[RFC PATCH] Add bridge driver to connect sensors to CIO2 device via software nodes on ACPI platforms

Daniel Scally djrscally at gmail.com
Thu Sep 17 21:25:03 UTC 2020

On 17/09/2020 15:14, Andy Shevchenko wrote:
> On Thu, Sep 17, 2020 at 4:53 PM Dan Scally <djrscally at gmail.com> wrote:
>> Hi Andy, thanks for input (as always)
> You're welcome! I'm really impressed by your activity in this area.
Thanks - it's pretty fun so far
>>> Ah, I think you misinterpreted the meaning of above. The above is a switch how
>>> camera device appears either as PCI or an ACPI. So, it effectively means you
>>> should *not* have any relation for this HID until you find a platform where the
>>> device is for real enumerated via ACPI.
>> Ah, ok. So that was never going to work. Thanks. That does raise another
>> question; we have had some testers report failure, which turns out to be
>> because on their platforms the definition of their cameras in ACPI is
>> never translated into an i2c_client so the cio2-bridge doesn't bind.
>> Those have a similar conditional in the _STA method, see CAM1 in this
>> DSDT for example:
>> https://raw.githubusercontent.com/linux-surface/acpidumps/master/surface_go/dsdt.dsl.
>> Is there anything we can do to enable those cameras to be discovered too?
> It means that this
Is the rest of this comment missing?
>>>>>> +#define PROPERTY_ENTRY_NULL                       \
>>>>>> +((const struct property_entry) { })
>>>>> Alignment. Same appears to apply to other macros (please indent).
>>>> Yep
>>>>>> +#define SOFTWARE_NODE_NULL                        \
>>>>>> +((const struct software_node) { })
>>> Why?!
>> It felt ugly to have the other definitions be macros and not this one,
>> but I can change it.
> My point is that those macros are simply redundant. The point is to
> have a terminator record (all 0:s in the last entry of an array) which
> is usually being achieved by allocating memory with kcalloc() which
> does implicitly this for you.
Ah I see. TIL - thanks, I'll make that change too.

More information about the devel mailing list