-
Notifications
You must be signed in to change notification settings - Fork 321
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[BUG] Platform-specific IPC4 code in platform-independent source files #7549
Comments
Some documentation on what this required component is doing is also useful for porting other platforms (we want to switch to IPC4 at NXP, and in my quick look this was one of the places where I stopped, not knowing how to progress further). Why is it even required in the first place no matter what? |
Tracking the basefw part in #8391 |
No assigned person and one week to go before branch point, kicking to v2.9. |
@andyross @cujomalainey we will tackle this in Q1, there is ongoing work already to split IPC logic out of modules as 1st step so we have module-common.c, module-ipc3.c and module-ipc4.c. We can then move this into other places so that common codebase is generic. |
Stable-v2.9 branched, this didn't make the cut, bumping to 2.10. |
With #9060 merged, we can close this. All vendor specific code has been now removed. Only exception is SOF_TELEMETRY, but that is behind separate ifdefs. |
Recent IPC4 changes have been introducing Intel DSP details into the generic SOF source layer.
The src/audio/base_fw.c file (despite the name, this is an IPC4-only source file) is implementing a protocol that seems directly related to Intel hardware details. In particular the handling in basefw_mem_state_info() is doing direct register reads to retrieve hardware state. It seems like there's a facility in place to structure the output of this function via "tuple" key/data pairs, but there's no documentation on the format it nor indirection API available for platforms.
Similarly the component driver module handling in ipc4/helper.c:ipc4_get_comp_drv() assumes that there is a Intel rimage-generated manifest format in memory from which to load modules. There is a "LIBRARY_MANAGER" feature that seems to be intended to replace this, I think? As of right now neither mechanism is in use in the main tree.
All hardware-specific features like this need to be provided behind an API that allows for other platforms to redefine it as needed, and ideally be defined/documented in a way that makes as few assumptions as possible about the underlying hardware.
See also comments and commit messages is PR #7531
The text was updated successfully, but these errors were encountered: