[edk2-devel] [edk2-discuss] PLDM messages via MCTP over KCS
Konstantin Aladyshev
aladyshev22 at gmail.com
Wed Oct 4 15:58:58 UTC 2023
Hi Chang!
Thanks!
There were a couple of trivial compilation errors, but after I've
fixed them everything seems to work fine!
Just in case I've tested the OpenBMC side with the mctp Linux kernel
driver approach
(https://github.com/Kostr/PLDM/tree/master/mctp-kernel)
The latest kernel patches can be found here:
https://lore.kernel.org/lkml/20231003131505.337-1-aladyshev22@gmail.com/
Here is a fix for the build errors that I've found:
```
diff --git a/Features/ManageabilityPkg/Universal/MctpProtocol/Common/MctpProtocolCommon.c
b/Features/ManageabilityPkg/Universal/MctpProtocol/Common/MctpProtocolCommon.c
index 79501d27aa..345c6da81a 100644
--- a/Features/ManageabilityPkg/Universal/MctpProtocol/Common/MctpProtocolCommon.c
+++ b/Features/ManageabilityPkg/Universal/MctpProtocol/Common/MctpProtocolCommon.c
@@ -193,7 +193,7 @@ SetupMctpRequestTransportPacket (
//
// Generate PEC follow SMBUS 2.0 specification.
- *MctpKcsTrailer->Pec = HelperManageabilityGenerateCrc8
(MCTP_KCS_PACKET_ERROR_CODE_POLY, 0, ThisPackage,
MctpKcsHeader->ByteCount);
+ MctpKcsTrailer->Pec = HelperManageabilityGenerateCrc8
(MCTP_KCS_PACKET_ERROR_CODE_POLY, 0, ThisPackage,
MctpKcsHeader->ByteCount);
*PacketBody = (UINT8 *)ThisPackage;
*PacketBodySize = MctpKcsHeader->ByteCount;
*PacketTrailer = (MANAGEABILITY_TRANSPORT_TRAILER)MctpKcsTrailer;
diff --git a/Features/ManageabilityPkg/Universal/MctpProtocol/Dxe/MctpProtocol.c
b/Features/ManageabilityPkg/Universal/MctpProtocol/Dxe/MctpProtocol.c
index 863b8d471c..247d032b9b 100644
--- a/Features/ManageabilityPkg/Universal/MctpProtocol/Dxe/MctpProtocol.c
+++ b/Features/ManageabilityPkg/Universal/MctpProtocol/Dxe/MctpProtocol.c
@@ -79,17 +79,17 @@ MctpSubmitMessage (
}
//
- // Chec source EID and destination EDI.
+ // Check source EID and destination EID
//
if ((MctpSourceEndpointId >= MCTP_RESERVED_ENDPOINT_START_ID) &&
- MctpSourceEndpointId <= MCTP_RESERVED_ENDPOINT_END_ID)
+ (MctpSourceEndpointId <= MCTP_RESERVED_ENDPOINT_END_ID)
) {
DEBUG ((DEBUG_ERROR, "%a: The value of MCTP source EID (%x) is
reserved.\n", func, MctpSourceEndpointId));
return EFI_INVALID_PARAMETER;
}
if ((MctpDestinationEndpointId >= MCTP_RESERVED_ENDPOINT_START_ID) &&
- MctpDestinationEndpointId <= MCTP_RESERVED_ENDPOINT_END_ID)
+ (MctpDestinationEndpointId <= MCTP_RESERVED_ENDPOINT_END_ID)
) {
DEBUG ((DEBUG_ERROR, "%a: The value of MCTP destination EID (%x)
is reserved.\n", func, MctpDestinationEndpointId));
return EFI_INVALID_PARAMETER;
```
Best regards,
Konstantin Aladyshev
On Wed, Oct 4, 2023 at 2:52 PM Chang, Abner <Abner.Chang at amd.com> wrote:
>
> [AMD Official Use Only - General]
>
> Hi Aladyshev,
> I have updated the change you made and put those code on below link,
> https://github.com/changab/edk2-platforms/commit/1c8d0d3fa403b47a34667f7f690add7822163111
>
> I combined MCTP over KCS changes and IPMI over KCS functionality in KcsCommonLib.c. I also created MANAGEABILITY_MCTP_KCS_TRAILER as you suggested. The source EID and destination EID are checked in MctpSubmitCommand as well. IPMI/KCS functionality is verified and works fine after this change.
> As I am no able to use the corresponding change you made on OpenBMC site at my end, could you please help to verify my updates on your machine? Let's see how it works.
> I also consider to migrate the code that generates MCTP over KCS header/trailer from MctpProtocolCommon.c to KcsCommonLib.c, maybe after we verifying PLDM->MCTP->KCS route works well on ManageabilityPkg.
