| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
| |
The team working on the XXV + EOE implementation (25G PL Ethernet using SFP+)
observed a bug in iperf3 which causes a roughly 50% drop in the bitrate (from
9.57Gbps down to 5.35Gbps). On scarthgap iperf3 is at version 3.16; this bug
is fixed in version 3.18. Instead of bumping the entire code base to 3.18 it
would be safer to simply apply the fix for this bug onto 3.16 instead.
Link: https://github.com/esnet/iperf/pull/1708/commits/ac6b9f7fd335ddebc5212eed40083ef4cd3cb86d
Signed-off-by: Trevor Woerner <trevor.woerner@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Signed-off-by: John Toomey <john.toomey@amd.com>
Also expanded the DESCRIPTION of 'ISP'
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
runqemu and related support requires a Linux host operating system to work.
We still want to be able to export a QEMU binary for use on Windows, but the
wrapper scripts and related will still require Linux.
The dependencies are primarily related to python3, as well as bootgen's xcrypt
not being available for mingw32.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Use boot_files to dump images all the images and remove bh, plm and
pcm_cdo dump options.
Also add print messages for plm and pmc_cdo load address which helps
for debugging.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix runqemu QEMU MB instance below warnings.
qemu-system-microblazeel: -device loader,file=/scratch/sandeep/yocto/2025.1/yp-bc/build/tmp/deploy/images/versal-vek280-sdt-seg/BOOT-versal-vek280-sdt-seg_bh.bin,addr=0xF201E000,force-raw: warning: short-form boolean option 'force-raw' deprecated
Please use force-raw=on instead
qemu-system-microblazeel: -device loader,file=/scratch/sandeep/yocto/2025.1/yp-bc/build/tmp/deploy/images/versal-vek280-sdt-seg/CDO/pmc_cdo.bin,addr=0xf2000000,force-raw: warning: short-form boolean option 'force-raw' deprecated
Please use force-raw=on instead
So use "force-raw=on" in QEMU_HW_FW_FILES arguments.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
| |
Add nativesdk to BBCLASSEXTEND to match meta-openembedded
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
In meta-virtualization, when used with meta-xilinx-core, the XEN dtsi was
always enabled, even in a non-Linux based build. This caused compilation
issues as cortex-r5 and other CPUs are missing required parameters for Xen.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
| |
deploy should only be performed in the native case, otherwise a conflict
and error can occur.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Allow the qemu SDK to have full access to all of the runqemu related work.
This will allow someone with the SDK to use runqemu with a qemuboot.conf file.
Remove 'unset OECORE_NATIVE_SYSROOT', this should only be in
buildtools-tarball, not in a regular SDK.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Used to combine multiple qemuboot.conf files, non-blank values replace prior values.
./qemuboot-tool: <command> <argument> ...
load <file> - load a new file
merge <file> - load and fill any new/empty variables
remove <var> - Remove a variable
Commands can be chained, such as:
load file1 remove val1 merge file2
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Instead of using the hard coded path, ${DEPLOY_DIR_IMAGE} we should be using
@DEPLOY_DIR_IMAGE@ where applicable. The exception is QB_DTB where the
qemuboot.conf export function will sanitize the path for us.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
On AMD FPGA SoC machines, we almost always include the serial port options
with the core QB_OPT_APPEND value. If a machine requires QB_SERIAL_OPT
it can override them.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
| |
In the same way that an image recipe would generate a qemuboot.conf, also
generate one for the boot.bin. This will allow us to use runqemu with
just the boot.bin, as well as merge boot.bin specific items with common
(non-machine specific) filesystem qemuboot.conf.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
| |
This will allow the caller of runqemu to override the APU's qemu dtb using
the standard DEVICE_TREE environment option to runqemu.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
To ensure it is easier to override the APU device-tree, support -dtb as an
analog for -hw-dtb to the APU. Additionally, runqemu needs to define a kernel
to pass -dtb, so ignore the -kernel parameter as it doesn't do anything in
the system emulated APU.
Also check and error if an APU DTB is not passed in, either via -dtb or -hw-dtb.
