that the BSP is never being
> unplugged.
>
> Please see the apic_designate_bsp() call sites in the QEMU source code
> -- there is exactly one of them, in "target/i386/cpu.c":
>
> /* We hard-wire the BSP to the first CPU. */
> apic_designate_bsp(cpu->apic_state, s->cpu_index == 0);
>
>
> By adding the business logic to SmmCpuFeaturesRendezvousExit() instead
> of CpuUnplugExitWork, said logic will actually be included in the
> PiSmmCpuDxeSmm binary -- but that's fine. That's why the
> SmmCpuFeaturesRendezvousExit() API exists, as a platform customization
> hook for PiSmmCpuDxeSmm.
>
>
> A formal comment in closing -- modifying two packages in the same patch,
> such as OvmfPkg and UefiCpuPkg, is almost always forbidden. It is
> permitted exceptionally, when there is absolutely no way to solve an
> issue with a gradual approach (i.e., by modifying the packages in
> question with separate patches).
Thanks for letting me know. I'll avoid that in future versions.
>
> I think I'll skip reviewing this version beyond this email, as the
> required changes appear quite intrusive.
Could you take a look at patch 8? That's really the crux of the series and
your comments on that would be helpful.
Thanks
Ankur
>
> Thanks
> Laszlo
>
>>
>> ---
>> MdeModulePkg/Core/PiSmmCore/PiSmmCore.c | 1 +
>> MdePkg/Include/Pi/PiMmCis.h | 16 ++++++++++++++++
>> MdePkg/Include/Pi/PiSmmCis.h | 2 ++
>> MdePkg/Library/MmServicesTableLib/MmServicesTableLib.c | 11 +++++++++++
>> UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c | 1 +
>> UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h | 1 +
>> 6 files changed, 32 insertions(+)
>>
>> diff --git a/MdeModulePkg/Core/PiSmmCore/PiSmmCore.c b/MdeModulePkg/Core/PiSmmCore/PiSmmCore.c
>> index cfa9922cbdb5..9d883bb06633 100644
>> --- a/MdeModulePkg/Core/PiSmmCore/PiSmmCore.c
>> +++ b/MdeModulePkg/Core/PiSmmCore/PiSmmCore.c
>> @@ -39,6 +39,7 @@ EFI_SMM_SYSTEM_TABLE2 gSmmCoreSmst = {
>> SmmFreePool,
>> SmmAllocatePages,
>> SmmFreePages,
>> + NULL, // SmmShutdownAp
>> NULL, // SmmStartupThisAp
>> 0, // CurrentlyExecutingCpu
>> 0, // NumberOfCpus
>> diff --git a/MdePkg/Include/Pi/PiMmCis.h b/MdePkg/Include/Pi/PiMmCis.h
>> index fdf0591a03d6..237bd8dcba76 100644
>> --- a/MdePkg/Include/Pi/PiMmCis.h
>> +++ b/MdePkg/Include/Pi/PiMmCis.h
>> @@ -77,6 +77,14 @@ EFI_STATUS
>> IN OUT VOID *ProcArguments OPTIONAL
>> );
>>
>> +typedef
>> +EFI_STATUS
>> +(EFIAPI *EFI_MM_SHUTDOWN_AP)(
>> + IN UINTN CpuNumber,
>> + IN BOOLEAN IsBSP
>> + );
>> +
>> +
>> /**
>> Function prototype for protocol install notification.
>>
>> @@ -242,6 +250,13 @@ VOID
>> IN CONST EFI_MM_ENTRY_CONTEXT *MmEntryContext
>> );
>>
>> +EFI_STATUS
>> +EFIAPI
>> +MmRegisterShutdownInterface (
>> + IN EFI_MM_SHUTDOWN_AP Procedure
>> + );
>> +
>> +
>> ///
>> /// Management Mode System Table (MMST)
>> ///
>> @@ -282,6 +297,7 @@ struct _EFI_MM_SYSTEM_TABLE {
>> ///
>> /// MP service
>> ///
>> + EFI_MM_SHUTDOWN_AP MmShutdownAp;
>> EFI_MM_STARTUP_THIS_AP MmStartupThisAp;
>>
>> ///
>> diff --git a/MdePkg/Include/Pi/PiSmmCis.h b/MdePkg/Include/Pi/PiSmmCis.h
>> index 06ef4aecd7b5..296dc01f6703 100644
>> --- a/MdePkg/Include/Pi/PiSmmCis.h
>> +++ b/MdePkg/Include/Pi/PiSmmCis.h
>> @@ -49,6 +49,7 @@ EFI_STATUS
>> IN UINTN TableSize
>> );
>>
>> +typedef EFI_MM_SHUTDOWN_AP EFI_SMM_SHUTDOWN_AP;
>> typedef EFI_MM_STARTUP_THIS_AP EFI_SMM_STARTUP_THIS_AP;
>> typedef EFI_MM_NOTIFY_FN EFI_SMM_NOTIFY_FN;
>> typedef EFI_MM_REGISTER_PROTOCOL_NOTIFY EFI_SMM_REGISTER_PROTOCOL_NOTIFY;
>> @@ -137,6 +138,7 @@ struct _EFI_SMM_SYSTEM_TABLE2 {
>> ///
>> /// MP service
>> ///
