Re: [PATCH v4 00/11] Add spi-hid transport driver

[fQwQf]

Hi Jingyuan,


> This series picks up the spi-hid driver work originally started by
> Microsoft. The patch breakdown has been modified and the implementation
> has been refactored to address upstream feedback and testing issues. We
> are submitting this as a new series while keeping the original sign-off
> chain to reflect the history.


I am working on Linux support for the Surface Laptop 7 (13.8-inch, Snapdragon X series), whose touchpad is a HID-over-SPI device behind a Qualcomm GENI QSPI controller. I noticed that v4 (June 9) is the
latest revision of this series and has so far only seen automated review feedback, so I would like to coordinate before preparing any upstream submission of my own.

Current state on my side:

- I have a working touchpad using a downstream Qualcomm GENI QSPI +  spi-hid stack (originally from scuggo's x1e-nixos work, imported  via ELLX-Kernel). My local adaptations move the transport to spi-mem and add framing, response matching, and DMA/error-path  hardening.
- Normal touchpad use works on my machine. The latest hardening currently has only build and software-test coverage; I have not validated s.
2. The SL7 adds a different transport requirement (quad-SPI via GENI, through spi-mem) on topuspend/resume, and I have not yet run your v4 series on this hardware.

I believe our work may complement each other in two ways:

1. The automated review of the series raised DMA cacheline-alignment concerns for SPI transfer buffers and unchecked reset return values in the init/resume paths. Both overlap with the hardening I have been doing downstream, and I would be glad to contribute fixes there.
2. The SL7 adds a different transport requirement (quad-SPI via GENI, through spi-mem) on top of the same HID-over-SPI protocol, which looks like a natural fit for your generic driver rather than a standalone one.

Questions:

- What tree or series do you recommend working against? Is v4 still your current baseline, or do you have a newer development branch?
- Would adding the SL7 quad-SPI transport requirements to your generic driver be the preferred upstream approach? Is anyone already working on this?
- Are you waiting on maintainer review before a v5? A second hardware platform may help move things along, and I am happy to provide testing on SL7.

I am also tracing the downstream provenance of the Qualcomm QSPI code with the original authors to ensure a clean sign-off chain; I appreciate that this series handles its own history the same way.

Happy to share technical details or a preliminary diff if useful.

I'm not subscribed to the lists; please keep me in CC.

Best regards,
Jizhou Tong