>
> Thank you
> Abner
>
>
> > -----Original Message-----
> > From: Konstantin Aladyshev <aladyshev22 at gmail.com>
> > Sent: Friday, September 29, 2023 2:18 AM
> > To: Chang, Abner <Abner.Chang at amd.com>
> > Cc: devel at edk2.groups.io; discuss at edk2.groups.io
> > Subject: Re: [edk2-devel] [edk2-discuss] PLDM messages via MCTP over KCS
> >
> > Caution: This message originated from an External Source. Use proper caution
> > when opening attachments, clicking links, or responding.
> >
> >
> > Hi, Chang!
> >
> > Did you have time to test libmctp MCTP KCS binding solution?
> >
> > Here are some updates from my end. As I was saying, I was working on
> > the Linux kernel binding solution.
> > And now I've finished the initial implementation of the Linux kernel
> > binding driver for the MCTP-over-KCS binding and proposed all the
> > patches upstream
> > (https://www.spinics.net/lists/kernel/msg4949173.html).
> > I've also updated instructions in my repo
> > https://github.com/Kostr/PLDM (the guide for the kernel binding
> > solution and all the necessary Linux kernel patches can be found here
> > https://github.com/Kostr/PLDM/tree/master/mctp-kernel).
> > So now you can use Linux driver instead of the libmctp utility on the BMC side.
> >
> > Couple of things that I've noticed in the development process:
> > - `MctpSubmitCommand` receives
> > 'MctpSourceEndpointId'/'MctpDestinationEndpointId' as arguments. But
> > these values aren't actually used. The current code just uses EIDs
> > that were set via PCDs
> > (https://github.com/tianocore/edk2-
> > platforms/blob/d03a60523a6086d200d3eb1e2f25530bf1cb790e/Features/
> > ManageabilityPkg/Universal/MctpProtocol/Common/MctpProtocolCommon.
> > c#L178)
> > - According to the specification DSP0236 (section 8.2) MCTP EID 0 to 7
> > are reserved. It is critical that we do not use them since MCTP Linux
> > kernel subsystem checks that part. So we probably need to add some
> > check to the `MctpSubmitCommand` that would verify that we don't use
> > reserved EIDs.
> >
> > Best regards,
> > Konstantin Aladyshev
> >
> > On Thu, Sep 21, 2023 at 5:32 AM Chang, Abner <Abner.Chang at amd.com>
> > wrote:
> > >
> > > [AMD Official Use Only - General]
> > >
> > > Hi Aladyshev,
> > > Thanks for providing the details, I will take a look at your code first,
> > implement it at my end and then response to your question.
> > >
> > > Abner
> > >
> > > > -----Original Message-----
> > > > From: Konstantin Aladyshev <aladyshev22 at gmail.com>
> > > > Sent: Friday, September 8, 2023 8:57 PM
> > > > To: Chang, Abner <Abner.Chang at amd.com>
> > > > Cc: devel at edk2.groups.io; discuss at edk2.groups.io
> > > > Subject: Re: [edk2-devel] [edk2-discuss] PLDM messages via MCTP over
> > KCS
> > > >
> > > > Caution: This message originated from an External Source. Use proper
> > caution
> > > > when opening attachments, clicking links, or responding.
> > > >
> > > >
> > > > Hi, Chang!
> > > >
> > > > I've finished my initial implementation of the MCTP over KCS binding.
> > > > You can find 'edk2-platform' patches and 'openbmc' patches along with
> > > > all of the instructions in my repository
> > > > https://github.com/Kostr/PLDM. I hope you'll be able to reproduce
> > > > everything on your hardware configuration. Feel free to ask any
> > > > questions.
> > > > Also I've sent all the openbmc patches upstream, hope they will get
> > > > accepted soon.
> > > > As for the 'edk2-platform' patches, right now I don't fully understand
> > > > how to write them correctly to keep IPMI over KCS stack working. I
> > > > think here I would need your help. Right now I've commited them to my
> > > > `edk2-platforms` fork
> > > > https://github.com/tianocore/edk2-
> > > > platforms/commit/99a6c98a63b37f955c0d0480149b84fcc3a03f74
> > > >
> > > > Couple of questions/notices:
> > > > 1) You've said that we can differentiate MCTP by the transfer token,
> > > > but it is not passed to the 'KcsTransportSendCommand' function
> > > > https://github.com/tianocore/edk2-
> > > >
> > platforms/blob/bb6841e3fd1c60b3f8510b4fc0a380784e05d326/Features/
> > > >
> > ManageabilityPkg/Library/ManageabilityTransportKcsLib/Common/KcsComm
> > > > on.c#L414
> > > >
> > > > 2) What function should know about the
> > > > MANAGEABILITY_MCTP_KCS_HEADER?
> > > > Keep in mind that this header includes 'ByteCount' for the incoming
> > > > data size that we need to read.
> > > > - KcsTransportSendCommand or CommonMctpSubmitMessage ?
> > > > (https://github.com/tianocore/edk2-
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Library/ManageabilityTra
> > > > nsportKcsLib/Common/KcsCommon.c)
> > > > (https://github.com/tianocore/edk2-
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Universal/MctpProtocol/
> > > > Common/MctpProtocolCommon.c)?