>> + EFI_SMM_SHUTDOWN_AP SmmShutdownAp;
>> EFI_SMM_STARTUP_THIS_AP SmmStartupThisAp;
>>
>> ///
>> diff --git a/MdePkg/Library/MmServicesTableLib/MmServicesTableLib.c b/MdePkg/Library/MmServicesTableLib/MmServicesTableLib.c
>> index 27f9d526e396..c7d81a0dc193 100644
>> --- a/MdePkg/Library/MmServicesTableLib/MmServicesTableLib.c
>> +++ b/MdePkg/Library/MmServicesTableLib/MmServicesTableLib.c
>> @@ -55,3 +55,14 @@ MmServicesTableLibConstructor (
>>
>> return EFI_SUCCESS;
>> }
>> +
>> +EFI_STATUS
>> +EFIAPI
>> +MmRegisterShutdownInterface (
>> + IN EFI_MM_SHUTDOWN_AP Procedure
>> + )
>> +{
>> + gMmst->MmShutdownAp = Procedure;
>> +
>> + return EFI_SUCCESS;
>> +}
>> diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c b/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c
>> index db68e1316ec5..f2f67e85e5e9 100644
>> --- a/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c
>> +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.c
>> @@ -22,6 +22,7 @@ SMM_CPU_PRIVATE_DATA mSmmCpuPrivateData = {
>> NULL, // Pointer to CpuSaveStateSize array
>> NULL, // Pointer to CpuSaveState array
>> { {0} }, // SmmReservedSmramRegion
>> + NULL, // SmmShutdownAp
>> {
>> SmmStartupThisAp, // SmmCoreEntryContext.SmmStartupThisAp
>> 0, // SmmCoreEntryContext.CurrentlyExecutingCpu
>> diff --git a/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h b/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h
>> index b8aa9e1769d3..7672834a2f70 100644
>> --- a/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h
>> +++ b/UefiCpuPkg/PiSmmCpuDxeSmm/PiSmmCpuDxeSmm.h
>> @@ -247,6 +247,7 @@ typedef struct {
>> VOID **CpuSaveState;
>>
>> EFI_SMM_RESERVED_SMRAM_REGION SmmReservedSmramRegion[1];
>> + EFI_SMM_SHUTDOWN_AP SmmShutdownAp;
>> EFI_SMM_ENTRY_CONTEXT SmmCoreEntryContext;
>> EFI_SMM_ENTRY_POINT SmmCoreEntry;
>>
>>
>
next prev parent reply other threads:[~2021-01-07 21:19 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-07 19:55 [PATCH v2 00/10] support CPU hot-unplug Ankur Arora
2021-01-07 19:55 ` [PATCH v2 01/10] OvmfPkg/CpuHotplugSmm: move CPU Hotplug into PlugCpus() Ankur Arora
2021-01-07 19:55 ` [PATCH v2 02/10] OvmfPkg/CpuHotplugSmm: handle Hot-unplug events Ankur Arora
2021-01-07 19:55 ` [PATCH v2 03/10] OvmfPkg/CpuHotplugSmm: add Qemu CpuStatus helper Ankur Arora
2021-01-07 19:55 ` [PATCH v2 04/10] OvmfPkg/CpuHotplugSmm: handle CPU hot-unplug Ankur Arora
2021-01-07 19:55 ` [PATCH v2 05/10] MdePkg: add MmRegisterShutdownInterface() Ankur Arora
2021-01-07 20:48 ` [edk2-devel] " Laszlo Ersek
2021-01-07 21:00 ` Laszlo Ersek
2021-01-07 21:19 ` Ankur Arora [this message]
2021-01-07 21:50 ` Laszlo Ersek
2021-01-07 21:45 ` Laszlo Ersek
2021-01-07 23:42 ` Ankur Arora
2021-01-07 19:55 ` [PATCH v2 06/10] UefiCpuPkg/PiSmmCpuDxeSmm: initialize IsBsp Ankur Arora
2021-01-07 19:55 ` [PATCH v2 07/10] UefiCpuPkg/SmmCpuFeaturesLib: add IsBsp as a param to SmmCpuFeaturesRendezvousExit() Ankur Arora
2021-01-07 19:55 ` [PATCH v2 08/10] OvmfCpuPkg/CpuHotplug: add a hot-unplug handler called at SMI exit Ankur Arora
2021-01-07 19:55 ` [PATCH v2 09/10] OvmfPkg/SmmControl2Dxe: negotiate ICH9_LPC_SMI_F_CPU_HOT_UNPLUG Ankur Arora
2021-01-07 19:55 ` [PATCH v2 10/10] MdePkg: use CpuPause() in CpuDeadLoop() Ankur Arora
2021-01-07 22:30 ` [edk2-devel] " Michael D Kinney
2021-01-07 23:43 ` Ankur Arora
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-list from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=6893343b-bfb0-2b2b-4954-72a2ad838c48@oracle.com \
--to=devel@edk2.groups.io \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox