Plans for BMC i2c to host bridge via IPMI

Plans for BMC i2c to host bridge via IPMI

[Peter Hanson]

Ave!

This email captures proposed actions from an exchange earlier today on
the #openbmc IRC channel.

Basic goal is to reach devices on a BMC I2C bus from IPMI commandline at host.

I2C commands and responses to be carried as OEM extension messages,
i.e., Network Function Codes 2Eh / 2Fh in the Intel specification.
Those message forms require a three byte IANA Enterprise Number, and
our original plan used Google's code.

We would like to carry the feature in OpenBmc, so I wanted to confirm
if this is acceptable, and/or what modifications would be needed.

Decisions:

1. Ok to use OEM extensions. Patrick noted that OpenBmc already uses
some, albeit under the IBM IANA Enterprise Number.

2. Switch to OpenBmc IANA Enterprise Number when available.

3. Patrick proposed to create a new phosphor-ipmi-oemproviders repo to
hold the code, message documentation, etc.

4. All uses of the OpenBmc IANA shall be enumerated in place - almost
certainly an include file in the new repo - so we don't end up with
conflicting uses.

Actions:

A1. peterh: send this email to the mailing list.
A2. stwcx: request IANA Enterprise Number.
A3. stwcx: create new phosphor-ipmi-oemproviders repo.
A4. peterh+brendanhiggins: propose detailed message designs.
A5. peterh: {net-ipmid,host-ipmi} += support for these commands.

A key point of this email is to reap your collective wisdom, so please
if you have something to contribute, please do. In particular,
Patrick, please correct anything I misunderstood or incorrectly
extrapolated from the chat.

  -- peterh

Re: Plans for BMC i2c to host bridge via IPMI

[tomjose]

Hello Peter,

As i understand you would be leveraging the BT(Block Transfer) interface 
to route commands from host to BMC.

 From the IPMI specification it looks like this can be a generic one 
implementation instead of implementing
OEM commands.

Implementing the IPMB interface (Section 7) in the IPMI specification 
would probably suit your requirement.
The following commands should help us achieve the goal: Master 
Write-Read command, Get Message command
and Send message command.

I am thinking that IPMB interface could be a dbus service which parses 
host IPMI commands and maps to an I2C read/write.

Let me know your thoughts.

Regards,
Tom


On Thursday 06 April 2017 07:10 AM, Peter Hanson wrote:
> Ave!
>
> This email captures proposed actions from an exchange earlier today on
> the #openbmc IRC channel.
>
> Basic goal is to reach devices on a BMC I2C bus from IPMI commandline at host.
>
> I2C commands and responses to be carried as OEM extension messages,
> i.e., Network Function Codes 2Eh / 2Fh in the Intel specification.
> Those message forms require a three byte IANA Enterprise Number, and
> our original plan used Google's code.
>
> We would like to carry the feature in OpenBmc, so I wanted to confirm
> if this is acceptable, and/or what modifications would be needed.
>
> Decisions:
>
> 1. Ok to use OEM extensions. Patrick noted that OpenBmc already uses
> some, albeit under the IBM IANA Enterprise Number.
>
> 2. Switch to OpenBmc IANA Enterprise Number when available.
>
> 3. Patrick proposed to create a new phosphor-ipmi-oemproviders repo to
> hold the code, message documentation, etc.
>
> 4. All uses of the OpenBmc IANA shall be enumerated in place - almost
> certainly an include file in the new repo - so we don't end up with
> conflicting uses.
>
> Actions:
>
> A1. peterh: send this email to the mailing list.
> A2. stwcx: request IANA Enterprise Number.
> A3. stwcx: create new phosphor-ipmi-oemproviders repo.
> A4. peterh+brendanhiggins: propose detailed message designs.
> A5. peterh: {net-ipmid,host-ipmi} += support for these commands.
>
> A key point of this email is to reap your collective wisdom, so please
> if you have something to contribute, please do. In particular,
> Patrick, please correct anything I misunderstood or incorrectly
> extrapolated from the chat.
>
>    -- peterh
>

Re: Plans for BMC i2c to host bridge via IPMI

[Peter Hanson]

On Wed, Apr 12, 2017 at 12:51 PM, tomjose <tomjose@linux.vnet.ibm.com> wrote:
> Hello Peter,
>
> As i understand you would be leveraging the BT(Block Transfer) interface to
> route commands from host to BMC.
>
> From the IPMI specification it looks like this can be a generic one
> implementation instead of implementing
> OEM commands.
>
> Implementing the IPMB interface (Section 7) in the IPMI specification would
> probably suit your requirement.
> The following commands should help us achieve the goal: Master Write-Read
> command, Get Message command
> and Send message command.
>
> I am thinking that IPMB interface could be a dbus service which parses host
> IPMI commands and maps to an I2C read/write.

Choosing the best interface is a very important early step, so I very
much appreciate specific suggestions in that area - like this - and
will be as specific as possible in my response, so that I can promptly
correct holes in my thinking.

So here goes:

1. IPMB plumbing alone is significantly heavier than this feature. The
resulting Send Message and Get Message API seems awkward for our
simple case. Someone with actual IPMB use cases should drive that
work.

2. The key message the host would want to send - Master Write-Read
Command - does not address the full problem at a modern BMC. My
reading of Table 22- in section 22.11 puts the maximum number of
private SMBuses at 8. We have 9 i2c _controllers_ on a typical system;
some are further connected to multiplexers, further multiplying the
number of logical buses.

Our proposed OEM master write-read feature addresses both concerns.

> Let me know your thoughts.
>
> Regards,
> Tom

Re: Plans for BMC i2c to host bridge via IPMI

[Anton D. Kachalov]

[-- Attachment #1: Type: text/html, Size: 4453 bytes --]

Re: Plans for BMC i2c to host bridge via IPMI

[Chris Austen]

[-- Attachment #1: Type: text/plain, Size: 4172 bytes --]


Anton,
you have it licensed as GPL.  Tom will not be able to use your code as the
base code for OpenBMC is Apache licensed.

Chris Austen
POWER Systems Enablement Manager


"openbmc" <openbmc-bounces+austenc=us.ibm.com@lists.ozlabs.org> wrote on
04/17/2017 04:31:52 AM:

> From: "Anton D. Kachalov" <mouse@yandex-team.ru>
> To: tomjose <tomjose@linux.vnet.ibm.com>, Peter Hanson
> <peterh@google.com>, OpenBMC Maillist <openbmc@lists.ozlabs.org>
> Date: 04/17/2017 04:41 AM
> Subject: Re: Plans for BMC i2c to host bridge via IPMI
> Sent by: "openbmc" <openbmc-bounces+austenc=us.ibm.com@lists.ozlabs.org>
>
> Hello,
>
> I'm already works on this approach:
>
> https://github.com/ya-mouse/meta-openbmc-yandex/blob/master/meta-
>
yandex/meta-openrack/meta-shaosi/recipes-kernel/linux/linux-obmc/ipmi_i2c.c
>
> This creates a OpenIPMI-compatible device for the requested i2c
> slave addr + master (you may use ipmitool/freeipmi after all). DTS
example:
>
> https://github.com/ya-mouse/meta-openbmc-yandex/blob/master/meta-
> yandex/meta-openrack/meta-shaosi/recipes-kernel/linux/linux-obmc/
> overlays/shaosi-CB.dts#L121-L125
>
> In order to operate in multi-master env, you also need additional
> patche such as:
> https://github.com/ya-mouse/meta-openbmc-yandex/blob/master/meta-
> yandex/meta-openrack/meta-shaosi/recipes-kernel/linux/linux-obmc/
> 086-ipmi-hacks.patch
>
> This driver still has some issues + not supported incoming requests
> (it only operates like a "client").
>
> 12.04.2017, 22:51, "tomjose" <tomjose@linux.vnet.ibm.com>:
> Hello Peter,
>
> As i understand you would be leveraging the BT(Block Transfer) interface
> to route commands from host to BMC.
>
>  From the IPMI specification it looks like this can be a generic one
> implementation instead of implementing
> OEM commands.
>
> Implementing the IPMB interface (Section 7) in the IPMI specification
> would probably suit your requirement.
> The following commands should help us achieve the goal: Master
> Write-Read command, Get Message command
> and Send message command.
>
> I am thinking that IPMB interface could be a dbus service which parses
> host IPMI commands and maps to an I2C read/write.
>
> Let me know your thoughts.
>
> Regards,
> Tom
>
>
> On Thursday 06 April 2017 07:10 AM, Peter Hanson wrote:
>  Ave!
>
>  This email captures proposed actions from an exchange earlier today on
>  the #openbmc IRC channel.
>
>  Basic goal is to reach devices on a BMC I2C bus from IPMI
> commandline at host.
>
>  I2C commands and responses to be carried as OEM extension messages,
>  i.e., Network Function Codes 2Eh / 2Fh in the Intel specification.
>  Those message forms require a three byte IANA Enterprise Number, and
>  our original plan used Google's code.
>
>  We would like to carry the feature in OpenBmc, so I wanted to confirm
>  if this is acceptable, and/or what modifications would be needed.
>
>  Decisions:
>
>  1. Ok to use OEM extensions. Patrick noted that OpenBmc already uses
>  some, albeit under the IBM IANA Enterprise Number.
>
>  2. Switch to OpenBmc IANA Enterprise Number when available.
>
>  3. Patrick proposed to create a new phosphor-ipmi-oemproviders repo to
>  hold the code, message documentation, etc.
>
>  4. All uses of the OpenBmc IANA shall be enumerated in place - almost
>  certainly an include file in the new repo - so we don't end up with
>  conflicting uses.
>
>  Actions:
>
>  A1. peterh: send this email to the mailing list.
>  A2. stwcx: request IANA Enterprise Number.
>  A3. stwcx: create new phosphor-ipmi-oemproviders repo.
>  A4. peterh+brendanhiggins: propose detailed message designs.
>  A5. peterh: {net-ipmid,host-ipmi} += support for these commands.
>
>  A key point of this email is to reap your collective wisdom, so please
>  if you have something to contribute, please do. In particular,
>  Patrick, please correct anything I misunderstood or incorrectly
>  extrapolated from the chat.
>
>     -- peterh
>
>
>
>
> --
> Anton D. Kachalov
>
> ITO, Systems Architect
> Tel: 7 (495) 739-70-00 ext.7613
>

[-- Attachment #2: Type: text/html, Size: 6089 bytes --]

Re: Plans for BMC i2c to host bridge via IPMI

[Anton D. Kachalov]

[-- Attachment #1: Type: text/html, Size: 6803 bytes --]

Re: Plans for BMC i2c to host bridge via IPMI

[Peter Hanson]

Anton,

Thank you for the interesting contributions. As noted earlier, the
Master send-receive message is inadequate to describe the private bus
array in a modern Bmc. And has other senseless limitations. But using
another message seems like a small difference. So I have been studying
your design for its key ideas.

I have some basic high-level questions and/or guesses.

A higher level entity sends messages to one `adapter', which then
sends them to i2c. More-or-less separately, a slave adapter receives
replies and converts them back to messages to that entity.

I now think ipmid can act as that entity. If so, it would be useful to
understand completely how that works.

I see what looks like usable slave I2C adapter code. Is it for
multimaster I2C bus? Or do infernal messaging flows also use it?

Thanks again,
 -- peterh