> > > >
> > > > 3) As I've said earlier I think it would be good to add
> > > > MANAGEABILITY_MCTP_KCS_TRAILER to the Mctp.h
> > > >
> > > > 4) Not sure if it is a good idea to pass these parameters to the
> > > > MctpSubmitCommand protocol function:
> > > > ```
> > > > UINT8 MctpType,
> > > > BOOLEAN RequestDataIntegrityCheck,
> > > > ```
> > > > (https://github.com/tianocore/edk2-
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Include/Protocol/MctpPr
> > > > otocol.h)
> > > > Shouldn't it be in the `RequestData` directly since it is more of a
> > > > payload than a header for the MCTP? I don't know the specification
> > > > very well, but what if the RequestDataIntegrityCheck would be set in
> > > > the response? Who would need to check the integrity of the payload
> > > > buffer in that case? MCTP library or the code calling the MCTP
> > > > library?
> > > >
> > > > 5) Haven't tested the PldmProtocol, maybe it also needs some corrections.
> > > >
> > > > Feel free to ask any questions about my solution.
> > > >
> > > > Right now I'll probably focus on the Linux kernel driver for the MCTP
> > > > over KCS binding. So if you want to finish edk2-platforms code based
> > > > on my patches, feel free to do it. If not, I'll try to get back to it
> > > > after I finish the Linux kernel driver.
> > > >
> > > > Best regards,
> > > > Konstantin Aladyshev
> > > >
> > > > On Fri, Sep 1, 2023 at 8:58 AM Chang, Abner <Abner.Chang at amd.com>
> > > > wrote:
> > > > >
> > > > > [AMD Official Use Only - General]
> > > > >
> > > > > See my answer below,
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: devel at edk2.groups.io <devel at edk2.groups.io> On Behalf Of
> > > > > > Konstantin Aladyshev via groups.io
> > > > > > Sent: Friday, September 1, 2023 12:02 AM
> > > > > > To: Chang, Abner <Abner.Chang at amd.com>
> > > > > > Cc: devel at edk2.groups.io; discuss at edk2.groups.io
> > > > > > Subject: Re: [edk2-devel] [edk2-discuss] PLDM messages via MCTP
> > over
> > > > KCS
> > > > > >
> > > > > > Caution: This message originated from an External Source. Use proper
> > > > caution
> > > > > > when opening attachments, clicking links, or responding.
> > > > > >
> > > > > >
> > > > > > As I've said there is nothing like "KCS completion code" in the MCTP
> > > > > > over KCS binding specification
> > > > > >
> > > >
> > (https://www.dmtf.org/sites/default/files/standards/documents/DSP0254_1
> > > > > > .0.0.pdf).
> > > > > > The response packet should have the same structure as a request. I.e.
> > > > > > a packet with MANAGEABILITY_MCTP_KCS_HEADER and PEC.
> > > > > >
> > > > > > Currently I'm writing "MCTP over KCS" binding for the OpenBMC
> > libmctp
> > > > > > project. So I can send whatever I want, I don't think my output would
> > > > > > be any useful to you. But I've asked this question in the community
> > > > > > and they also confirmed that the response packet has the same
> > > > > > structure.
> > > > > >
> > > >
> > (https://discord.com/channels/775381525260664832/7787906385638850
> > > > > > 86/1146782595334549554)
> > > > > >
> > > > > > Now I'm a little bit confused about the `KcsTransportSendCommand`
> > > > > > function. It has the following arguments for the function output:
> > > > > > ```
> > > > > > OUT UINT8 *ResponseData OPTIONAL,
> > > > > > IN OUT UINT32 *ResponseDataSize OPTIONAL
> > > > > > ```
> > > > > > Should we include
> > MCTP_TRANSPORT_HEADER/MCTP_MESSAGE_HEADER
> > > > to
> > > > > > this
> > > > > > output or not?
> > > > > If the MCTP KCS packet for host->BMC and BMC->host are in the same
> > > > structure, then yes, the response data from BMC should includes
> > > > MCTP_TRANSPORT_HEADER/MCTP_MESSAGE_HEADER in my opinion, as
> > this
> > > > is defined in MCTP base protocol.
> > > > >
> > > > > So let me explain the implementation for MCTP over KCS and what do we
> > > > miss now.
> > > > >
> > > > > A. MCTP protocol driver linked with KCS Transport interface library
> > > > > - In MCTP protocol driver, if the transport interface is KCS then
> > > > > 1. MCTP protocol driver builds up the MCTP KCS transport header,
> > which
> > > > is DefBody, NetFunc and ByeCount
> > > > > 2. MCTP protocol driver builds up the MCTP payload
> > > > > 3. MCTP protocol driver builds up the MCTP KCS transport trailer,
> > which
> > > > is PEC.
> > > > > Above three steps are already implemented.
> > > > > PEC is calculated by MCTP protocol driver and should be verified by
> > KCS
> > > > using the same algorithm in my understanding of spec.
