Security Working Group meeting - Wednesday December 8

Security Working Group meeting - Wednesday December 8

[Joseph Reynolds]

This is a reminder of the OpenBMC Security Working Group meeting 
scheduled for this Wednesday December 8 at 10:00am PDT.

We'll discuss the following items on the agenda 
<https://docs.google.com/document/d/1b7x9BaxsfcukQDqbvZsU2ehMq4xoJRQvLxxsDUWmAOI>, 
and anything else that comes up:

1.



Access, agenda and notes are in the wiki:
https://github.com/openbmc/openbmc/wiki/Security-working-group 
<https://github.com/openbmc/openbmc/wiki/Security-working-group>

- Joseph

Re: Security Working Group meeting - Wednesday December 8 - results

[Joseph Reynolds]

On 12/7/21 3:55 PM, Joseph Reynolds wrote:
> This is a reminder of the OpenBMC Security Working Group meeting 
> scheduled for this Wednesday December 8 at 10:00am PDT.
>
> We'll discuss the following items on the agenda 
> <https://docs.google.com/document/d/1b7x9BaxsfcukQDqbvZsU2ehMq4xoJRQvLxxsDUWmAOI>, 
> and anything else that comes up:
>

Attendance: James, Joseph, Anton, Dhananjay, Ratan


1 OpenBMC CNA onboarding

DISCUSSION:

James started the process to onboard the OpenBMC project as a CNA.  (See 
agenda item 2 from 2021-11-10.)   Onboarding process is next week for 
James, Joseph, and Dhananjay.

Onboarding time commitment: unknown  -  watch  videos

Here are the training links:

please view the six on-boarding videos, available on the CNA On-Boarding 
Channel on YouTube--> Click 
here<https://www.youtube.com/playlist?list=PLWfD9RQVdJ6c4eMkdqbOKqF7zPCqXkgX3 
<https://www.youtube.com/playlist?list=PLWfD9RQVdJ6c4eMkdqbOKqF7zPCqXkgX3>>


   1.  CVE Program Overview

   2.  Becoming a CNA

   3.  CNA Processes

   4.  Assigning CVE IDs

   5.  CVE Record (previously “CVE Entry”) Creation

   6.  CVE Record Submission Process to the MITRE Top-Level Root Only

      *   CVE Record (previously “CVE Entry”) GitHub Submissions


Softcopies of the presentations are available here  
(https://www.cve.org/ResourcesSupport/Resources#CVENumberingAuthorities 
<https://www.cve.org/ResourcesSupport/Resources#CVENumberingAuthorities>)


2 Daemon privilege separation design doc for review 
<https://gerrit.openbmc-project.xyz/c/openbmc/docs/+/49100>(PoC 
<https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/42748>change for 
ACLs)

DISCUSSION:

This is a multi-stage project, and having a design will make it easier 
to move forward.

Next steps:

  *

    approve design doc (need reviewers),

  *

    then write acl rules

  *

    Then change process to an unique user

  *

    List all services which need to participate  - all D-bus service
    owners and clients

  *

    Move to a role-based approach?

Idea: Complete the privilege separation work for a service to use as a 
model for other services.  When this is done, repo maintainers will have 
an easier time to understand what changes are needed.

We briefly talked through an example set of rules for bmcweb and  ipmid 
talking to phosphor-user-manager.


3 Move meeting earlier by 1 hour?  Let’s renegotiate the meeting time.


4 Progress on BMC secure boot?

AST2600 hardware secure U-boot  boot, then secure booting the Linux 
kernel. No additional pieces.

See the AST security guide.  How is signing-key management done?

Dhananjay to  follow up.



Joseph


>
>
> Access, agenda and notes are in the wiki:
> https://github.com/openbmc/openbmc/wiki/Security-working-group 
> <https://github.com/openbmc/openbmc/wiki/Security-working-group>
>
> - Joseph

Re: Security Working Group meeting - Wednesday December 8 - results

[Andrew Jeffery]

On Thu, 9 Dec 2021, at 05:44, Joseph Reynolds wrote:

>
> 4 Progress on BMC secure boot?
>
> AST2600 hardware secure U-boot  boot, then secure booting the Linux 
> kernel. No additional pieces.
>
> See the AST security guide.  How is signing-key management done?
>
> Dhananjay to  follow up.

