Hi Sumit,

> -----Original Message-----
> From: Sumit Garg <sumit.g...@linaro.org>
> 
> Hi Chris,
> 
> On Sat, 3 Nov 2018 at 05:25, Chris Co <christopher...@microsoft.com> wrote:
> >
> > Hi Sumit,
> >
> > > -----Original Message-----
> > > From: Sumit Garg <sumit.g...@linaro.org>
> > >
> > > + OP-TEE ML.
> > >
> > > On Fri, 2 Nov 2018 at 06:11, Chris Co <christopher...@microsoft.com>
> wrote:
> > > >
> > > > Hi Sumit,
> > > >
> > > > Our full OpteeClientPkg has:
> > > > - Our OpteeClientAPI implementation. I was monitoring the merge
> > > > progress
> > > on OpteeLib and will look into moving over now that it is available.
> > > > - The fTPM and AuthVar TA binaries. In our current design, the TA
> > > > binaries
> > > are loaded at runtime. We could host the binaries themselves
> > > elsewhere on the filesystem, but we do not want these binaries as
> > > early/pseudo TAs. Is there a plan for OpteeLib to support loading full 
> > > TAs?
> > >
> > > Early TAs [1] are basically full TAs only, running in Secure EL0 mode.
> > > So instead of loading TA from normal world file-system, they are
> > > linked into a special data section in the OP-TEE core blob.
> > >
> > > Also I don't think loading TAs dynamically especially during boot
> > > makes much sense due to following reasons:
> > > 1. Increased boot time.
> > > 2. Fixed TAs like in your case which could be linked as early TAs as well.
> > >
> >
> > We prefer to load TAs dynamically for a more flexible servicing story. My
> understanding is that Early TAs are coupled with the OP-TEE binary itself, so
> to update an Early TA, a new OP-TEE binary would need to be created and
> pushed. We want to avoid rolling a new OP-TEE and only update the TA
> binary in this scenario.
> >
> 
> Are you referring to run-time updates on the device in the field? If this is 
> the
> case then how do you think to update TAs, is it via some custom capsule
> update method?
> 

Yes, run-time TA updates. Currently, our fTPM and Authvar TAs get packaged 
inside our UEFI binary. So an update to a TA means a UEFI update via firmware 
capsule.
The discussion of these TA binaries living on the filesystem were ideas we were 
discussing internally but are not fully baked or committed to.

> I do consider these TAs used during boot as essential secure services provided
> by the secure firmware (OP-TEE in this case). So these TAs should be part of
> firmware itself and updates for them should come through firmware capsule
> updates only.
> 

I agree in principle and I think I see where the misalignment is, mostly coming 
from my end.
The security guarantees (termed TCPS) we want to provide on the current 
hardware we support (NXP i.MX6), mean OP-TEE becomes prohibitively difficult to 
update. This is due to a hardware resource limitation (not enough fuse space). 
If this limitation were not present, we could freely update OP-TEE and package 
these TAs as EarlyTAs.

Info on TCPS (whitepaper at bottom of post) - 
https://www.microsoft.com/en-us/microsoft-365/blog/2018/04/24/trusted-cyber-physical-systems-looks-to-protect-your-critical-infrastructure-from-modern-threats-in-the-world-of-iot/

I'm not sure how you want to handle this from an OpteeLib vs custom platform 
package perspective.

> > > And you mentioned filesystem, are you referring to root filesystem?
> > >
> >
> > We have not implemented this yet, but we were thinking to have the TA
> binaries present in the EFI partition.
> >
> 
> AFAIK, EFI partition is shared among Linux and UEFI. This provides Linux
> access to secure firmware TAs that could be a security concern (denial of
> service could be one of them).
> 

Note - we are booting Windows, though your point here is still valid. The TAs 
living in the filesystem is not what is implemented today. It was an idea we 
were discussing internally.