> > > > >
> > > > > B. In KCS Transport interface library
> > > > > 1. KCS Transport interface library sends the transport header got from
> > > > TransportToken. Same behavior for IPMI protocol, but different content.
> > KCS
> > > > Transport interface library doesn't have to understand this.
> > > > > 2. KCS Transport interface library sends the payload
> > > > > 3. KCS Transport interface library sends the transport trailer got from
> > > > TransportToken. Same behavior for IPMI protocol, but different content.
> > KCS
> > > > Transport interface library doesn't have to understand this.
> > > > > Above three steps are already implemented.
> > > > >
> > > > > Then, if Manageability protocol is MCTP, we skip reading responses
> > > > header (Not implemented)
> > > > > For reading response data
> > > > > 1. If the ResponseData and ResponseSize is valid in the given
> > > > TransportToken, then we read ResponseSize data from KCS. (Already
> > > > implemented)
> > > > > 2. if Manageability protocol is MCTP, then we skip reading responses
> > > > header again (Not implemented)
> > > > > Now the response is returned to MCTP protocol driver
> > > > >
> > > > > C. In MCTP protocol driver, we expect to get the whole MCTP over KCS
> > > > packet response, that includes DefBody, NetFunc and ByeCount, MCTP
> > > > message and PEC.
> > > > > 1. MCTP protocol driver verifies the returned PEC with the payload.
> > > > > 2. Strip out DefBody, NetFunc, ByeCount and PEC and then returns it
> > to
> > > > upper layer (e.g., MCTP transport interface library). Returns only MCTP
> > > > Transport header and MCTP packet payload as it shows in MCTP base
> > protocol
> > > > spec.
> > > > > Above is not implemented
> > > > >
> > > > > D. In MCTP transport interface library, we can strip out MCTP transport
> > > > header and then return it to upper layer (e.g., PLDM protocol driver).
> > > > > Above is not implemented.
> > > > >
> > > > > E. In PLDM protocol driver,
> > > > > 1. we verify the Message Integrity Check if the Message Type
> > requests it.
> > > > > 2. we can remove MCTP message type then return it to upper layer
> > (e.g.,
> > > > PLDM SMBIOS transfer)
> > > > > Above is not implemented.
> > > > >
> > > > > We didn’t implement BMC->Host in step C, D and E as our current
> > demand is
> > > > to send the SMBIOS table to BMC, which doesn't require the response data
> > if I
> > > > am not wrong.
> > > > > Let me know if it is problematic in the above process.
> > > > >
> > > > > Thanks and regards,
> > > > > Abner
> > > > >
> > > > >
> > > > >
> > > > > >
> > > > > > Best regards,
> > > > > > Konstantin Aladyshev
> > > > > >
> > > > > > On Thu, Aug 31, 2023 at 6:52 PM Chang, Abner
> > > > <Abner.Chang at amd.com>
> > > > > > wrote:
> > > > > > >
> > > > > > > [AMD Official Use Only - General]
> > > > > > >
> > > > > > > But wait, wee my another comment below,
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: Chang, Abner
> > > > > > > > Sent: Thursday, August 31, 2023 11:42 PM
> > > > > > > > To: devel at edk2.groups.io; aladyshev22 at gmail.com
> > > > > > > > Cc: discuss at edk2.groups.io
> > > > > > > > Subject: RE: [edk2-devel] [edk2-discuss] PLDM messages via MCTP
> > > > over
> > > > > > KCS
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > > -----Original Message-----
> > > > > > > > > From: devel at edk2.groups.io <devel at edk2.groups.io> On Behalf
> > Of
> > > > > > > > > Konstantin Aladyshev via groups.io
> > > > > > > > > Sent: Thursday, August 31, 2023 10:57 PM
> > > > > > > > > To: Chang, Abner <Abner.Chang at amd.com>
> > > > > > > > > Cc: discuss at edk2.groups.io; devel at edk2.groups.io
> > > > > > > > > Subject: Re: [edk2-devel] [edk2-discuss] PLDM messages via
> > MCTP
> > > > over
> > > > > > KCS
> > > > > > > > >
> > > > > > > > > Caution: This message originated from an External Source. Use
> > proper
> > > > > > > > caution
> > > > > > > > > when opening attachments, clicking links, or responding.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > > (I don see what is the response header for MCTP KCS in spec
> > though,
> > > > > > does
> > > > > > > > it
> > > > > > > > > mention the KCS response?).