As someone who was involved in integrating the AST2600 secure-boot support into OpenBMC, what's going on here?

Andrew

Re: Security Working Group meeting - Wednesday December 8 - results

[Joseph Reynolds]

On 12/9/21 4:21 AM, Andrew Jeffery wrote:
> On Thu, 9 Dec 2021, at 05:44, Joseph Reynolds wrote:
>
>> 4 Progress on BMC secure boot?
>>
>> AST2600 hardware secure U-boot  boot, then secure booting the Linux
>> kernel. No additional pieces.
>>
>> See the AST security guide.  How is signing-key management done?
>>
>> Dhananjay to  follow up.
> As someone who was involved in integrating the AST2600 secure-boot support into OpenBMC, what's going on here?

Someone asked the question you see above, and Dhananjay tried to 
answer.  The consensus was that there is support for AST2600 secure 
booting U-Boot, support for U-Boot securely loading the Linux kernel, 
and no additional support.  Your summary would be appreciated.

Is there a document which what a system integrator needs to use this 
function?  Ideally it would be linked from 
https://github.com/openbmc/docs/blob/master/features.md

Joseph

>
> Andrew

Re: Security Working Group meeting - Wednesday December 8 - results

[Dhananjay Phadke]

>> On Thu, 9 Dec 2021, at 05:44, Joseph Reynolds wrote:
>>
>>> 4 Progress on BMC secure boot?
>>>
>>> AST2600 hardware secure U-boot  boot, then secure booting the Linux
>>> kernel. No additional pieces.
>>>
>>> See the AST security guide.  How is signing-key management done?
>>>
>>> Dhananjay to  follow up.
>> As someone who was involved in integrating the AST2600 secure-boot support into OpenBMC, what's going on here?
>
>Someone asked the question you see above, and Dhananjay tried to 
>answer.  The consensus was that there is support for AST2600 secure 
>booting U-Boot, support for U-Boot securely loading the Linux kernel, 
>and no additional support.  Your summary would be appreciated.
>
>Is there a document which what a system integrator needs to use this 
>function?  Ideally it would be linked from 
>https://github.com/openbmc/docs/blob/master/features.md

Right, I noted recent submissions to U-Boot and Kernel.
(1) HACE/ARCY support in U-Boot
(2) OTP sysfs access for logging Secure Boot status.

Need clarity regarding OTP programming.
(1) There's Linux tool and U-Boot patches floating somewhere.
(2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
boot (ABR).
(3) Any interest in using encryption besides SHA/RSA auth?

Signing of FIT is handled by Yocto/poky anyway.

Dhananjay

Re: Security Working Group meeting - Wednesday December 8 - results

[Andrew Jeffery]

On Fri, 10 Dec 2021, at 04:01, Dhananjay Phadke wrote:
>>> On Thu, 9 Dec 2021, at 05:44, Joseph Reynolds wrote:
>>>
>>>> 4 Progress on BMC secure boot?
>>>>
>>>> AST2600 hardware secure U-boot  boot, then secure booting the Linux
>>>> kernel. No additional pieces.
>>>>
>>>> See the AST security guide.  How is signing-key management done?
>>>>
>>>> Dhananjay to  follow up.
>>> As someone who was involved in integrating the AST2600 secure-boot support into OpenBMC, what's going on here?
>>
>>Someone asked the question you see above, and Dhananjay tried to 
>>answer.  The consensus was that there is support for AST2600 secure 
>>booting U-Boot, support for U-Boot securely loading the Linux kernel, 
>>and no additional support.  Your summary would be appreciated.
>>
>>Is there a document which what a system integrator needs to use this 
>>function?  Ideally it would be linked from 
>>https://github.com/openbmc/docs/blob/master/features.md

There's not much documentation as yet. p10bmc can be used as an example 
of a system that enables it.

https://github.com/openbmc/openbmc/blob/ade3e145ead0beedad181394fcaa63856176bdee/meta-ibm/conf/machine/p10bmc.conf#L39-L56

Given the lack of documentation it's probably also reviewing these 
patches in the context of the configuration above:

https://gerrit.openbmc-project.xyz/q/topic:%22secure-boot%22+(status:open%20OR%20status:merged)

>
> Right, I noted recent submissions to U-Boot and Kernel.
> (1) HACE/ARCY support in U-Boot
> (2) OTP sysfs access for logging Secure Boot status.
>
> Need clarity regarding OTP programming.
> (1) There's Linux tool

I assume this refers to socsec? The socsec repo provides two tools: 
`socsec` and `otptool`. `otptool` can be used to generate the OTP image 
and exercise signature validity.

https://github.com/AspeedTech-BMC/socsec/

> and U-Boot patches floating somewhere.

