NVIDIA EDK2 platform support
License: Other
This repository contains sources necessary to build UEFI for NVIDIA Platforms. UEFI for NVIDIA Platforms is based on EDK2.
See the wiki for more information, including build and flash instructions.
To build this platform, please refer to the instructions.
CapsuleApp uses the QueryCapsuleCapabilities UEFI RT service to detect the maximum supported capsule size which is currently set to 10MB. Current T194 and T234 capsules are over 11MB, so CapsuleApp fails with a capsule size error.
USB driver support did not make the rel-34 release. The attached pull requests add support for this feature
I was looking through this source code for a way to disable OP-TEE but wasn't able to find one. Is there a way I can disable OP-TEE so that no physical address range is reserved for OP-TEE? If this isn't possible, is there a way to find the physical address range that has been reserved for OP-TEE?
edk2_docker edk2-nvidia/Platform/NVIDIA/Jetson/build.sh
found in
https://github.com/NVIDIA/edk2-nvidia/wiki/Build-without-docker#build-uefi
but if i try with edk2_docker inculded i will get edk2_docker: command not found
The logic in board eeprom handling does not work correctly for customer part numbers, as it did with previous implementations.
While working on this pull request:
I discovered what appears to be a bug in the logic that is attempting to find a match to either the compatibility or full TnSpec. I think the simple fix is to add a break here:
The bug was exposed due to meta-tegra having additional bup specs that differed from the L4T 35.3.1 release (which was resolved in the PR). I was able to recreate it using the L4T release by simply adding this spec:
'boardid=3668;fab=200;boardsku=0001;boardrev=;fuselevel_s=1;chiprev=2;board=jetson-xavier-nx-devkit-emmc;rootdev=mmcblk0p1'
to the jetson-xavier-nx-devkit-emmc board of the t19x_spec in jetson_board_spec.cfg and then rebuilding the capsule per here. With the additional spec present the code will fail to break out of the loop and will match on the next entry, but since Found is now TRUE it will fall into the else clause that returns EFI_UNSUPPORTED.
Hi, the boot time of UEFI firmware costs a lot of time. I have changed PcdPlatformBootTimeOut from 5 to 0, but it still took about 16 seconds.
I noticed a large gap before booting the kernel in the log, so I wanted to ask if there is any way to shorten the startup time.
Thank you.
On Xavier NX platforms where some _b FW partitions are on an eMMC device, FMP capsule update may fail with write protect errors when updating boot chain B FW, E.g.
Overwriting active adsp-fw partition not allowed
Error writing adsp-fw, Offset=0: Write Protected
Failed to write image=adsp-fw: Write Protected
This is caused by the FwImageDxe driver loading before the FwPartitionBlockIoDxe driver which makes the _b partitions unavailable during the capsule update. A fix to the FwImageDxe Depex is being developed to ensure that the drivers load in the correct order.
When switching the rootfs, OS still thinks the switch has not happened despite of it actually having happened.
The TegraPlatformSpec and TegraPlatformCompatSpec UEFI variables read by FmpDeviceLib are not written with null termination bytes. UEFI shoud ensure its copies of these variables are null-terminated passing them to functions that require null-terminated strings.
If system is configured for ACPI mode in some boot flows a minimal DTB is installed by the OS launcher application and in this case there may be system crashes related to the application of device tree overlays.
When using a DVI to HDMI converter may fault and go into a reboot loop. VGA to HDMI and HDMI to HDMI work.
Need to shutdown network activity at UEFI exit especially in the case of ACPI boot as the memory set aside for that may be reclaimed by operating system.
Hello,
I've been updating our xavier platform to use the 35.1 using an AGX Xavier devkit and the new uefi bootloader. We are running a yocto meta-tegra kirkstone image, which by default does not come with a recovery partition. The head of kirkstone on meta-tegra right now also does not run nvbootctrl verify on boot and I ran into the following issue:
Once the retry count for slot A reached 0 the platform will no longer boot. Since there is no recovery partition it drops to the UEFI Shell. There does not appear to be any way from UEFI shell to reset the retry count, nor to force the platform to boot from slot A anyway, nor to place the device in forced-recovery mode to allow flashing over USB.
The only way I am aware of to get out of this state is to have someone physically present place the device into forced recovery mode, after which USB-C flashing works fine. This is very inconvenient for remote developers though.
