[edk2-devel] [PATCH v7 0/9] Ovmf: Disable the TPM2 platform hierarchy
Stefan Berger
stefanb at linux.ibm.com
Mon Sep 13 14:51:13 UTC 2021
On 9/10/21 10:46 PM, Yao, Jiewen wrote:
> If you want, I would suggest to take 2 steps (2 separate patch sets).
>
> 1) To add the TCG2 platform auth handling the security pkg (just move the code from min-platform to securitypkg)
> If nothing else is changed, it can be approved easily.
>
> 2) To enable QEMU support to make platform auth + TCG PP work together. (based upon 1)
> Need consider how to do it in a secure way.
I am not clear what it's going to take to get this right. Is there are
platform example that does things similar to Ovmf but does it in the
right order?
Several packages are using BdsEntry() from here:
https://github.com/tianocore/edk2/blob/master/MdeModulePkg/Universal/BdsDxe/BdsEntry.c#L661
That's where the split of PlatformBootManagerBeforeConsole() and
...AfterConsole() comes from. It looks like we would have to do TPM PPI
handling in the BeforeConsole function but cannot do it since there's no
console at this point but end-of-dxe is triggered there and that SMM
locking signal is also sent in that function.
EndOfDxe:
https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBootManagerLib/BdsPlatform.c#L380
Smm Lock:
https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBootManagerLib/BdsPlatform.c#L394
If we now move the console initialization ('Connect consoles') to
'before' PlatformBootManagerBeforeConsole() is that then correct? Or
should the PPI module initialize the console when it needs it?
Stefan
>
> Thank you
> Yao Jiewen
>
>> -----Original Message-----
>> From: Yao, Jiewen
>> Sent: Saturday, September 11, 2021 10:38 AM
>> To: Stefan Berger <stefanb at linux.ibm.com>; devel at edk2.groups.io;
>> stefanb at linux.vnet.ibm.com
>> Cc: mhaeuser at posteo.de; spbrogan at outlook.com;
>> marcandre.lureau at redhat.com; kraxel at redhat.com
>> Subject: RE: [edk2-devel] [PATCH v7 0/9] Ovmf: Disable the TPM2 platform
>> hierarchy
>>
>> Hi Stefan
>> I notice you signal EndOfDxe at PlatformBootManagerBeforeConsole()
>> https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBoo
>> tManagerLib/BdsPlatform.c#L380
>> I would say, if PP is done after EndOfDxe, then the order is NOT right.
>>
>> This topic has been debated for years. Finally, we reach the conclusion with the
>> trusted console concept.
>>
>> The recommended way is to connect *trusted console only* and process PP
>> before EndOfDxe, to ensure no 3rd party code can touch the platform hierarchy.
>> We did that at PlatformBootManagerBeforeConsole(). Here is console means all
>> console, including the trusted console and untrusted console populated by
>> untrusted device. The full console list can still be connected after EndOfDxe.
>> The platform can decide which console is trusted v.s. not-trusted.
>>
>> Thank you
>> Yao Jiewen
>>
>>
>>> -----Original Message-----
>>> From: Stefan Berger <stefanb at linux.ibm.com>
>>> Sent: Saturday, September 11, 2021 12:15 AM
>>> To: Yao, Jiewen <jiewen.yao at intel.com>; devel at edk2.groups.io;
>>> stefanb at linux.vnet.ibm.com
>>> Cc: mhaeuser at posteo.de; spbrogan at outlook.com;
>>> marcandre.lureau at redhat.com; kraxel at redhat.com
>>> Subject: Re: [edk2-devel] [PATCH v7 0/9] Ovmf: Disable the TPM2 platform
>>> hierarchy
>>>
>>>
>>> On 9/10/21 11:32 AM, Yao, Jiewen wrote:
>>>> According to the security policy, PP request must be processed before
>>> EndOfDxe.
>>>> May I know when you trigger PP request?