This can cause qemu to lock up, try to prevent this error.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
| |
Adjust the qemu-system-aarch64-multiarch script to remove reference to these.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
The common machine are missing the necessary boot firmware, so clear the QB
(qemu) settings. This will make it easier to 'merge' in the board specific
configurations, where applicable.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Copy all dependent files into ${B}
Construct the bif with relative paths in the ${B} (build) directory
Deploy the resulting build to:
boot.bin-extracted
This is similar behavior to the SDT workflow's 'extracted' directory and
will enable people to re-run bootgen, as necessary, outside of the build
environmet.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
| |
In order to coordinate the deployed filename and machine qemu setting, use
the common variable PMU_ROM.
Note, the variable needs to be defined in a global location as native
recipes are not run with machine.conf loaded.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
| |
Currently the qemu flash drive bin files has static qspi suffix for
both qspi and ospi boot mode, Instead of generating the qspi always
based on OSPI boot mode value generate the qemu ospii flash bin file.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
| |
Refactor QEMU_HW_BOOT_MODE to accept only boot mode value and move
"-boot mode" to qb_extra_args. This makes checking for boot mode value
easier in getVar operations.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The commit:
commit 4b39c6111b965ddd43b899302f9f41d49cec47c4
Author: Jonathan Stroud <jonathan.stroud@amd.com>
Date: Wed Jan 15 16:31:24 2025 -0700
Backport pmufw fix to 2024.2 that fixes a bug in GEM IO register mapping
PMU FW has a bug that stomps on the MIO registers when intializing GEM.
Backport this patch from master to 2024.2 builds
was applied incorrectly due to recent changes in the way the
software was being patched. Correct this and match the current
embeddedsw-source implementation.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
| |
When we've disabled a recipe due to preferred provider status, we do NOT
want to cache the result. If the result is cached, and the user adjusts
the value it may not be recognized requiring the user to delete the
cache directory as a workaround.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
gen-machine-conf is moving to using TFA_CONSOLE to match the newer TFA
recipe implementations. Adjust the older arm-trusted-firmware to match
while retaining compatibility with prior ATF_CONSOLE if required.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QB_MACHINE is now set in each base machine. This ensures that the default
configuration is correct, and can still be overriden if necessary.
This solves a problem where bitbake -e could fail to expand properly
in some cases.
QB_DEFAULT_KERNEL was removed, and restored back to the default settings.
The defaults now work properly with all tested configurations.
Note: simpleImage.mb is obsolete and no longer used in this context.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
QB_ variables are automatically included in the generated qemuboot.conf
files, even if they are not used by runqemu. Rename various AMD specific
values to avoid them being added to the file.
QB_BOOT_MODE -> QEMU_HW_BOOT_MODE
QB_SD_DRIVE_INDEX -> QEMU_HW_SD_DRIVE_INDEX
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
| |
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Add nativesdk recipe for qemu-xilinx-system multiarch which is required
for runqemu when you source sdk scripts and use runqemu script.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For some use case user would like to provide custom machine path for
qemu boot. By default we will create a temporary files using the
tempfile.mkdtemp method, when user provides a custom machine path from
runqemu command line args then skip the tempfile.mkdtemp method.
Here is how user can specify a custom machine path from qemuparams
option.
$ runqemu nographic qemuparams="-machine-path /<custom-mach-path>"
Also note when using custom machine path its user responsibility to
clean that path before running the runqemu launch.
QEMU machine-path option is a socket path. Being a socket path it’s
limited to a max of 107 characters. Since we’re passing in a path,
the socket itself is part of the path so it’s probably closer to 90
characters.
Also add missing Copyright and SPDX-License-Identifier.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
| |
Backport bmap-writer recipe from meta-openembedded master
Signed-off-by: John Toomey <john.toomey@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Add apu_ss attibute name to versal bif files as this is required linux
power management which is enabled by default in kernel defconfig.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Enable examples and other packageconfig which is required for
packagegroup-xilinx-qt.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Multiconfig dependencies are configured to be specific to a single from
and to multiconfig, for example:
mc:default:xilinx-image-recovery:core-image-tiny-initramfs:do_image_complete
The above says, when in the 'default' multiconfig, we depend on the
'xilinx-image-recovery' multiconfig, and need to run the do_image_complete
task within the core-image-tiny-initramfs recipe.
Two workarouns are present, the first adjusts:
mc::to_mc:recipe:task
to be
mc:${BB_CURRENT_MC}:to_mc:recipe_task
This will allow the mcdependencies produced by gen-machine-conf to wor in
any multiconfig the user may have. In the future we will adjust
gen-machine-conf to use ${BB_CURRET_MC} instead.