> > > > > > > > >
> > > > > > > > > The spec doesn't explicitly mention that the format of a send and
> > > > > > > > > response packets differ. So I assume it is the same and it is
> > > > > > > > > described at the "Figure 1 – MCTP over KCS Packet Format"
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > (https://www.dmtf.org/sites/default/files/standards/documents/DSP0254_1
> > > > > > > > > .0.0.pdf)
> > > > > > > > > Therefore the format of a response would look like this:
> > > > > > > > > ```
> > > > > > > > > MANAGEABILITY_MCTP_KCS_HEADER
> > > > > > > > > (https://github.com/tianocore/edk2-
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Include/Library/Managea
> > > > > > > > > bilityTransportMctpLib.h)
> > > > > > > > > MCTP_TRANSPORT_HEADER
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > (https://github.com/tianocore/edk2/blob/master/MdePkg/Include/Industry
> > > > > > > > > Standard/Mctp.h)
> > > > > > > > > MCTP_MESSAGE_HEADER
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > (https://github.com/tianocore/edk2/blob/master/MdePkg/Include/Industry
> > > > > > > > > Standard/Mctp.h)
> > > > > > > > > < response data>
> > > > > > > > What do you see the KCS response from BMC? You probably right as
> > > > the
> > > > > > > > header and trailer are labeled in different colors 😊. Could you
> > please
> > > > > > enable
> > > > > > > > the debug message to capture it?
> > > > > > > >
> > > > > > > > > PEC
> > > > > > > > > (Probably we need to define
> > > > > > MANAGEABILITY_MCTP_KCS_TRAILER)
> > > > > > > > > ```
> > > > > > > > We have MANAGEABILITY_MCTP_KCS_HEADER defined but no
> > > > > > > > MANAGEABILITY_MCTP_KCS_TRAILER as it is hardcoded to one
> > byte.
> > > > > > > > If the KCS response is PEC, then yes, we can create
> > > > > > > > MANAGEABILITY_MCTP_KCS_TRAILER for KCS transport interface.
> > > > > > >
> > > > > > > In the implementation, PEC is calculated in MCTP protocol and send
> > > > through
> > > > > > KCS as the KCS packet trailer. So, when we send the MCTP request
> > through
> > > > > > KCS, KCS shouldn't respond the PEC to upper stack, right? I still think
> > the
> > > > > > response should be the KCS completion code. The debug message from
> > > > your
> > > > > > end may help to clarify this as your BMC has the MCTP KCS
> > > > implementation.
> > > > > > >
> > > > > > > Thanks
> > > > > > > Abner
> > > > > > >
> > > > > > > >
> > > > > > > > >
> > > > > > > > > So in the "KcsTransportSendCommand"
> > > > > > > > > (https://github.com/tianocore/edk2-
> > > > > > > > >
> > > > > >
> > > >
> > platforms/blob/14553d31c72afa7289f6a2555b6e91f4f715a05a/Features/
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > ManageabilityPkg/Library/ManageabilityTransportKcsLib/Common/KcsComm
> > > > > > > > > on.c#L414)
> > > > > > > > > we can check if we transfer is MCTP (based on
> > > > > > > > > "TransportToken->ManagebilityProtocolSpecification == MCTP" like
> > > > > > > > > you've suggested) and handle response accordingly.
> > > > > > > > >
> > > > > > > > > But which headers should we check in this function? Only
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > MANAGEABILITY_MCTP_KCS_HEADER/MANAGEABILITY_MCTP_KCS_TRAILER
> > > > > > > > > ?
> > > > > > > > Yes, only check header and trailer for transport interface.
> > > > > > > >
> > > > > > > > > What about
> > > > MCTP_TRANSPORT_HEADER/MCTP_MESSAGE_HEADER?
> > > > > > Do
> > > > > > > > we
> > > > > > > > > need to
> > > > > > > > > check them here as well? Or do we need to check them
> > somewhere
> > > > > > upper
> > > > > > > > > the call stack?
> > > > > > > > We should leave this to MCTP protocol driver as this is belong to
> > > > protocol
> > > > > > > > layer, the upper layer stack.
> > > > > > > >
> > > > > > > > Thanks
> > > > > > > > Abner
> > > > > > > >
> > > > > > > > >
> > > > > > > > > Best regards,
> > > > > > > > > Konstantin Aladyshev
> > > > > > > > >
> > > > > > > > > On Thu, Aug 31, 2023 at 7:59 AM Chang, Abner
> > > > > > <Abner.Chang at amd.com>
> > > > > > > > > wrote:
> > > > > > > > > >
> > > > > > > > > > [AMD Official Use Only - General]
> > > > > > > > > >
> > > > > > > > > > Hi Aladyshev,
> > > > > > > > > >
> > > > > > > > > > > -----Original Message-----
> > > > > > > > > > > From: Konstantin Aladyshev <aladyshev22 at gmail.com>
> > > > > > > > > > > Sent: Wednesday, August 30, 2023 11:09 PM
> > > > > > > > > > > To: Chang, Abner <Abner.Chang at amd.com>
> > > > > > > > > > > Cc: discuss at edk2.groups.io; devel at edk2.groups.io
> > > > > > > > > > > Subject: Re: [edk2-discuss] PLDM messages via MCTP over KCS
> > > > > > > > > > >
> > > > > > > > > > > Caution: This message originated from an External Source. Use
> > > > proper
> > > > > > > > > caution
> > > > > > > > > > > when opening attachments, clicking links, or responding.
