On Mon, 2019-12-16 at 12:05 +0000, Paul Olaru (OSS) wrote:
Hello, I would like panics on our DSP to actually be properly reported to the kernel.
Currently on the Intel platforms for panic there is custom code for each platform to e.g. read from the shim area and check with the panic magic mask whether there is a panic.
My suggestion is to instead craft a specific message type and a very short message (possibly only that one word with the panic mask, which will not be confused with a regular DSP request header) which is then handled in a generic fashion within snd_sof_ipc_msgs_rx.
Ok, so you mean have a global IPC type for panic and send as regular IPC ?
I am aware that such a change will affect several platforms so I'm currently just raising the issue as I want to know if it's a good idea before starting work on it.
We do have partial support for this on Intel platforms but we also have the side channel of using other shared IO areas for panic information.
Also, does this qualify as a minor ABI bump?
Yes.
I would also go further here and invoke a polling GDB stuff (Kconfig option - with all IRQs OFF) after the panic IPC. This way the user could connect GDB for further debug.
Liam
Sound-open-firmware mailing list Sound-open-firmware@alsa-project.org https://mailman.alsa-project.org/mailman/listinfo/sound-open-firmware