[edk2-devel] [edk2-platforms][PATCH V1 0/3] Add FW Boot Media Device Indicator
Kubacki, Michael A
michael.a.kubacki at intel.com
Thu Oct 3 07:54:37 UTC 2019
Thanks for sharing. You articulated well why I also don't think this is required in MdePkg. My
description of the reset vector problem was kind of incomplete. While the x86 reset vector
is fixed at the top of the address space, that does not mean that a FV has to be mapped to it
(or any SPI flash based content). SEC has made that assumption. For example, the top
address space could be mapped to another random access memory with x86 instructions.
In this case, assumptions previously made in SecCore such as calculating the BFV size as follows are invalid:
SecCoreData.BootFirmwareVolumeSize = (UINTN)(0x100000000ULL - (UINTN) BootFirmwareVolume);
In a system with NAND media, the BFV could be located in CAR at a base of 4GB-xKB/MB with
execution entry via a jump from the code at top of memory. In this particular case, the FvLength
field can more reliably determine the FV size.
These patches use a GUIDed HOB. It just hides that detail behind a function API largely added
for convenience so firmware boot media consumers do not need to handle the HOB related logic.
Thanks,
Michael
From: afish at apple.com <afish at apple.com>
Sent: Wednesday, October 2, 2019 9:20 PM
To: devel at edk2.groups.io; Kubacki, Michael A <michael.a.kubacki at intel.com>
Cc: Chaganty, Rangasai V <rangasai.v.chaganty at intel.com>; Dong, Eric <eric.dong at intel.com>; Gao, Liming <liming.gao at intel.com>; Ni, Ray <ray.ni at intel.com>
Subject: Re: [edk2-devel] [edk2-platforms][PATCH V1 0/3] Add FW Boot Media Device Indicator
Since I was around back in the Intel Tiano days and I've worked on all the PI specs I can share the history.
The reset vector is a hardware thing. It is usually at the top or bottom of the address space. For x86 it is at the TOP of the ROM and that is why the FV has a VoluteTop file GUID that places the file at the end of the FV, thus the end of that file is the reset vector. If the reset vector is low then the FV header starts with a 16 byte ZeroVector that can contain the reset jump instruction. The other architecture that is out there is a mask ROM, and the mask ROM is the reset vector, and that code hands off to the 2nd level, and sometimes that can be in DRAM or ROM.
The PI FVs (Firmware Volumes) are abstracted by Firmware Volume Blocks (FVB) the FVB instances abstract the storage media for the firmware. Thus from a PI spec point of view there is not need to define the storage type of the "ROM". That is just an implementation detail, that needs to be managed by the implementation.
If you think about in UEFI disk are abstracted by the Block IO protocol. We can boot an OS from any Block device, and the firmware does not need to know what kind of disk it is. The purpose of the Block IO protocol is to abstract all possible implementations o block devices. For the "Boot ROM", FVB is the same kind of abstraction. So you can implement PI with out knowing the media type. Managing the media type was is a platform abstraction, thus there is no need for a boot media abstraction in the MdePkg.
That being said the PI architecture is abstract to a fault. It was designed to abstract all possible future platforms. If there are a family of platforms that share common properties it makes sense to build a platform abstraction package that a set of platforms can share. This is the intent of the PI architecture and the EDKII source base.
The MdePkg implements the UEFI and PI specs, and other industry standards, with some common libs thrown in. So the MdePkg is not the place to put some implementation hint about the Media Device. You could use a GUIDed HOB for that if needed?
Thanks,
Andrew Fish
On Oct 2, 2019, at 8:02 PM, Kubacki, Michael A <michael.a.kubacki at intel.com<mailto:michael.a.kubacki at intel.com>> wrote:
In platforms built for boot media other than SPI flash there has been a compelling
need for silicon and platform code to be aware of the firmware boot media but
apart from the UEFI variable driver (which is a special case being addressed
here - https://github.com/makubacki/edk2/tree/storage_agnostic_uefi_variables),
this has not been the case for edk2 repository packages. In some cases, code in SEC
has made assumptions about the reset vector or FIT pointer that do not necessarily
translate to storage media that does not support MMIO. These cases have been
handled more gracefully than checking the firmware boot media technology. This is
just an observation, not necessarily a case for it to stay in IntelSiliconPkg (which
does make it accessible to Intel silicon and platform code).