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > Hi!
> > > > > > > > > > >
> > > > > > > > > > > I've started to implement MCTP over KCS binding for the
> > libmctp
> > > > > > > > > > > (https://github.com/openbmc/libmctp) and test it with the
> > > > current
> > > > > > code
> > > > > > > > > > > in the ManageabilityPkg.
> > > > > > > > > > >
> > > > > > > > > > > I was able successfully send the MCTP packet to the BMC, but
> > right
> > > > > > now
> > > > > > > > > > > I'm having some troubles with receiving the answer back.
> > > > > > > > > > >
> > > > > > > > > > > I think I've found some bug in the
> > `KcsTransportSendCommand`
> > > > code.
> > > > > > > > > > >
> > > > > > > > > > > After it sends data over KCS in expects a responce starting with
> > a
> > > > > > > > > > > 'IPMI_KCS_RESPONSE_HEADER'
> > > > > > > > > > > https://github.com/tianocore/edk2-
> > > > > > > > > > >
> > > > > > > > >
> > > > > >
> > > >
> > platforms/blob/14553d31c72afa7289f6a2555b6e91f4f715a05a/Features/
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > ManageabilityPkg/Library/ManageabilityTransportKcsLib/Common/KcsComm
> > > > > > > > > > > on.c#L476
> > > > > > > > > > >
> > > > > > > > > > > Isn't it wrong, assuming that the right header in case of MCTP
> > > > should
> > > > > > > > > > > be 'MANAGEABILITY_MCTP_KCS_HEADER' ?
> > > > > > > > > > >
> > > > > > > > > > > I guess the 'IpmiHelperCheckCompletionCode' check after the
> > data
> > > > > > > > > > > receive is also not relevant for the MCTP.
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > This is something I don’t really sure as I can't verify the response
> > > > > > payload
> > > > > > > > > because our BMC doesn't have the code to handle MCTP over KCS
> > > > > > > > command.
> > > > > > > > > However it is appreciated if community can help to verify this. As I
> > can
> > > > > > > > > remember, I can see the return KCS status is 0xC1, the invalid
> > > > command.
> > > > > > > > Thus I
> > > > > > > > > think if we do a MCTP over KCS, the first response is still KCS
> > response
> > > > > > > > header.
> > > > > > > > > > This is not what do you see on the BCM it does support MCTP
> > over
> > > > > > KCS? If
> > > > > > > > > so, then I would like to have your help to correct this code.
> > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > Since 'ManageabilityTransportKcsLib' can be used both for IPMI
> > > > and
> > > > > > > > > > > MCTP, how should we deal with this?
> > > > > > > > > > If KcsCommon.c, we can have different code path for the given
> > > > protocol
> > > > > > > > > GUID. e.g., if (TransportToken->ManagebilityProtocolSpecification
> > ==
> > > > > > MCTP).
> > > > > > > > > > Then skip reading the KCS_REPOSNSE_HEADER or to read the
> > > > > > > > > MCTP_RESPONSE_HEADER (I don see what is the response header
> > for
> > > > > > MCTP
> > > > > > > > > KCS in spec though, does it mention the KCS response?).
> > > > > > > > > >
> > > > > > > > > > Thanks
> > > > > > > > > > Abner
> > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > Best regards,
> > > > > > > > > > > Konstantin Aladyshev
> > > > > > > > > > >
> > > > > > > > > > > On Wed, Aug 23, 2023 at 5:18 AM Chang, Abner
> > > > > > > > > <Abner.Chang at amd.com>
> > > > > > > > > > > wrote:
> > > > > > > > > > > >
> > > > > > > > > > > > [AMD Official Use Only - General]
> > > > > > > > > > > >
> > > > > > > > > > > > Please see my answers inline.
> > > > > > > > > > > >
> > > > > > > > > > > > > -----Original Message-----
> > > > > > > > > > > > > From: discuss at edk2.groups.io <discuss at edk2.groups.io>
> > On
> > > > > > Behalf
> > > > > > > > Of
> > > > > > > > > > > > > Konstantin Aladyshev via groups.io
> > > > > > > > > > > > > Sent: Wednesday, August 23, 2023 1:54 AM
> > > > > > > > > > > > > To: Chang, Abner <Abner.Chang at amd.com>
> > > > > > > > > > > > > Cc: discuss at edk2.groups.io; devel at edk2.groups.io
> > > > > > > > > > > > > Subject: Re: [edk2-discuss] PLDM messages via MCTP over
> > KCS
> > > > > > > > > > > > >
> > > > > > > > > > > > > Caution: This message originated from an External Source.
> > Use
> > > > > > proper
> > > > > > > > > > > caution
> > > > > > > > > > > > > when opening attachments, clicking links, or responding.
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > > Thanks for the answer!