Shell> setvar RootfsInfo -guid 781E084C-A330-417C-B678-38E696380CB9 -bs -rt -nv
781E084C-A330-417C-B678-38E696380CB9 - RootfsInfo - 0004 Bytes
0C C3 00 00
Shell> setvar RootfsInfo -guid 781E084C-A330-417C-B678-38E696380CB9 -bs -rt -nv
=0x0030C30C
Shell> setvar RootfsInfo -guid 781E084C-A330-417C-B678-38E696380CB9 -bs -rt -nv
781E084C-A330-417C-B678-38E696380CB9 - RootfsInfo - 0004 Bytes
0C C3 30 00
Shell> mm 0x0C39041C -w 4 -n
MEM 0x000000000C39041C : 0x000CFACE
Shell> mm 0x0C39041C -w 4 -n =0x003CFACE
MEM 0x000000000C39041C : 0x003CFACE
Shell> mm 0x0C39041C -w 4 -n
MEM 0x000000000C39041C : 0x003CFACE
neither reset -c nor reset -w following either of the above changes had any effect.
Attempting to boot always results in
Jetson UEFI firmware (version 1.0-d7fb19b built on 2022-08-10T20:18:13-07:00)
Press ESCAPE for boot options ** WARNING: Test Key is used. **
......
L4TLauncher: Attempting Recovery Boot
Header not seen at either offset 0 or offset 0x1000
Failed to boot recovery:0 partition
...
Press ESC in 1 seconds to skip startup.nsh or any other key to continue.
Shell>
Based on my understanding of nvidia's documentation, these are what the reboot logic should be basing the decision to not boot to slot A on.
based on how "forced-recovery" reboots are handled by drivers/soc/tegra/pmc.c I tried to read 0xc362000 with the intention of setting the PCM_SCRATCH0_MODE_RCM bit, but this resulted in an exception and reset of the bootloader:
Shell> mm 0xC362000 -w 4 -n
MEM
Synchronous Exception at 0x0000000857C88874
Synchronous Exception at 0x0000000857C88874
ASSERT [ArmCpuDxe] /dvs/git/dirty/git-master_linux/out/nvidia/bootloader/uefi/Jetson_RELEASE/edk2/ArmPkg/Library/DefaultExceptionHandlerLib/AArch64/DefaultExceptionHandler.c(333): ((BOOLEAN)(0==1))
Resetting the system in 5 seconds.
When I build edk2 - nvidia/Platform/nvidia/Jetson/build. Sh, there will be such a mistake.
Traceback (most recent call last):
File "edk2/BaseTools/Edk2ToolsBuild.py", line 177, in
main()
File "edk2/BaseTools/Edk2ToolsBuild.py", line 173, in main
Edk2ToolsBuild().Invoke()
File "/home/user/Jetson_UEFI/nvidia-uefi/venv/lib/python3.8/site-packages/edk2toolext/base_abstract_invocable.py", line 146, in Invoke
retcode = self.Go()
File "edk2/BaseTools/Edk2ToolsBuild.py", line 148, in Go
raise Exception("Failed to build.")
Exception: Failed to build.
Hi,
When following the docs to build UEFI without docker(https://github.com/NVIDIA/edk2-nvidia/wiki/Build-without-docker)
Encountered below error with
[main]$ edk2-nvidia/Platform/NVIDIA/Jetson/build.sh
nvidia: building from workspace rooted at: /home/fadhelh/project/octo_v1/uefistrap/nvidia-uefi
nvidia: found command: python3.
nvidia: found command: virtualenv.
nvidia: found command: mono.
nvidia: found command: aarch64-linux-gnu-gcc.
nvidia: found Python 3.9 or later.
nvidia: Activating Python virtual environment.
nvidia: Updating build environment (edk2-nvidia/Platform/NVIDIA/Jetson/PlatformBuild.py).
Unable to use edk2-nvidia/Platform/NVIDIA/Jetson/PlatformBuild.py as a UpdateSettingsManager
Did you mean to use a different kind of invocable?
Traceback (most recent call last):
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toolext/edk2_invocable.py", line 355, in ParseCommandLineOptions
self.PlatformModule = import_module_by_file_name(os.path.abspath(settingsArg.platform_module))
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toollib/utility_functions.py", line 347, in import_module_by_file_name
spec.loader.exec_module(ImportedModule)
File "", line 848, in exec_module
File "", line 219, in _call_with_frames_removed
File "/home/fadhelh/project/octo_v1/uefistrap/nvidia-uefi/edk2-nvidia/Platform/NVIDIA/Jetson/PlatformBuild.py", line 10, in
from edk2nv.stuart import NVIDIASettingsManager, NVIDIAPlatformBuilder
File "/home/fadhelh/project/octo_v1/uefistrap/nvidia-uefi/edk2-nvidia/Silicon/NVIDIA/edk2nv/stuart/init.py", line 10, in
from .builder import * # noqa
File "/home/fadhelh/project/octo_v1/uefistrap/nvidia-uefi/edk2-nvidia/Silicon/NVIDIA/edk2nv/stuart/builder.py", line 12, in
from edk2toolext.environment.uefi_build import UefiBuilder
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toolext/environment/uefi_build.py", line 24, in
from edk2toollib.uefi.edk2.parsers.dsc_parser import DscParser
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toollib/uefi/edk2/parsers/dsc_parser.py", line 13, in
from edk2toollib.uefi.edk2.parsers.base_parser import HashFileParser
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toollib/uefi/edk2/parsers/base_parser.py", line 13, in
from edk2toollib.uefi.edk2 import path_utilities
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toollib/uefi/edk2/path_utilities.py", line 21, in
class Edk2Path(object):
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toollib/uefi/edk2/path_utilities.py", line 329, in Edk2Path
def GetContainingModules(self, input_path: str) -> list[str]:
TypeError: 'type' object is not subscriptable
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/fadhelh/.local/bin/stuart_update", line 8, in
sys.exit(main())
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toolext/invocables/edk2_update.py", line 120, in main
Edk2Update().Invoke()
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toolext/base_abstract_invocable.py", line 200, in Invoke
self.ParseCommandLineOptions()
File "/home/fadhelh/.local/lib/python3.8/site-packages/edk2toolext/edk2_invocable.py", line 367, in ParseCommandLineOptions
Module = self.PlatformModule
AttributeError: 'Edk2Update' object has no attribute 'PlatformModule'
Please comment
Regards
Changed console settings under Boot Menu => Boot Maintenance Manager => Console Options => Console [Input|Output|Standard Error] Device Select do not persist through a device reboot.
[2023-10-24 16:38:22] Invalid EFI_I2C_DEVICE reported by I2c Enumerate protocol.
[2023-10-24 16:38:22] Invalid EFI_I2C_DEVICE reported by I2c Enumerate protocol.
[2023-10-24 16:38:22] Invalid EFI_I2C_DEVICE reported by I2c Enumerate protocol.
[2023-10-24 16:38:22] Invalid EFI_I2C_DEVICE reported by I2c Enumerate protocol.
[2023-10-24 16:38:22] Invalid EFI_I2C_DEVICE reported by I2c Enumerate protocol.
[2023-10-24 16:38:23] Installed Fat filesystem on 8221EE398
[2023-10-24 16:38:23] [ext4] Needs journal recovery, mounting read-only
[2023-10-24 16:38:25] TegraI2cStartRequest: Timeout waiting for Packet Complete
[2023-10-24 16:38:25] TegraI2cStartRequest: Timeout waiting for Packet Complete
[2023-10-24 16:38:25] TegraI2cStartRequest: Timeout waiting for Packet Complete
[2023-10-24 16:38:25] TegraI2cStartRequest: Timeout waiting for Packet Complete
[2023-10-24 16:38:26] TegraI2cStartRequest: Timeout waiting for Packet Complete
[2023-10-24 16:38:26] PrintBootOrder: Current BootOrder:
[2023-10-24 16:38:26] PrintBootOrder: BootOrder[0] = 0x0008 = UEFI I-O DATA USB Flash Disk 1242609221250144
[2023-10-24 16:38:26] PrintBootOrder: BootOrder[1] = 0x0001 = UEFI eMMC Device
It's important boot time on some embedded scene. so, We investigate how many seconds spent each compornent. We found the above time out messages.
Hi, now when i switch branch main to r35.3.1-updates, it compiled failed
ERROR - Invalid file path: ('DynamicTablesPkg/Library/Acpi/Arm/AcpiSlitLibArm/AcpiSlitLibArm.inf',)
ERROR - Build.py #000 from /home/ultwcz/work/prj/nx/nvidia-uefi-test/edk2-nvidia/Platform/NVIDIA/Jetson/Jetson.dsc(941): File/directory not found in workspace
What should i do?
Thank you.
Hi,
We are working on Nvidia Orin platform.
We observed the a longer boot time and need to optimize it.
Please show us how to disable BootLogo,
UEFI branch: R35.4.1
Regards
As I was reading through the code I noticed that
probably was meant to be: FwPackageCopyImageName (PkgName, PkgImageInfo, sizeof (PkgName));instead of the current:
FwPackageCopyImageName (PkgName, PkgImageInfo, FW_IMAGE_NAME_LENGTH);as the length is in terms of bytes, and not CHAR16s from my reading of the function's description, and other usages of it in this file.