I suppose the firmware boot media properties could be treated in a similar manner
to Boot Mode as defined in the Platform Initialization spec. If this was done, it may
make more sense to abstract the technology impact onto firmware, for example,
whether it supports MMIO or not (NOR vs NAND flash) instead of what is defined
here with specific technologies such as eMMC and UFS. Otherwise, the specification
describing this would be subject to constant expansion over time keeping pace with
new technologies and cross-generation code (not silicon or platform specific drivers)
may base conditionals on something like eMMC when its algorithm really applies to
all NAND media (which, again, has been the proven observation thus far outside
silicon and platform code).
With that said, I see the firmware boot technology details being more pertinent to
silicon and platform code so that's why this change is made now. It can immediately
address existing needs for these details in silicon and platform code. Some form of
the firmware boot media details in a more generic package could be useful but it
will likely not align closely with the scope of information needed at this level and is
an undertaking, that in my opinion, is separate but compatible with the work done here.
-----Original Message-----
From: Chaganty, Rangasai V <rangasai.v.chaganty at intel.com<mailto:rangasai.v.chaganty at intel.com>>
Sent: Wednesday, October 2, 2019 4:03 PM
To: Kubacki, Michael A <michael.a.kubacki at intel.com<mailto:michael.a.kubacki at intel.com>>;
devel at edk2.groups.io<mailto:devel at edk2.groups.io>
Cc: Dong, Eric <eric.dong at intel.com<mailto:eric.dong at intel.com>>; Gao, Liming <liming.gao at intel.com<mailto:liming.gao at intel.com>>;
Ni, Ray <ray.ni at intel.com<mailto:ray.ni at intel.com>>
Subject: RE: [edk2-platforms][PATCH V1 0/3] Add FW Boot Media Device
Indicator
I am not sure if there is a silicon scope around the FirmwareBootMediaLib.
Have we considered adding this interface to MdePkg, instead?
-----Original Message-----
From: Kubacki, Michael A
Sent: Monday, September 30, 2019 6:16 PM
To: devel at edk2.groups.io<mailto:devel at edk2.groups.io>
Cc: Chaganty, Rangasai V <rangasai.v.chaganty at intel.com<mailto:rangasai.v.chaganty at intel.com>>; Dong, Eric
<eric.dong at intel.com<mailto:eric.dong at intel.com>>; Gao, Liming <liming.gao at intel.com<mailto:liming.gao at intel.com>>; Ni, Ray
<ray.ni at intel.com<mailto:ray.ni at intel.com>>
Subject: [edk2-platforms][PATCH V1 0/3] Add FW Boot Media Device
Indicator
This patch series introduces a mechanism for determining the firmware boot
media device. This allows the firmware boot media to be discovered through
a standardized API.
Traditionally, most systems have only supported firmware storage on SPI
flash. Presently, several other storage technologies are being used to store
boot system firmware such as eMMC, UFS, and NVMe.
The API for all board, platform, and silicon code to consume the firmware
boot media device is provided by the FirmwareBootMediaLib in
IntelSiliconPkg.
A driver (FirmwareBootMediaInfoPei) is added to BoardModulePkg to serve
as a consistent location for reporting the firmware boot device information.
In order to abstract the potentially hardware-specific details to determine
the boot media (for platforms that support multiple firmware boot media
devices), the driver retrieves the boot media information using a new library
class introduced called FirmwareBootMediaInfoLib. A default instance of this
library class is provided in BoardModulePkg that always returns SPI flash. This
is intended to serve as a default implementation of the library for the most
common scenario and to easily allow a board package to substitute the logic
required to determine the boot media in more complex scenarios.
Ultimately, FirmwareBootMediaInfoPei produces a HOB containing the
firmware boot media device information so it can be used in the HOB
consumer phase.