>>> OVMF has 3 implementations invoking it in
>> PlatformBootManagerAfterConsole():
>>>
>> https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBoo
>>> tManagerLib/BdsPlatform.c#L1517
>>>
>>>
>> https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBoo
>>> tManagerLibBhyve/BdsPlatform.c#L1451
>>>
>>>
>> https://github.com/tianocore/edk2/blob/master/OvmfPkg/Library/PlatformBoo
>>> tManagerLibGrub/BdsPlatform.c#L1316
>>>
>>> Stefan
>>>
>>>
>>>> Thank you
>>>> Yao Jiewen
>>>>
>>>>> -----Original Message-----
>>>>> From: Stefan Berger <stefanb at linux.ibm.com>
>>>>> Sent: Friday, September 10, 2021 10:25 PM
>>>>> To: devel at edk2.groups.io; stefanb at linux.vnet.ibm.com
>>>>> Cc: mhaeuser at posteo.de; spbrogan at outlook.com;
>>>>> marcandre.lureau at redhat.com; kraxel at redhat.com; Yao, Jiewen
>>>>> <jiewen.yao at intel.com>
>>>>> Subject: Re: [edk2-devel] [PATCH v7 0/9] Ovmf: Disable the TPM2 platform
>>>>> hierarchy
>>>>>
>>>>>
>>>>> On 9/9/21 1:35 PM, Stefan Berger wrote:
>>>>>> This series imports code from the edk2-platforms project related to
>>>>>> disabling the TPM2 platform hierarchy in Ovmf. It addresses the Ovmf
>>>>>> aspects of the following bugs:
>>>>>>
>>>>>> https://bugzilla.tianocore.org/show_bug.cgi?id=3510
>>>>>> https://bugzilla.tianocore.org/show_bug.cgi?id=3499
>>>>>>
>>>>>> I have patched the .dsc files and successfully test-built with most of
>>>>>> them. Some I could not build because they failed for other reasons
>>>>>> unrelated to this series.
>>>>>>
>>>>>> I tested the changes with QEMU on x86 following the build of
>>>>>> OvmfPkgX64.dsc.
>>>>>>
>>>>>> Neither one of the following commands should work anymore on first
>>>>>> try when run on Linux:
>>>>>>
>>>>>> With IBM tss2 tools:
>>>>>> tsshierarchychangeauth -hi p -pwdn newpass
>>>>>>
>>>>>> With Intel tss2 tools:
>>>>>> tpm2_changeauth -c platform newpass
>>>>> While disabling the platform hierarchy works, the unfortunate problem is
>>>>> now that the signal to disable the TPM 2 platform hierarchy is received
>>>>> before handling the physical presence interface (PPI) opcodes, which is
>>>>> bad because some of the opcodes will not go through. The question now is
>>>>> what is wrong? Are the PPI opcodes handled too late or the signal is
>>>>> sent to early or is it the wrong signal?
>>>>>
>>>>> Event = EfiCreateProtocolNotifyEvent (
>>>>> &gEfiDxeSmmReadyToLockProtocolGuid,
>>>>> TPL_CALLBACK,
>>>>> SmmReadyToLockEventCallBack,
>>>>> NULL,
>>>>> &Registration
>>>>> );
>>>>>
>>>>> Stefan
>>>>>
>>>>>> Regards,
>>>>>> Stefan
>>>>>>
>>>>>> v7:
>>>>>> - Ditched ARM support in this series
>>>>>> - Using Tcg2PlatformDxe and Tcg2PlaformPei from edk2-platforms now
>>>>>> and revised most of the patches
>>>>>>
>>>>>> v6:
>>>>>> - Removed unnecessary entries in .dsc files
>>>>>> - Added support for S3 resume failure case
>>>>>> - Assigned unique FILE_GUID to NULL implementation
>>>>>>
>>>>>> v5:
>>>>>> - Modified patch 1 copies the code from edk2-platforms
>>>>>> - Modified patch 2 fixes bugs in the code
>>>>>> - Modified patch 4 introduces required PCD
>>>>>>
>>>>>> v4:
>>>>>> - Fixed and simplified code imported from edk2-platforms
>>>>>>
>>>>>> v3:
>>>>>> - Referencing Null implementation on Bhyve and Xen platforms
>>>>>> - Add support in Arm
>>>>>>
>>>>>>
>>>>>> Stefan Berger (9):
>>>>>> SecurityPkg/TPM: Import PeiDxeTpmPlatformHierarchyLib.c from
>>>>>> edk2-platforms
>>>>>> SecurityPkg/TPM: Fix bugs in imported PeiDxeTpmPlatformHierarchyLib
>>>>>> SecrutiyPkg/Tcg: Import Tcg2PlatformDxe from edk2-platforms
>>>>>> SecurityPkg/Tcg: Make Tcg2PlatformDxe buildable
>>>>>> SecurityPkg: Introduce new PCD PcdRandomizePlatformHierarchy
>>>>>> OvmfPkg: Reference new Tcg2PlatformDxe in the build system for
>>>>>> compilation
>>>>>> SecurityPkg/Tcg: Import Tcg2PlatformPei from edk2-platforms
>>>>>> SecurityPkg/Tcg: Make Tcg2PlatformPei buildable
>>>>>> OvmfPkg: Reference new Tcg2PlatformPei in the build system
>>>>>>
>>>>>> OvmfPkg/AmdSev/AmdSevX64.dsc | 8 +
>>>>>> OvmfPkg/AmdSev/AmdSevX64.fdf | 2 +
>>>>>> OvmfPkg/OvmfPkgIa32.dsc | 8 +
>>>>>> OvmfPkg/OvmfPkgIa32.fdf | 2 +
>>>>>> OvmfPkg/OvmfPkgIa32X64.dsc | 8 +
>>>>>> OvmfPkg/OvmfPkgIa32X64.fdf | 2 +
>>>>>> OvmfPkg/OvmfPkgX64.dsc | 8 +
>>>>>> OvmfPkg/OvmfPkgX64.fdf | 2 +
>>>>>> .../Include/Library/TpmPlatformHierarchyLib.h | 27 ++
>>>>>> .../PeiDxeTpmPlatformHierarchyLib.c | 255 ++++++++++++++++++
>>>>>> .../PeiDxeTpmPlatformHierarchyLib.inf | 44 +++
>>>>>> SecurityPkg/SecurityPkg.dec | 6 +
>>>>>> .../Tcg/Tcg2PlatformDxe/Tcg2PlatformDxe.c | 85 ++++++
>>>>>> .../Tcg/Tcg2PlatformDxe/Tcg2PlatformDxe.inf | 43 +++
>>>>>> .../Tcg/Tcg2PlatformPei/Tcg2PlatformPei.c | 107 ++++++++
>>>>>> .../Tcg/Tcg2PlatformPei/Tcg2PlatformPei.inf | 51 ++++
>>>>>> 16 files changed, 658 insertions(+)
>>>>>> create mode 100644
>>> SecurityPkg/Include/Library/TpmPlatformHierarchyLib.h
>>>>>> create mode 100644
>> SecurityPkg/Library/PeiDxeTpmPlatformHierarchyLib/PeiDxeTpmPlatformHierar
>>>>> chyLib.c
>>>>>> create mode 100644
>> SecurityPkg/Library/PeiDxeTpmPlatformHierarchyLib/PeiDxeTpmPlatformHierar
>>>>> chyLib.inf
>>>>>> create mode 100644
>>> SecurityPkg/Tcg/Tcg2PlatformDxe/Tcg2PlatformDxe.c
>>>>>> create mode 100644
>>> SecurityPkg/Tcg/Tcg2PlatformDxe/Tcg2PlatformDxe.inf
>>>>>> create mode 100644
>> SecurityPkg/Tcg/Tcg2PlatformPei/Tcg2PlatformPei.c
>>>>>> create mode 100644
>>> SecurityPkg/Tcg/Tcg2PlatformPei/Tcg2PlatformPei.inf
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#80564): https://edk2.groups.io/g/devel/message/80564
Mute This Topic: https://groups.io/mt/85498425/1813853
Group Owner: devel+owner at edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [edk2-devel-archive at redhat.com]
-=-=-=-=-=-=-=-=-=-=-=-
More information about the edk2-devel-archive
mailing list