> > > > - We have two client drivers: a firmware TPM TA driver and an
> > > authenticated variable TA driver. These talk through the
> > > tee-supplicant to their respective TAs.
> > > >
> > >
> > > Here from tee-supplicant apart from loading TAs, what other services
> > > are you expecting? If you are looking for secure storage via RPMB,
> > > that could be an enhancement to OpteeLib adding corresponding RPC
> handling here [2].
> > >
> >
> > For RPC handling, we are looking for the following callback support:
> > - OPTEE_SMC_RPC_FUNC_ALLOC
> > - OPTEE_SMC_RPC_FUNC_FREE
> > - OPTEE_SMC_RPC_FUNC_CMD
> >         - OPTEE_MSG_RPC_CMD_LOAD_TA
> 
> Please see above comments for this.
> 
> >         - OPTEE_MSG_RPC_CMD_RPMB
> >         - OPTEE_MSG_RPC_CMD_GET_TIME
> 
> Can you share the usage of OPTEE_MSG_RPC_CMD_GET_TIME? AFAIK, this is
> used to get REE time from OP-TEE.
> 

I dug further and found that this was being used in our fTPM TA for debug logs. 
It has since been deprecated so we do not need this RPC command.

> >         - OPTEE_MSG_RPC_CMD_SHM_ALLOC
> >         - OPTEE_MSG_RPC_CMD_SHM_FREE
> >         - OPTEE_MSG_RPC_CMD_WAIT_QUEUE
> 
> I don't think we need OPTEE_MSG_RPC_CMD_WAIT_QUEUE implementation
> in UEFI as its a single threaded execution flow on boot core.
> 

Agreed. Our implementation is effectively a no-op. We don't need this either.

> BTW, I am not sure if I could get time to work on RPC handling anytime soon.
> So patches are welcome and I am happy to review them.
> 

I'll see if I can find time to port over our RPC handlers. Will add you to any 
patches for review.

