Re: [PATCH] Revert "Bluetooth: hci_qca: Set BDA quirk bit if fwnode exists in DT"

[Linux regression tracking (Thorsten Leemhuis)]

Bluetooth Maintainers, what's...

On 14.03.24 16:07, Johan Hovold wrote:
> On Thu, Mar 14, 2024 at 10:30:36AM -0400, Luiz Augusto von Dentz wrote:
>> On Thu, Mar 14, 2024 at 4:44 AM Johan Hovold <johan+linaro@xxxxxxxxxx> wrote:
> 
>>> This reverts commit 7dcd3e014aa7faeeaf4047190b22d8a19a0db696.
>>>
>>> Qualcomm Bluetooth controllers like WCN6855 do not have persistent
>>> storage for the Bluetooth address and must therefore start as
>>> unconfigured to allow the user to set a valid address unless one has
>>> been provided by the boot firmware in the devicetree.
>>>
>>> A recent change snuck into v6.8-rc7 and incorrectly started marking the
>>> default (non-unique) address as valid. This specifically also breaks the
>>> Bluetooth setup for some user of the Lenovo ThinkPad X13s.
>>>
>>> Note that this is the second time Qualcomm breaks the driver this way
>>> and that this was fixed last year by commit 6945795bc81a ("Bluetooth:
>>> fix use-bdaddr-property quirk"), which also has some further details.
>>>
>>> Fixes: 7dcd3e014aa7 ("Bluetooth: hci_qca: Set BDA quirk bit if fwnode exists in DT")
>>> Cc: stable@xxxxxxxxxxxxxxx      # 6.8
>>> Cc: Janaki Ramaiah Thota <quic_janathot@xxxxxxxxxxx>
>>> Signed-off-by: Johan Hovold <johan+linaro@xxxxxxxxxx>
>>
>> Well I guess I will need to start asking for evidence that this works
>> on regular Linux distros then, because it looks like that is not the
>> environment Janaki and others Qualcomm folks are testing with.
>>
>> What I probably would consider as evidence is bluetoothd logs showing
>> that the controller has been configured correctly or perhaps there is
>> a simpler way?
> 
> Well, in this case we actually want the controller to remain
> unconfigured (e.g. to avoid having every user of the X13s unknowingly
> use the same default address). 
> 
> I'm not sure why Qualcomm insists on breaking these quirks, but I guess
> they just haven't understood why they exist. It's of course convenient
> to be able to use the default address during development without first
> having to provide an address, but that's not a valid reason to break the
> driver.
> 
> From what I hear the Qualcomm developers only care about Android and I
> believe they have some out-of-tree hack for retrieving the device
> address directly from the rootfs.
> 
> For the X13s, and as I think I've mentioned before, we have been trying
> to get Qualcomm to tell us how to access the assigned addresses that are
> stored in some secure world storage so that we can set it directly from
> the driver. But until we figure that out, users will need to continue
> setting the address manually.

..the plan forward here? This to me sounds like a case where a quick
revert is the right (interim?) solution, but nevertheless nothing
happened for ~10 days now afaics. Or am I missing something?

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
--
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
If I did something stupid, please tell me, as explained on that page.

#regzbot poke