> > > > > > > > > > > > >
> > > > > > > > > > > > > I was a little bit confused about the part, that in the same
> > > > package
> > > > > > I
> > > > > > > > > > > > > actually need to provide different library implementations
> > for
> > > > the
> > > > > > > > > > > > > same 'ManageabilityTransportLib', thanks for the
> > clarification!
> > > > > > > > > > > > > I think your DSC example should go into the package
> > > > > > documentation.
> > > > > > > > > > > > Yes, this is a good idea. I will update it.
> > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > > As for me, I'm working with the OpenBMC distribution
> > > > > > > > > > > > > (https://github.com/openbmc/openbmc) and my goal is to
> > > > > > transfer
> > > > > > > > > data
> > > > > > > > > > > > > from the BIOS to the BMC via MCTP/PLDM.
> > > > > > > > > > > > > Currently there is no solution for the MCTP over KCS binding
> > in
> > > > > > Linux,
> > > > > > > > > > > > > so I need to add this support:
> > > > > > > > > > > > > - either to the MCTP userspace library
> > > > > > > > > > > > > (https://github.com/openbmc/libmctp) [old OpenBMC
> > way,
> > > > but
> > > > > > > > > probably
> > > > > > > > > > > > > easier]
> > > > > > > > > > > > > - or to the MCTP kernel binding
> > > > > > > > > > > > >
> > > > > > (https://github.com/torvalds/linux/tree/master/drivers/net/mctp)
> > > > > > > > > > > > > [modern mctp Linux driver approach]
> > > > > > > > > > > > >
> > > > > > > > > > > > > Both don't sound like an easy task, so can I ask, what MC
> > (i.e.
> > > > > > > > > > > > > management controller) device and firmware do you use on
> > > > the
> > > > > > other
> > > > > > > > > > > > > side of the MCTP KCS transmissions?
> > > > > > > > > > > >
> > > > > > > > > > > > We use OpenBMC as well, but as you mention there are some
> > > > > > missing
> > > > > > > > > pieces
> > > > > > > > > > > to fully support manageability between host and BMC.
> > > > > > > > > > > > We don’t have code to handle MCTP IPMI either, the edk2
> > > > > > > > > ManageabilityPkg
> > > > > > > > > > > provides the framework while MCTP/PLDM/KCS
> > implementation
> > > > > > > > provides
> > > > > > > > > a
> > > > > > > > > > > sample other than IPMI/KCS to prove the flexibility of
> > > > > > ManageabilityPkg.
> > > > > > > > > > > > Actually, MCTP over KCS is not supported in our BMC
> > firmware
> > > > yet,
> > > > > > > > thus
> > > > > > > > > > > BMC just returns the invalid command. However, the transport
> > > > > > > > framework
> > > > > > > > > > > has been verified to make sure the implementation works fine
> > as
> > > > > > expect.
> > > > > > > > > > > > We need help from community to provide more
> > manageability
> > > > > > > > protocols
> > > > > > > > > and
> > > > > > > > > > > transport interface libraries to this package.
> > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > > You've also mentioned PLDM SMBIOS, isn't it covered by
> > the
> > > > > > > > > > > > > https://github.com/tianocore/edk2-
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Universal/PldmSmbiosTr
> > > > > > > > > > > > > ansferDxe/PldmSmbiosTransferDxe.c
> > > > > > > > > > > > > ?
> > > > > > > > > > > > Ah hah, yes I forget I upstream it.
> > > > > > > > > > > >
> > > > > > > > > > > > Please just feel free to send patch to make more
> > functionalities to
> > > > > > this
> > > > > > > > > > > package.
> > > > > > > > > > > > Thanks
> > > > > > > > > > > > Abner
> > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > > Best regards,
> > > > > > > > > > > > > Konstantin Aladyshev
> > > > > > > > > > > > >
> > > > > > > > > > > > > On Tue, Aug 22, 2023 at 7:26 PM Chang, Abner
> > > > > > > > > > > <Abner.Chang at amd.com>
> > > > > > > > > > > > > wrote:
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > [AMD Official Use Only - General]
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Hi Aladyshev,
> > > > > > > > > > > > > > We use library class to specify the desire transport
> > interface
> > > > for
> > > > > > the
> > > > > > > > > > > > > management protocol, such as MCTP, PLDM and IPMI. This
> > > > way
> > > > > > we
> > > > > > > > can
> > > > > > > > > > > > > flexibly support any transport interface for the management
> > > > > > protocol.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Here is the example of using ManageabilityPkg, which is
> > > > PLDM
> > > > > > over
> > > > > > > > > MCTP
> > > > > > > > > > > > > over KCS.