Cc: Sai Chaganty <rangasai.v.chaganty at intel.com<mailto:rangasai.v.chaganty at intel.com>>
Cc: Eric Dong <eric.dong at intel.com<mailto:eric.dong at intel.com>>
Cc: Liming Gao <liming.gao at intel.com<mailto:liming.gao at intel.com>>
Cc: Ray Ni <ray.ni at intel.com<mailto:ray.ni at intel.com>>
Signed-off-by: Michael Kubacki <michael.a.kubacki at intel.com<mailto:michael.a.kubacki at intel.com>>
Michael Kubacki (3):
IntelSiliconPkg/FirmwareBootMediaLib: Add library
BoardModulePkg/FirmwareBootMediaInfoLib: Add library
BoardModulePkg/FirmwareBootMediaInfoPei: Add module
Platform/Intel/BoardModulePkg/BoardModulePkg.dec
| 3 +
Silicon/Intel/IntelSiliconPkg/IntelSiliconPkg.dec | 4
+-
Platform/Intel/BoardModulePkg/BoardModulePkg.dsc
| 5 +
Silicon/Intel/IntelSiliconPkg/IntelSiliconPkg.dsc | 4
+-
Platform/Intel/BoardModulePkg/FirmwareBootMediaInfo/FirmwareBootMe
diaInfoPei.inf | 46 +++++++++
Platform/Intel/BoardModulePkg/Library/PeiFirmwareBootMediaInfoLib/Pei
FirmwareBootMediaInfoLib.inf | 35 +++++++
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/DxeSmmFirm
wareBootMediaLib.inf | 43 ++++++++
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/PeiFirmwareB
ootMediaLib.inf | 38 +++++++
Platform/Intel/BoardModulePkg/Include/Library/FirmwareBootMediaInfoLi
b.h | 26 +++++
Silicon/Intel/IntelSiliconPkg/Include/Library/FirmwareBootMediaLib.h
| 106 +++++++++++++++++++
Platform/Intel/BoardModulePkg/FirmwareBootMediaInfo/FirmwareBootMe
diaInfoPei.c | 76 ++++++++++++++
Platform/Intel/BoardModulePkg/Library/PeiFirmwareBootMediaInfoLib/Pei
FirmwareBootMediaInfoLib.c | 24 +++++
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/DxeSmmFirm
wareBootMediaLib.c | 107 +++++++++++++++++++
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/FirmwareBoo
tMediaLib.c | 109 ++++++++++++++++++++
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/PeiFirmwareB
ootMediaLib.c | 82 +++++++++++++++
15 files changed, 706 insertions(+), 2 deletions(-) create mode 100644
Platform/Intel/BoardModulePkg/FirmwareBootMediaInfo/FirmwareBootMe
diaInfoPei.inf
create mode 100644
Platform/Intel/BoardModulePkg/Library/PeiFirmwareBootMediaInfoLib/Pei
FirmwareBootMediaInfoLib.inf
create mode 100644
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/DxeSmmFirm
wareBootMediaLib.inf
create mode 100644
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/PeiFirmwareB
ootMediaLib.inf
create mode 100644
Platform/Intel/BoardModulePkg/Include/Library/FirmwareBootMediaInfoLi
b.h
create mode 100644
Silicon/Intel/IntelSiliconPkg/Include/Library/FirmwareBootMediaLib.h
create mode 100644
Platform/Intel/BoardModulePkg/FirmwareBootMediaInfo/FirmwareBootMe
diaInfoPei.c
create mode 100644
Platform/Intel/BoardModulePkg/Library/PeiFirmwareBootMediaInfoLib/Pei
FirmwareBootMediaInfoLib.c
create mode 100644
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/DxeSmmFirm
wareBootMediaLib.c
create mode 100644
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/FirmwareBoo
tMediaLib.c
create mode 100644
Silicon/Intel/IntelSiliconPkg/Library/PeiDxeSmmBootMediaLib/PeiFirmwareB
ootMediaLib.c
--
2.16.2.windows.1
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#48412): https://edk2.groups.io/g/devel/message/48412
Mute This Topic: https://groups.io/mt/34352432/1813853
Group Owner: devel+owner at edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [edk2-devel-archive at redhat.com]
-=-=-=-=-=-=-=-=-=-=-=-
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/edk2-devel-archive/attachments/20191003/7f5f776e/attachment.htm>
More information about the edk2-devel-archive
mailing list