AMD Sensor Fusion Hub Laptop Driver Unlikely To Land For Linux 5.7

[phoronix]

Phoronix: AMD Sensor Fusion Hub Laptop Driver Unlikely To Land For Linux 5.7

While we were hoping to see the AMD Sensor Fusion Hub driver introduced in Linux 5.7 for improving the AMD Ryzen Linux laptop experience, that now looks quite unlikely...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-SHF-Driver-Needs-Work

[Adarion]

Reminds me of the time when I was reviewing student's BSc theses, protocols and the like. Or common publications for the professional journals.
If one looks closely you're always bound to find something.
Sometimes I wonder what mistakes I might find when I look back at my own works of qualification...

Let's still hope this driver get polished soon and goes mainline.

[ddriver]

Oh wow, intel employees sandbagging amd products... unprecedented.

[TemplarGR]

Oh wow, intel employees sandbagging amd products... unprecedented.

Yeah. I am pretty sure if one actually took the time he could find similar errors on their stuff but of course no one bothers.

[ddriver]

I've spent enough time in the industry to know duckhead nitpicking when I see it, and that's like 95% of the alleged "review".

[Azultra]

I've spent enough time in the industry to know duckhead nitpicking when I see it, and that's like 95% of the alleged "review".

That's most of the reviews for Linux patches, telling to use the proper kernel functions/macros, minimizing redundances, using the the proper formatting (which this patch doesn't follow at all)

Since the reviewer is not familiar with the hardware he's missing the elephants in the room for this series of patches : the authors are duplicating the MP2 pci driver from i2c-amd-mp2, and even using their own inferior unreviewed version that was far from acceptance back then, before it was revamped by an external contributor.

[finalzone]

Since the reviewer is not familiar with the hardware he's missing the elephants in the room for this series of patches : the authors are duplicating the MP2 pci driver from i2c-amd-mp2, and even using their own inferior unreviewed version that was far from acceptance back then, before it was revamped by an external contributor.

As one of HP Envy Ryzen 2500u convertible owners, it will be great someone does a cross examination of the reviews at the Linux kernel mailing list. Testing the patches showed the gyroscope was not even detected and used at all.

[ernstp]

It might be a mix between somewhat unnecessary nitpicking and very well founded feedback. ;-)

[schard]

As I noted on the respective bug report, the problem with the current driver is that on Ryzen 5 2500U (and possibly other) systems, the BIOS or firmware does not correctly provide the SFH's register

AMD_P2C_MSG3

, which stores a bitmask for the enabled sensors. Consequently on these systems none of the connected sensors are being activated.
With my linked patch, which I also provided on patchwork.kernel.org you can get the accelerometer on these systems working by loading the driver with

force_accel=1

.
Sandeep Singh, the main deleveloper of the SFH patch, already replied to me via email, saying that these module parameters will be taken into the next revision of the patch series.

[finalzone]

As I noted on the respective bug report, the problem with the current driver is that on Ryzen 5 2500U (and possibly other) systems, the BIOS or firmware does not correctly provide the SFH's register

AMD_P2C_MSG3

, which stores a bitmask for the enabled sensors. Consequently on these systems none of the connected sensors are being activated.

Sadly the provided patch failed to apply on the latest git snapshot of the kernel. Would you mind fixing it?