Thanks,
Chris

> Regards,
> Sumit
> 
> >
> > Thanks,
> > Chris
> >
> > > [1]
> > > https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit
> > > hub.c
> > > om%2FOP-
> > >
> TEE%2Foptee_os%2Fblob%2Fmaster%2Fdocumentation%2Foptee_design.md
> > > %23early-trusted-
> > >
> applications&amp;data=02%7C01%7CChristopher.Co%40microsoft.com%7C4a
> > >
> 7d8c01e4804365f4eb08d640837a15%7C72f988bf86f141af91ab2d7cd011db47%
> > >
> 7C1%7C0%7C636767330779998429&amp;sdata=yaDWw5Z6yuux1o89kxzbknVp
> > > b%2B1OHUagbB%2FOGS4dAcU%3D&amp;reserved=0
> > > [2]
> > > https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit
> > > hub.c
> > >
> om%2Ftianocore%2Fedk2%2Fblob%2Fmaster%2FArmPkg%2FLibrary%2FOpteeL
> > >
> ib%2FOptee.c%23L147&amp;data=02%7C01%7CChristopher.Co%40microsoft.c
> > >
> om%7C4a7d8c01e4804365f4eb08d640837a15%7C72f988bf86f141af91ab2d7cd
> > >
> 011db47%7C1%7C0%7C636767330779998429&amp;sdata=Lsplb1L7Ugd2C6cXG
> > > 8gBo40Ei8UQPtIA7fNEDL1t%2Fbg%3D&amp;reserved=0
> > >
> > > Regards,
> > > Sumit
> > >
> > > > Chris
> > > >
> > > > > -----Original Message-----
> > > > > From: Sumit Garg <sumit.g...@linaro.org>
> > > > > Sent: Thursday, November 1, 2018 3:55 AM
> > > > > To: Chris Co <christopher...@microsoft.com>; Leif Lindholm
> > > > > <leif.lindh...@linaro.org>
> > > > > Cc: edk2-devel@lists.01.org; Ard Biesheuvel
> > > > > <ard.biesheu...@linaro.org>; Michael D Kinney
> > > > > <michael.d.kin...@intel.com>
> > > > > Subject: Re: [PATCH edk2-platforms 01/27] Platform/Microsoft:
> > > > > Add OpteeClientPkg dec
> > > > >
> > > > > Hi Christopher,
> > > > >
> > > > > Optee Client library has recently been merged to edk2 source code.
> > > > > It tries to provide a generic interface [1] to OP-TEE based
> > > > > trusted applications (pseudo/early).
> > > > >
> > > > > AFAIK, you don't need any platform specific hook in client
> > > > > interface to work with upstream OP-TEE. So instead you should use
> Optee library.
> > > > >
> > > > > [1]
> > > > > https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2
> > > > > Fgit
> > > > > hub.c
> > > > >
> > >
> om%2Ftianocore%2Fedk2%2Fblob%2Fmaster%2FArmPkg%2FInclude%2FLibrary
> > > > >
> > >
> %2FOpteeLib.h&amp;data=02%7C01%7CChristopher.Co%40microsoft.com%7C
> > > > >
> > >
> c19b84ef7f8f4213424108d63fe88f66%7C72f988bf86f141af91ab2d7cd011db47
> > > > >
> > >
> %7C1%7C0%7C636766665404786500&amp;sdata=m24akbKtoyCERVN77meoSU
> > > > > H6E%2Bpf8W2P5MF7nvU5y7I%3D&amp;reserved=0
> > > > >
> > > > > Regards,
> > > > > Sumit
> > > > >
> > > > > On Thu, 1 Nov 2018 at 02:13, Leif Lindholm
> > > > > <leif.lindh...@linaro.org>
> > > wrote:
> > > > > >
> > > > > > +Sumit (just to loop you two together). Is there anything
> > > > > > +Microsoft
> > > > > > platform specific about what will go in here?
> > > > > >
> > > > > > /
> > > > > >     Leif
> > > > > >
> > > > > > On Fri, Sep 21, 2018 at 08:25:53AM +0000, Chris Co wrote:
> > > > > > > On Windows IoT Core devices with ARM TrustZone capabilities,
> > > > > > > EDK2 runs in normal world and we use OP-TEE to execute
> > > > > > > secure world operations. The overall package will contain
> > > > > > > client-side support to invoke EDK2 services implemented as
> > > > > > > OP-TEE trusted applications that run in secure world.
> > > > > > >
> > > > > > > This commit adds the initial dec file to add some PCD
> > > > > > > settings needed by other packages.
> > > > > > >
> > > > > > > Contributed-under: TianoCore Contribution Agreement 1.1
> > > > > > > Signed-off-by: Christopher Co <christopher...@microsoft.com>
> > > > > > > Cc: Ard Biesheuvel <ard.biesheu...@linaro.org>
> > > > > > > Cc: Leif Lindholm <leif.lindh...@linaro.org>
> > > > > > > Cc: Michael D Kinney <michael.d.kin...@intel.com>
> > > > > > > ---
> > > > > > >  Platform/Microsoft/OpteeClientPkg/OpteeClientPkg.dec | 49
> > > > > > > ++++++++++++++++++++
> > > > > > >  1 file changed, 49 insertions(+)
> > > > > > >
> > > > > > > diff --git
> > > > > > > a/Platform/Microsoft/OpteeClientPkg/OpteeClientPkg.dec