I'm not sure what patches you're referring to here, can you clarify?

> (2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
> boot (ABR).

There's no real preference. My intent is to add a recipe that can 
consume a platform-specific otptool json config and spit out the OTP 
binary as a build artefact. Currently I just have the config captured 
in a separate repo internally and I generate binaries from that using 
make.

> (3) Any interest in using encryption besides SHA/RSA auth?

SHA/RSA is what IBM will ship with in current platforms, so not from us.

Andrew

RE: Security Working Group meeting - Wednesday December 8 - results

[Troy Lee]

Hi,
> -----Original Message-----
> From: openbmc <openbmc-
> bounces+troy_lee=aspeedtech.com@lists.ozlabs.org> On Behalf Of Andrew
> Jeffery
> Sent: Friday, December 10, 2021 7:35 AM
> To: Dhananjay Phadke <dphadke@linux.microsoft.com>; Joseph Reynolds
> <jrey@linux.ibm.com>
> Cc: openbmc@lists.ozlabs.org
> Subject: Re: Security Working Group meeting - Wednesday December 8 -
> results
> 
> 
> 
> On Fri, 10 Dec 2021, at 04:01, Dhananjay Phadke wrote:
> >>> On Thu, 9 Dec 2021, at 05:44, Joseph Reynolds wrote:
> >>>
> >>>> 4 Progress on BMC secure boot?
> >>>>
> >>>> AST2600 hardware secure U-boot  boot, then secure booting the Linux
> >>>> kernel. No additional pieces.
> >>>>
> >>>> See the AST security guide.  How is signing-key management done?
> >>>>
> >>>> Dhananjay to  follow up.
> >>> As someone who was involved in integrating the AST2600 secure-boot
> support into OpenBMC, what's going on here?
> >>
> >>Someone asked the question you see above, and Dhananjay tried to
> >>answer.  The consensus was that there is support for AST2600 secure
> >>booting U-Boot, support for U-Boot securely loading the Linux kernel,
> >>and no additional support.  Your summary would be appreciated.
> >>
> >>Is there a document which what a system integrator needs to use this
> >>function?  Ideally it would be linked from
> >>https://github.com/openbmc/docs/blob/master/features.md
> 
> There's not much documentation as yet. p10bmc can be used as an example
> of a system that enables it.
> 
> https://github.com/openbmc/openbmc/blob/ade3e145ead0beedad181394f
> caa63856176bdee/meta-ibm/conf/machine/p10bmc.conf#L39-L56
> 
> Given the lack of documentation it's probably also reviewing these patches in
> the context of the configuration above:
> 
> https://gerrit.openbmc-project.xyz/q/topic:%22secure-
> boot%22+(status:open%20OR%20status:merged)
> 
> >
> > Right, I noted recent submissions to U-Boot and Kernel.
> > (1) HACE/ARCY support in U-Boot
> > (2) OTP sysfs access for logging Secure Boot status.
> >
> > Need clarity regarding OTP programming.
> > (1) There's Linux tool
> 
> I assume this refers to socsec? The socsec repo provides two tools:
> `socsec` and `otptool`. `otptool` can be used to generate the OTP image and
> exercise signature validity.
> 
> https://github.com/AspeedTech-BMC/socsec/
> 
> > and U-Boot patches floating somewhere.
> 
> I'm not sure what patches you're referring to here, can you clarify?
> 
> > (2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
> > boot (ABR).
> 
> There's no real preference. My intent is to add a recipe that can consume a
> platform-specific otptool json config and spit out the OTP binary as a build
> artefact. Currently I just have the config captured in a separate repo
> internally and I generate binaries from that using make.
> 
> > (3) Any interest in using encryption besides SHA/RSA auth?
> 
> SHA/RSA is what IBM will ship with in current platforms, so not from us.
> 
> Andrew

There is a secure boot document in review.
https://gerrit.openbmc-project.xyz/c/openbmc/docs/+/26169

The secure boot hardware verifies root of trust image, i.e. u-boot-spl.bin, and chain-of-trust image (u-boot.bin and kernel/initramfs) are verified by u-boot verified boot feature.

