Create kernel_qemu_aarch64_run rule #1060
Conversation
bazel/linux/runner.bzl
Outdated
| @@ -195,6 +201,13 @@ def create_runner(ctx, archs, code, runfiles = None, extra = {}): | |||
| ), | |||
| ) | |||
|
|
|||
| if ki.arch == "aarch64": | |||
There was a problem hiding this comment.
just a dumb note, nothing to do --- the compiler and bazel both use aarch64, but the Linux kernel uses arm64. it so dumb these are different in different places... i think what you have here is correct. when writing anything bazel use aarch64.
sudhakar-mamillapalli
force-pushed
the
master
branch
from
47b86b2
to
75375ab
Compare
|
Rebased. |
sudhakar-mamillapalli
force-pushed
the
master
branch
2 times, most recently
from
c3d9764
to
7a010db
Compare
sudhakar-mamillapalli
force-pushed
the
master
branch
from
7a010db
to
ff91d66
Compare
bazel/linux/qemu.bzl
Outdated
| echo 1>&2 '$' "$QEMU_BINARY" "${{QEMU_FLAGS[@]}}" "${{WRAPPER_OPTS[@]}}" | ||
| if [ -z "$INTERACTIVE" -a -z "$SINGLE" ]; then | ||
| "$QEMU_BINARY" "${{QEMU_FLAGS[@]}}" "${{WRAPPER_OPTS[@]}}" </dev/null | tee "$OUTPUT_FILE" | ||
| "$QEMU_BINARY" "${{QEMU_FLAGS[*]}}" "${{KERNEL_FLAGS[*]}}" "${{WRAPPER_OPTS[*]}}" </dev/null | tee "$OUTPUT_FILE" |
There was a problem hiding this comment.
The original code had:
QEMU_FLAGS+=("-append" "${{KERNEL_FLAGS[*]}} ${{KERNEL_OPTS[*]}}")
... what happened to KERNEL_OPTS in the new code? Can't find anywhere the -append, and handling of KERNEL_OPTS?
There was a problem hiding this comment.
@ccontavalli Very sorry. Seems like I dropped KERNEL_OPTS when I mucked with args. As for -append I pass all the args (kernel/qemu and wrapper args) separately to the model-wrapper-socket script in internal repo and do the append there (a different PR https://github.com/enfabrica/internal/pull/40251).
Any way thinking again, I went back the old way of doing things since if I do the above I realized there are other model-wrapper scripts which need to be changed too. So I just pass the arguments similar to how we use to do so before.
Also I just use the machine name now to know if the qemu to use is aarch64 or x86 one, instead of passing it as a wrapper option. That way do not have to muck with wrapper opts. If you rather have the arch passed as a wapper opts, I can change that. Please let me know.
Also to make it easy to read I am force pushing the diff. Please take a look again. Sorry again for the mess-up and thanks for finding the issue.
sudhakar-mamillapalli
force-pushed
the
master
branch
from
ff91d66
to
6db3c57
Compare
* For launching aarch64 qemu acf-machine a bootram image is needed.
This image is loaded into SPS bootram and on qemu launch execution
starts there. The actual image (bootloader + kernel + initramfs + devicetree) is loaded
into DRAM. The bootram image does mimimal stuff and jumps to bootloader
which continues rest of the boot.
* This commit just creates a provider for the BootRam image.
* We want to be able to run a qemu machine emulating
E1 system complex which connect to the SFA FE model
via netpci for dev and testing.
* So based on kernel_qemu_run which works for x86 arch,
kernel_qemu_aarch64_run target is introduced which
pretty much follows the same flow as kernel_qemu_run
with changes needed for aarch64.
sudhakar-mamillapalli
force-pushed
the
master
branch
from
6db3c57
to
eb0565d
Compare
|
rebase |
| ), | ||
| **{ | ||
| "archs": attr.string_list( | ||
| default = ["aarch64"], |
There was a problem hiding this comment.
if we are passing the archs, do we need a separate kernel_qemu_run rule ? Could we not just use the existing one ?
Create kernel_qemu_aarch64_run rule.
We want to be able to run a aarch64 qemu machine emulating E1 system complex which connect to the SFA FE model via netpci for dev and testing.
So based on kernel_qemu_run which works for x86 arch, kernel_qemu_aarch64_run target is introduced which pretty much follows the same flow as kernel_qemu_run with changes needed for aarch64.