On 03.05.2024 16:39, Pali Rohár wrote: > On Friday 03 May 2024 16:09:03 Heiner Kallweit wrote: >> So far each new Dell device with an accelerometer requires a patch. >> All devices, with one older system as an exception, use address 0x29. >> So I think we can safely scan for the correct address > > This is too risky. Poking random smbus address can lead to the lockup or > crash of the whole system. I really dislike such change which is going > to unconditionally on every system to scan or access fixed smbus address. > Crashing kernel on some new model in future or on some older models is > the worst user experience which can happen. > The scan is protected by is_dell_system_with_lis3lv02d(). So we know it's a Dell system with an accelerometer. The only potentially problematic scenario would be: - It's a Dell system with an accelerometer and - Accelerometer is at an address different from 0x29 and - System has some other device at address 0x29 If there ever should be such a case, we still would have the option to provide a more specific probe function to i2c_new_scanned_device(). At least for now I don't really see a risk. > The hardcoded table with model name and address is the safest option to > prevent crashes or other unexpected behavior. > > Instead, I would suggest to contact Dell if they can provide a way to > read accelerometer's i2c/smbus address from ACPI/WMI/DMI. And if there See comment in the driver: "Accelerometer's I2C address is not specified in DMI nor ACPI" > is not any way right now then ask Dell to include it for new products. > So we can avoid having hardcoded table for new products released in > future. This is a good point. Not sure whether anybody tried this yet. Therefore +Dell.Client.Kernel@xxxxxxxx