Hi,
I am using leetop agx xavier which installed 250GB SSD, Samsung SSD 980 series. but, it is stuck and do not disappeared booting logo end of progress bar after I installed 2TB SSD the same brand. After some seconds, it try to reboot and repeat over and again.
I could only see message from UEFI.
Jetson UEFI firmware (version 2.1-32413640 built on 2023-01-24T23:12:27+00:00)
ESC to enter Setup.
F11 to enter Boot Manager Menu.
Enter to continue boot.
** WARNING: Test Key is used. **
......L4TLauncher: Attempting Recovery Boot
EFI stub: Booting Linux Kernel...
EFI stub: Using DTB from configuration table
EFI stub: Loaded initrd from LINUX_EFI_INITRD_MEDIA_GUID device path
EFI stub: Exiting boot services and installing virtual address map...
Is there any way to debug this issue?
BR,
Mark
I recently opened a forum post about this issue. In diagnosing I added logs around why the nvme was failing to be recognized and found that the NvmExpress Driver was failing to create a second queue for the Completion and Submission queues.
As a simple test I hardcoded the NvmeCreateIoCompletionQueues and NvmeCreateIoSubmissionQueues functions to stop after the first queue which did work to get the system to boot from my NVMe drive, but I know it is not a valid long term solution that would work for all drives.
When a device is attached to the /dev/ttyTHS0 uart port, if it streams data it interrupts the boot process. This behavior isn't expected, as it is the only portion of the boot process which outputs or reads from this serial port.
On Xavier NX SD platforms where some _b FW partitions are on an SD device, FMP capsule update will fail with write protect errors when updating boot chain B FW, E.g.
Overwriting active adsp-fw partition not allowed
Error writing adsp-fw, Offset=0: Write Protected
Failed to write image=adsp-fw: Write Protected
This is caused by the FwPartitionBlockIoDxe driver excluding SD partitions based on device path which makes those _b partitions unavailable during the capsule update.
EQOS is not starting in UEFI correctly causing the device to not work correctly.
Add a .gitattributes for the binary assets; check in source files in a manner that line endings are not CRLF but utilize the best practices for source modules being checked into a repository having their line endings as LF. This is controlled by a client using the git config setting core.autocrlf on Windows set to something other than false.
.gitattributes example:
This shows up when attempting to apply patches from a machine running Linux using bitbake as a build system.
Possible command to utilize:
find . -type f ! -path './.git/' ! -path '.bmp' -print | xargs -L 1 dos2unix
SMBIOS table type 19 is used to get memory information in UEFI menu.
Linux does not work with the GICR region described in ACPI.
During UEFI boot, the system sometimes hits an unhandled exception in EL3.
Hello, two questions here
What is edk2-nvidia? Is it a version of nvidia's UEFI for Jetson?
How do I change the Jetson's UEFI variables and settings (I'm using a Jetson AGX Orin)? How do I use the information on this VariableList page?
I think my questions may help others too.
Thank you
Jetson Linux 34.1 suffers from an issue where booting Orin with a 4K display attached results in a framebuffer overrun and a boot failure. This has been tracked as known bootloader issue 3581719.
Pull request #3 corrects the framebuffer size computation, fixing this issue.
Hi,
When following steps https://github.com/NVIDIA/edk2-nvidia/wiki/Build-with-docker
I'm seeing permission denied when giving
$ edk2_docker edkrepo clone nvidia-uefi NVIDIA-Platforms main
Syncing the global manifest repository: /home/fadhelh/.edkrepo/edk2-edkrepo-manifest-main
Syncing the global manifest repository: /home/fadhelh/.edkrepo/nvidia
Error: [Errno 13] Permission denied: '/home/fadhelh/project/octo_v1/uefi_docker/build/nvidia-uefi
When tried with
$sudo edk2_docker edkrepo clone nvidia-uefi NVIDIA-Platforms main
sudo: edk2_docker: command not found
I could even see hello when giving
edk2_docker echo hello
Please guide.
Regards
I was trying to enter BIOS menu but the keyboard is not working at boot up.
ESC, F11 and ENTER buttons are not working.
Specs:
Jetson Xavier NX with SEEED A206 carrier board
Jetpack 5.1 (L4T 35.2.1)
Keyboard works up to edk2-nvidia-r35.1-updates-20220930 6bc2699.
Keyboard is not working starting edk2-nvidia-202209.0 a87aded.
Tested by flasing uefi_jetson.bin in tarball from releases.
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.