> > > > > > > > > > > > > >
> > > > > > > > ManageabilityPkg/Universal/IpmiProtocol/Dxe/IpmiProtocolDxe.inf
> > > > > > > > > {
> > > > > > > > > > > > > > <LibraryClasses>
> > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > ManageabilityTransportLib|ManageabilityPkg/Library/ManageabilityTranspor
> > > > > > > > > > > > > tKcsLib/Dxe/DxeManageabilityTransportKcs.inf
> > > > > > > > > > > > > > }
> > > > > > > > > > > > > >
> > > > > > > > >
> > ManageabilityPkg/Universal/MctpProtocol/Dxe/MctpProtocolDxe.inf
> > > > {
> > > > > > > > > > > > > > <LibraryClasses>
> > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > ManageabilityTransportLib|ManageabilityPkg/Library/ManageabilityTranspor
> > > > > > > > > > > > > tKcsLib/Dxe/DxeManageabilityTransportKcs.inf
> > > > > > > > > > > > > > }
> > > > > > > > > > > > > >
> > > > > > > > >
> > ManageabilityPkg/Universal/PldmProtocol/Dxe/PldmProtocolDxe.inf
> > > > {
> > > > > > > > > > > > > > <LibraryClasses>
> > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > ManageabilityTransportLib|ManageabilityPkg/Library/ManageabilityTranspor
> > > > > > > > > > > > > tMctpLib/Dxe/DxeManageabilityTransportMctp.inf
> > > > > > > > > > > > > > }
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > So you can implement ManageabilityTransport library for
> > > > either
> > > > > > > > > industry
> > > > > > > > > > > > > standard or proprietary implementation for the specific
> > > > > > management
> > > > > > > > > > > > > protocol.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > BTW, We do have PLDM SMBIOS over MCTP
> > implementation
> > > > but
> > > > > > > > not
> > > > > > > > > > > > > upstream yet.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Hope this information helps.
> > > > > > > > > > > > > > Thanks
> > > > > > > > > > > > > > Abner
> > > > > > > > > > > > > >
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > > -----Original Message-----
> > > > > > > > > > > > > > > From: discuss at edk2.groups.io
> > <discuss at edk2.groups.io>
> > > > On
> > > > > > > > > Behalf Of
> > > > > > > > > > > > > > > Konstantin Aladyshev via groups.io
> > > > > > > > > > > > > > > Sent: Tuesday, August 22, 2023 7:00 PM
> > > > > > > > > > > > > > > To: discuss <discuss at edk2.groups.io>;
> > > > devel at edk2.groups.io
> > > > > > > > > > > > > > > Subject: [edk2-discuss] PLDM messages via MCTP over
> > KCS
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Caution: This message originated from an External
> > Source.
> > > > Use
> > > > > > > > > proper
> > > > > > > > > > > > > caution
> > > > > > > > > > > > > > > when opening attachments, clicking links, or responding.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Hi!
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > I'm trying to build `ManageabilityPkg` from the edk2-
> > > > platforms
> > > > > > > > > > > > > > > repo to issue PLDM messages via MCTP over KCS. Is it
> > > > possible
> > > > > > > > with
> > > > > > > > > > > > > > > the current code? I see all the building blocks, but have
> > > > trouble
> > > > > > > > > > > > > > > putting it all together.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > The main question that bothers me is what
> > implementation
> > > > > > should
> > > > > > > > I
> > > > > > > > > set
> > > > > > > > > > > > > > > for the `ManageabilityTransportLib`?
> > > > > > > > > > > > > > > By default it is set to dummy
> > > > > > `BaseManageabilityTransportNull.inf`
> > > > > > > > > > > > > > > (https://github.com/tianocore/edk2-
> > > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > >
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/ManageabilityPkg.dsc).
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > On one case to get PLDM via MCTP it looks that I need to
> > set
> > > > it
> > > > > > to
> > > > > > > > > > > > > > > `DxeManageabilityTransportMctp.inf`
> > > > > > > > > > > > > > > ManageabilityTransportLib|
> > > > > > > > > <...>/DxeManageabilityTransportMctp.inf
> > > > > > > > > > > > > > > (https://github.com/tianocore/edk2-
> > > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Library/ManageabilityTra
> > > > > > > > > > > > > > >
> > nsportMctpLib/Dxe/DxeManageabilityTransportMctp.inf)
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > But on the other case if I want MCTP over KCS I need to
> > set
> > > > it to
> > > > > > > > > > > > > > > `DxeManageabilityTransportKcs.inf`
> > > > > > > > > > > > > > > ManageabilityTransportLib|
> > > > > > > > <...>/DxeManageabilityTransportKcs.inf
> > > > > > > > > > > > > > > (https://github.com/tianocore/edk2-
> > > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > >
> > platforms/blob/master/Features/ManageabilityPkg/Library/ManageabilityTra
> > > > > > > > > > > > > > > nsportKcsLib/Dxe/DxeManageabilityTransportKcs.inf)
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > What is the right way to resolve this?
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > There are no platforms in the repo that actually
> > implement
> > > > > > > > > PLDM/MCTP
> > > > > > > > > > > > > > > functionality, so there is no example that I can use as a
> > > > > > reference.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Best regards,
> > > > > > > > > > > > > > > Konstantin Aladyshev
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > > >
> > > > >
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#109325): https://edk2.groups.io/g/devel/message/109325
Mute This Topic: https://groups.io/mt/100897530/1813853
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