Second there is a bug/oversight in bitbake where the BB_CURRENT_MC reports
'default', but the system is expecting '' (blank) when in the default Linux
multiconfig. We will do this conversion to avoid any issues. A patch for
bitbake is in progress.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
| |
We need to create a custom do_deploy_source_date_epoch to copy the file from
the shared location, where applicable.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Both python3-pybind11 and python3-pybind11-json exist in upstream meta-oe.
However, as part of another task, when python3-bokeh_3.5.1 was updated one of
its dependencies (python3-contourpy_1.2.1) required a newer version of
python3-pybind11 than what was available in meta-oe scarthgap. A second task
now needs python3-pybind11-json as well, which has a dependency on
python3-pybind11.
With the old:
inherit pypi python_setuptools_build_meta
python3-pybind11-json was not able to find python3-pybind11 during its
configure stage:
| CMake Error at CMakeLists.txt:26 (find_package):
| By not providing "Findpybind11.cmake" in CMAKE_MODULE_PATH this project has
| asked CMake to find a package configuration file provided by "pybind11",
| but CMake did not find one.
|
| Could not find a package configuration file provided by "pybind11"
| (requested version 2.2.4) with any of the following names:
|
| pybind11Config.cmake
| pybind11-config.cmake
|
| Add the installation prefix of "pybind11" to CMAKE_PREFIX_PATH or set
| "pybind11_DIR" to a directory containing one of the above files. If
| "pybind11" provides a separate development package or SDK, be sure it has
| been installed.
Switching to:
inherit cmake python_setuptools_build_meta
fixes this issue and allows the python3-pybind11-json recipe from upsream
meta-oe to configure (and build) successfully.
Signed-off-by: Trevor Woerner <trevor.woerner@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Create recipe to deploy cdo files from system-device-tree dir to deploy
dir.
bootgen will fail if the image path in the bif is larger, this will fix
that issue by using images from deploy dir.
Signed-off-by: Raju Kumar Pothuraju <rajukumar.pothuraju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
| |
Signed-off-by: John Toomey <john.toomey@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
'cp' is not always replacing an existing file, but appears to sometimes change
the file. This is not normally an issues but when using shared-source and
files that are hardlinks, we need to do a Copy-on-Write style operation
which install does by default.
Adjust all instances of:
cp Makefile.Linux Makefile
with:
install Makefile.Linux Makefile
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
| |
Per request, enable medany memory model with riscv 64-bit and newlib tclibc.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
| |
'b' extensions require binutils 2.43 or newer.
Update the generation script to correctly identify and set the TUNE_FEATURE
for riscv 32-bit and 64-bit. This may be needed in the future.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In order to support additional risc-v extensions, specifically 'b', binutils
2.43.1 is required. Backport binutils 2.43.1 from styhead as of commit:
commit b2cd1dea0459acec76e601987ad4b0661babfcc3
Author: Deepthi Hemraj <Deepthi.Hemraj@windriver.com>
Date: Tue Nov 12 04:28:09 2024 -0800
binutils: stable 2.43.1 branch update
Set this version as NOT preferred. We do not want to override the regular
Scarthgap binutils version unless we explicitly need this version.
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
We are using psmfw.elf in all flows to build and boot.
Remove the psmfw.bin dependency or support from psmfw recipe.
Use below command to generate the .bin from .elf.
objcopy -O binary psm-$MACHINE.elf psm.bin
Signed-off-by: Raju Kumar Pothuraju <rajukumar.pothuraju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
|
|
|
| |
We are using pmufw.elf in all flows to build and boot.
Remove the pmufw.bin dependency or support from pmufw recipe.
Use below command to generate the .bin from .elf.
objcopy -O binary pmu-$MACHINE.elf pmu.bin
Signed-off-by: Raju Kumar Pothuraju <rajukumar.pothuraju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|
|
|
|
|
|
|
|
| |
Remove installing pmc_cdo.bin to DEPLOYDIR and create pmc_cdo.bin
symlink in DEPLOYDIR.
Signed-off-by: Sandeep Gundlupet Raju <sandeep.gundlupet-raju@amd.com>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
|