My questions is how does root fs being verified? Are you using UBI FS authentication support for this purpose?

Thanks,
Troy Lee

Re: Security Working Group meeting - Wednesday December 8 - results

[Dhananjay Phadke]

On Fri, 10 Dec 2021, Andrew Jeffery wrote:

> There's not much documentation as yet. p10bmc can be used as an example
> of a system that enables it.
>
> https://github.com/openbmc/openbmc/blob/ade3e145ead0beedad181394fcaa63856176bdee/meta-ibm/conf/machine/p10bmc.conf#L39-L56
>
> Given the lack of documentation it's probably also reviewing these
> patches in the context of the configuration above:
>
> https://gerrit.openbmc-project.xyz/q/topic:%22secure-boot%22+(status:open%20OR%20status:merged)

Thank you for the pointer, I'll comments there.

>> Need clarity regarding OTP programming.
>> (1) There's Linux tool
>
> I assume this refers to socsec? The socsec repo provides two tools:
> `socsec` and `otptool`. `otptool` can be used to generate the OTP image
> and exercise signature validity.
>
> https://github.com/AspeedTech-BMC/socsec/

Yes, I was referring to these tools, socsec-sign.bbclass seems to cover
the workflow I was looking for.

>
>> and U-Boot patches floating somewhere.
>
> I'm not sure what patches you're referring to here, can you clarify?

https://github.com/AspeedTech-BMC/u-boot/commits/aspeed-master-v2019.04

cmd/otp.c has more changes compared to openbmc/u-boot.

>
>> (2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
>> boot (ABR).
>
> There's no real preference. My intent is to add a recipe that can
> consume a platform-specific otptool json config and spit out the OTP
> binary as a build artefact. Currently I just have the config captured
> in a separate repo internally and I generate binaries from that using
> make.

This is useful, having readable config and letting platform select
behavior such as alternate image in same SPI or alternate, etc.

Regards,
Dhananjay

Re: Security Working Group meeting - Wednesday December 8 - results

[Andrew Jeffery]

On Fri, 10 Dec 2021, at 12:19, Troy Lee wrote:
> Hi,
>> -----Original Message-----
>> From: openbmc <openbmc-
>> bounces+troy_lee=aspeedtech.com@lists.ozlabs.org> On Behalf Of Andrew
>> Jeffery
>> Sent: Friday, December 10, 2021 7:35 AM
>> To: Dhananjay Phadke <dphadke@linux.microsoft.com>; Joseph Reynolds
>> <jrey@linux.ibm.com>
>> Cc: openbmc@lists.ozlabs.org
>> Subject: Re: Security Working Group meeting - Wednesday December 8 -
>> results

>
> My questions is how does root fs being verified? Are you using UBI FS 
> authentication support for this purpose?

Verification of the rootfs is TBD, though we did experiment with dm-verity.

Andrew

Re: Security Working Group meeting - Wednesday December 8 - results

[Andrew Jeffery]

On Fri, 10 Dec 2021, at 15:31, Dhananjay Phadke wrote:
> On Fri, 10 Dec 2021, Andrew Jeffery wrote:
>> I'm not sure what patches you're referring to here, can you clarify?
>
> https://github.com/AspeedTech-BMC/u-boot/commits/aspeed-master-v2019.04
>
> cmd/otp.c has more changes compared to openbmc/u-boot.

Okay, I expect we'll sync up at some point.

