[virtio-dev] VIRTIO RPMB Front End drivers for Linux

[virtio-dev] VIRTIO RPMB Front End drivers for Linux

[Alex Bennée]

Hi,

I'm currently working on a vhost-user backend for virto-rpmb as defined
in the draft version of the OASIS spec (since 8c6acac2). I'm developing
it within the QEMU source tree but it is intended to be a portable
vhost-user binary with minimal POSIX like dependencies.

Part of the aim of the project is to experiment with a vhost-user
backend that has minimal access to guest memory and RPMB makes sense as
the vhost-user daemon only really needs to see the virt queues to pass
the data back and forth. One eventual architecture would be to move the
vhost-user driver into it's own isolated VM with only minimal memory
sharing between the guest and it's back-end.

I've gotten enough of the plumbing working now that I'm ready to start
implementing the RPMB spec itself. I've been looking for a front-end to
use in testing and I found the Linux driver sitting in Project ACRN's
guest kernel fork. With this kernel and VIRTIO_ID_RPMB tweaked to the
standard based one I get far enough that the driver is initialised
before my drivers falls over when an un-handled feature is requested
(0x130000000). This is slightly confusing because the spec doesn't
define any feature bits. So a few questions:

  - how divergent is the virtio-rpmb front/back-end in ACRN from the
    OASIS spec?

  - is there likely to be another submission to the upstream kernel?

I've also read the document on the ACRN website which talks about the
secure architecture:

  https://projectacrn.github.io/latest/tutorials/trustyACRN.html

I was wondering if there was any low level usage information. From
looking at the Kconfig options it seems the virtio-rpmb driver
interfaces with a wider RPMB sub-system which exposes a character device
to the kernels user-space. So:

  - are there any special user-space tools to use this interface?

  - is there a write-up somewhere of how to exercise the interface or
    any sort of test suite?

Thanks in advance,


-- 
Alex Bennée

---------------------------------------------------------------------
To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org

[virtio-dev] Re: VIRTIO RPMB Front End drivers for Linux

[Alex Bennée]

Winkler, Tomas <tomas.winkler@intel.com> writes:

> Hi, yes the OASIS spec is based around this ACRN solution. We are still planning on another Linux kernel submission effort, actually in the next release cycle. 
> The capability message which is not implemented in the virtio solution is due to NVMe support, which didn't make into OASIS spec. 
> There is an rpmb tool under tools/rpmb of the Linux kernel, also AOS guest on ACRN uses this interface, there should be doc under Documentation/driver-api/rpmb/. 

Thanks for the pointers they are really helpful.

> I'm not sure how updated is ACRN public tree, but I'll post
> https://github.com/tomasbw/linux-mei rpmb branch, I hope in a week or
> two.

Thanks. Looking at the ACRN tree when their base rpmb commit (01abd0e5)
as added on a 5.4 based machine. The commits have V10 references bellow
the --- which implies some changes from your series against v4.14-rc5.

I've cherry-picked the initial ones out of the ACRN tree and they seemed
to re-base onto v5.6.7 reasonably enough (untested branch):

  https://git.linaro.org/people/alex.bennee/linux.git/log/?h=review/rpmb

There are a bunch more commits touching that in the ACRN tree but
hopefully this will give me a minimum testing set without having
everything else that's in there.

It does make me wonder if any of these patches are on an upstream path
or are they likely to live in 3rd party trees forever?

> There is also a simulation HW drivers/char/rpmb/rpmb_sim.C it's based on eMMC part of the spec. 
> I do have some very simple tests to exercise this interface.  
>
> Thanks
> Tomas
>
>
>
>> -----Original Message-----
>> From: Alex Bennée <alex.bennee@linaro.org>
>> Sent: Friday, May 08, 2020 19:12
>> To: virtio-dev@lists.oasis-open.org; Winkler, Tomas
>> <tomas.winkler@intel.com>; Huang, Yang <yang.huang@intel.com>
>> Cc: Zhu, Bing <bing.zhu@intel.com>; Michael S.Tsirkin <mst@redhat.com>; Ilias
>> Apalodimas <ilias.apalodimas@linaro.org>
>> Subject: VIRTIO RPMB Front End drivers for Linux
>> 
>> 
>> Hi,
>> 
>> I'm currently working on a vhost-user backend for virto-rpmb as defined in the
>> draft version of the OASIS spec (since 8c6acac2). I'm developing it within the
>> QEMU source tree but it is intended to be a portable vhost-user binary with
>> minimal POSIX like dependencies.
>> 
>> Part of the aim of the project is to experiment with a vhost-user backend that
>> has minimal access to guest memory and RPMB makes sense as the vhost-user
>> daemon only really needs to see the virt queues to pass the data back and forth.
>> One eventual architecture would be to move the vhost-user driver into it's own
>> isolated VM with only minimal memory sharing between the guest and it's back-
>> end.
>> 
>> I've gotten enough of the plumbing working now that I'm ready to start
>> implementing the RPMB spec itself. I've been looking for a front-end to use in
>> testing and I found the Linux driver sitting in Project ACRN's guest kernel fork.
>> With this kernel and VIRTIO_ID_RPMB tweaked to the standard based one I get
>> far enough that the driver is initialised before my drivers falls over when an un-
>> handled feature is requested (0x130000000). This is slightly confusing because
>> the spec doesn't define any feature bits. So a few questions:
>> 
>>   - how divergent is the virtio-rpmb front/back-end in ACRN from the
>>     OASIS spec?
>> 
>>   - is there likely to be another submission to the upstream kernel?
>> 
>> I've also read the document on the ACRN website which talks about the secure
>> architecture:
>> 
>>   https://projectacrn.github.io/latest/tutorials/trustyACRN.html
>> 
>> I was wondering if there was any low level usage information. From looking at
>> the Kconfig options it seems the virtio-rpmb driver interfaces with a wider RPMB
>> sub-system which exposes a character device to the kernels user-space. So:
>> 
>>   - are there any special user-space tools to use this interface?
>> 
>>   - is there a write-up somewhere of how to exercise the interface or
>>     any sort of test suite?
>> 
>> Thanks in advance,
>> 
>> 
>> --
>> Alex Bennée


-- 
Alex Bennée

---------------------------------------------------------------------
To unsubscribe, e-mail: virtio-dev-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: virtio-dev-help@lists.oasis-open.org