SOF v2.3 #106
SOF v2.3 #106
Conversation
mengdonglin
requested review from
keqiaozhang,
kv2019i,
lgirdwood and
marc-hb
|
Still debugging the checksum difference between community-signed and Intel-signed FW binaries. |
|
@lgirdwood @marc-hb I think the PR is ready to merge. The PR BATS check failure is caused by a known issue on Zephyr side: The devicetree dependencies in .__device_handles_pass2 section has the same content but maybe in different order in different builds. This issue was fixed by commit scripts: gen_handles: Sort the device handles in Zephyr upstream. This fix is not in Zephyr snapshot for SOF v2.3 But with the fix we can compare different builds in future releases. |
Great find. Can we backport this to a new stable-v2.3 branch here? https://github.com/thesofproject/zephyr/branches We may need this stable branch in the future anyway for other reasons. Then we can tag and release SOF v2.3.1 with it and make sure there's no mismatch. |
|
Since this is a device tree ordering issue (where the DT data has the same values) and we have validated the binaries with this DT ordering then I'm good to merge this PR @mengdonglin. We can do v2.3.1 if we have runtime issues and include the DT ordering update. |
|
@marc-hb I adjusted the BATS test to allow checksum difference for SOF v2.3 FW binaries. Now we can pass all PR checks. @lgirdwood I added one commit to this PR to adjust BATs test as Marc suggested. We're good to merge. |
Add v2.3 firmware binaries for Intel Tiger Lake, Alder Lake, and Raptor Lake platforms. This is the first release using Zephyr as the RTOS on these 3 platforms: https://github.com/thesofproject/sof/releases/tag/v2.3 Name prefix sofz- is used for v2.3 FW to distinguish from v2.2.x FW using XTOS as the RTOS. v2.3.x/ └── sof-v2.3 ├── community │ ├── sofz-adl.ri -> sofz-tgl.ri │ ├── sofz-adl-s.ri -> sofz-tgl-h.ri │ ├── sofz-rpl.ri -> sofz-tgl.ri │ ├── sofz-rpl-s.ri -> sofz-tgl-h.ri │ ├── sofz-tgl-h.ri │ └── sofz-tgl.ri ├── intel-signed │ ├── sofz-adl.ri │ ├── sofz-adl-s.ri │ ├── sofz-rpl.ri -> sofz-adl.ri │ ├── sofz-rpl-s.ri -> sofz-adl-s.ri │ ├── sofz-tgl-h.ri │ └── sofz-tgl.ri ├── sofz-adl.ldc -> sofz-tgl.ldc ├── sofz-adl.ri -> intel-signed/sofz-adl.ri ├── sofz-adl-s.ldc -> sofz-tgl-h.ldc ├── sofz-adl-s.ri -> intel-signed/sofz-adl-s.ri ├── sofz-rpl.ldc -> sofz-tgl.ldc ├── sofz-rpl.ri -> intel-signed/sofz-rpl.ri ├── sofz-rpl-s.ldc -> sofz-tgl-h.ldc ├── sofz-rpl-s.ri -> intel-signed/sofz-rpl-s.ri ├── sofz-tgl-h.ldc ├── sofz-tgl-h.ri -> intel-signed/sofz-tgl-h.ri ├── sofz-tgl.ldc └── sofz-tgl.ri -> intel-signed/sofz-tgl.ri Signed-off-by: Mengdong Lin <mengdong.lin@intel.com>
For shell completion convenience. Signed-off-by: Mengdong Lin <mengdong.lin@intel.com>
Allow checksum differences for v2.3 FW binaries v2.3 is the 1st SOF release using Zephyr as RTOS. But there's an unintended dependency on python version that would result in differences in the output FW binaries built in different environment. This has been fixed in Zephyr mainline by commit zephyrproject-rtos/zephyr@f896fc2306 So future SOF binary releases will have same checksums. Signed-off-by: Mengdong Lin <mengdong.lin@intel.com>
Be consistency with zephyr.strip This will help with reproducibility issues like the one in thesofproject/sof-bin#106 Use the `strip_command` introduced by commit c060b07 ("cmake: toolchain: bintools abstraction") boot.mod is already deterministic because it has no debug symbols; no need to strip it. Signed-off-by: Marc Herbert <marc.herbert@intel.com>
|
These 2 lines in Zephyr should help: |
Be consistency with zephyr.strip This will help with reproducibility issues like the one in thesofproject/sof-bin#106 Use the `strip_command` introduced by commit c060b07 ("cmake: toolchain: bintools abstraction") boot.mod is already deterministic because it has no debug symbols; no need to strip it. Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Be consistency with zephyr.strip This will help with reproducibility issues like the one in thesofproject/sof-bin#106 Use the `strip_command` introduced by commit c060b07 ("cmake: boot.mod is already deterministic because it has no debug symbols; no need to strip it. (cherry picked from commit 57ea358) Original-toolchain: bintools abstraction") Original-Signed-off-by: Marc Herbert <marc.herbert@intel.com> GitOrigin-RevId: 57ea358 Change-Id: Id7d856377a86cffa63d50b4d253aae7d23ff2f3f Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/zephyr/+/4062711 Tested-by: Aaron Massey <aaronmassey@google.com> Commit-Queue: Aaron Massey <aaronmassey@google.com> Reviewed-by: Aaron Massey <aaronmassey@google.com> Reviewed-by: Abe Levkoy <alevkoy@chromium.org>
Be consistency with zephyr.strip This will help with reproducibility issues like the one in thesofproject/sof-bin#106 Use the `strip_command` introduced by commit c060b07 ("cmake: toolchain: bintools abstraction") boot.mod is already deterministic because it has no debug symbols; no need to strip it. Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Add v2.3 firmware binaries for Intel Tiger Lake, Alder Lake, and Raptor Lake platforms. This is the first release using Zephyr as the RTOS on these 3 platforms: https://github.com/thesofproject/sof/releases/tag/v2.3
Please note
fw_filename=sofz-<platform name>.ri