> > > > > > > b/Platform/Microsoft/OpteeClientPkg/OpteeClientPkg.dec
> > > > > > > new file mode 100644
> > > > > > > index 000000000000..4752eab39ce3
> > > > > > > --- /dev/null
> > > > > > > +++ b/Platform/Microsoft/OpteeClientPkg/OpteeClientPkg.dec
> > > > > > > @@ -0,0 +1,49 @@
> > > > > > > +## @file
> > > > > > > +#
> > > > > > > +#  OP-TEE client package
> > > > > > > +#
> > > > > > > +#  OP-TEE client package contains the client-side interface
> > > > > > > +to invoke OP-
> > > > > TEE TAs.
> > > > > > > +#  Certain EDKII services are implemented in Trusted
> > > > > > > +Applications running in #  the secure world OP-TEE OS.
> > > > > > > +#
> > > > > > > +#  Copyright (c) 2018 Microsoft Corporation. All rights reserved.
> > > > > > > +#
> > > > > > > +#  This program and the accompanying materials #  are
> > > > > > > +licensed and made available under the terms and conditions
> > > > > > > +of the BSD License # which accompanies this distribution.
> > > > > > > +The full text of the license may be found at #
> > > > > > > +https://na01.safelinks.protection.outlook.com/?url=http%3A%
> > > > > > > +2F%2
> > > > > > > +Fope
> > > > > > > +nsource.org%2Flicenses%2Fbsd-
> > > > > license.php&amp;data=02%7C01%7CChristo
> > > > > > >
> > > > >
> > >
> +pher.Co%40microsoft.com%7Cc19b84ef7f8f4213424108d63fe88f66%7C72f988
> > > > > > >
> > > > >
> > >
> +bf86f141af91ab2d7cd011db47%7C1%7C0%7C636766665404786500&amp;sda
> > > > > ta=1
> > > > > > >
> > > +MxFvlsMPhk19grEexBXo5VqRd0jZaCSRjxZCi87A2w%3D&amp;reserved=0
> > > > > > > +#
> > > > > > > +#  THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN
> > > > > > > +"AS
> > > IS"
> > > > > > > +BASIS, #  WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY
> > > > > > > +KIND,
> > > > > EITHER EXPRESS OR IMPLIED.
> > > > > > > +#
> > > > > > > +##
> > > > > > > +
> > > > > > > +[Defines]
> > > > > > > +  DEC_SPECIFICATION              = 0x0001001A
> > > > > > > +  PACKAGE_NAME                   = OpteeClientPkg
> > > > > > > +  PACKAGE_GUID                   = 77416fcb-10ec-4693-bdc0-
> > > 1bdd74ec9595
> > > > > > > +  PACKAGE_VERSION                = 0.01
> > > > > > > +
> > > > > > > +[Includes]
> > > > > > > +
> > > > > > > +[LibraryClasses]
> > > > > > > +
> > > > > > > +[Guids]
> > > > > > > +  gOpteeClientPkgTokenSpaceGuid   = { 0x04ad34ca, 0xdd25,
> 0x4156, {
> > > > > 0x90, 0xf5, 0x16, 0xf9, 0x40, 0xd0, 0x49, 0xe3 }}
> > > > > > > +
> > > > > > > +[PcdsFixedAtBuild]
> > > > > > > +
> > > > > > >
> > > > >
> > >
> +gOpteeClientPkgTokenSpaceGuid.PcdTpm2AcpiBufferBase|0|UINT64|0x0000
> > > > > > > +0005
> > > > > > > +
> > > > > > >
> > > > >
> > >
> +gOpteeClientPkgTokenSpaceGuid.PcdTpm2AcpiBufferSize|0|UINT32|0x0000
> > > > > > > +0006
> > > > > > > +
> > > > > > > +  ## The base address of the Trust Zone OpTEE OS private
> > > > > > > + memory region  # This memory is manager privately by the OpTEE
> OS.
> > > > > > > +
> > > > > > > +
> > > > >
> > >
> gOpteeClientPkgTokenSpaceGuid.PcdTrustZonePrivateMemoryBase|0xDEAD
> > > > > > > + 1|UINT64|0x00000001
> > > > > > > +
> > > > > > > +  ## The size of the Trust Zone OpTEE OS private memory
> > > > > > > + region
> > > > > > > +
> > > > > > > +
> > > > >
> gOpteeClientPkgTokenSpaceGuid.PcdTrustZonePrivateMemorySize|55|U
> > > > > IN
> > > > > > > + T64|0x00000002
> > > > > > > +
> > > > > > > +  ## The base address of the Trust Zone OpTEE OS shared
> > > > > > > + memory region
> > > > > > > +
> > > > > > > +
> > > > >
> > >
> gOpteeClientPkgTokenSpaceGuid.PcdTrustZoneSharedMemoryBase|0xDEAD2
> > > > > > > + |UINT64|0x00000003
> > > > > > > +
> > > > > > > +  ## The size of the Trust Zone OpTEE OS shared memory
> > > > > > > + region
> > > > > > > +
> > > > > > > +
> > > > >
> > >
> gOpteeClientPkgTokenSpaceGuid.PcdTrustZoneSharedMemorySize|0xAA|UI
> > > > > > > + NT64|0x00000004
> > > > > > > --
> > > > > > > 2.16.2.gvfs.1.33.gf5370f1
> > > > > > >
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel

Reply via email to