Andrew

RE: Security Working Group meeting - Wednesday December 8 - results

[Troy Lee]

Hi Dhananjay and Andrew,

> On Fri, 10 Dec 2021, Andrew Jeffery wrote:
> 
> > There's not much documentation as yet. p10bmc can be used as an
> > example of a system that enables it.
> >
> >
> https://github.com/openbmc/openbmc/blob/ade3e145ead0beedad181394f
> caa63
> > 856176bdee/meta-ibm/conf/machine/p10bmc.conf#L39-L56
> >
> > Given the lack of documentation it's probably also reviewing these
> > patches in the context of the configuration above:
> >
> > https://gerrit.openbmc-project.xyz/q/topic:%22secure-boot%22+(status:o
> > pen%20OR%20status:merged)
> 
> Thank you for the pointer, I'll comments there.
> 
> >> Need clarity regarding OTP programming.
> >> (1) There's Linux tool
> >
> > I assume this refers to socsec? The socsec repo provides two tools:
> > `socsec` and `otptool`. `otptool` can be used to generate the OTP
> > image and exercise signature validity.
> >
> > https://github.com/AspeedTech-BMC/socsec/
> 
> Yes, I was referring to these tools, socsec-sign.bbclass seems to cover the
> workflow I was looking for.
> 
> >
> >> and U-Boot patches floating somewhere.
> >
> > I'm not sure what patches you're referring to here, can you clarify?
> 
> https://github.com/AspeedTech-BMC/u-boot/commits/aspeed-master-
> v2019.04
> 
> cmd/otp.c has more changes compared to openbmc/u-boot.
> 
> >
> >> (2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
> >> boot (ABR).
> >
> > There's no real preference. My intent is to add a recipe that can
> > consume a platform-specific otptool json config and spit out the OTP
> > binary as a build artefact. Currently I just have the config captured
> > in a separate repo internally and I generate binaries from that using
> > make.
> 
> This is useful, having readable config and letting platform select behavior
> such as alternate image in same SPI or alternate, etc.

We have the recipe to generate otp-image with OpenBMC building
process and is working on upstreaming it into OpenBMC.
Could you help to review the recipe once we submit it into gerrit?

> 
> Regards,
> Dhananjay

Thanks,
Troy Lee

Re: Security Working Group meeting - Wednesday December 8 - results

[Andrew Jeffery]

On Fri, 10 Dec 2021, at 16:25, Troy Lee wrote:
> Hi Dhananjay and Andrew,
>
>> On Fri, 10 Dec 2021, Andrew Jeffery wrote:
>> 
>> > There's not much documentation as yet. p10bmc can be used as an
>> > example of a system that enables it.
>> >
>> >
>> https://github.com/openbmc/openbmc/blob/ade3e145ead0beedad181394f
>> caa63
>> > 856176bdee/meta-ibm/conf/machine/p10bmc.conf#L39-L56
>> >
>> > Given the lack of documentation it's probably also reviewing these
>> > patches in the context of the configuration above:
>> >
>> > https://gerrit.openbmc-project.xyz/q/topic:%22secure-boot%22+(status:o
>> > pen%20OR%20status:merged)
>> 
>> Thank you for the pointer, I'll comments there.
>> 
>> >> Need clarity regarding OTP programming.
>> >> (1) There's Linux tool
>> >
>> > I assume this refers to socsec? The socsec repo provides two tools:
>> > `socsec` and `otptool`. `otptool` can be used to generate the OTP
>> > image and exercise signature validity.
>> >
>> > https://github.com/AspeedTech-BMC/socsec/
>> 
>> Yes, I was referring to these tools, socsec-sign.bbclass seems to cover the
>> workflow I was looking for.
>> 
>> >
>> >> and U-Boot patches floating somewhere.
>> >
>> > I'm not sure what patches you're referring to here, can you clarify?
>> 
>> https://github.com/AspeedTech-BMC/u-boot/commits/aspeed-master-
>> v2019.04
>> 
>> cmd/otp.c has more changes compared to openbmc/u-boot.
>> 
>> >
>> >> (2) Any specific OTP straps preferred by OpenBMC, e.g. enabling alt
>> >> boot (ABR).
>> >
>> > There's no real preference. My intent is to add a recipe that can
>> > consume a platform-specific otptool json config and spit out the OTP
>> > binary as a build artefact. Currently I just have the config captured
>> > in a separate repo internally and I generate binaries from that using
>> > make.
>> 
>> This is useful, having readable config and letting platform select behavior
>> such as alternate image in same SPI or alternate, etc.
>
> We have the recipe to generate otp-image with OpenBMC building
> process and is working on upstreaming it into OpenBMC.
> Could you help to review the recipe once we submit it into gerrit?

Yes, please push it for review.

Andrew