Message-ID: <dd381dc1-454f-4ecd-adb7-55de2e15d592@broadcom.com>
Date: Wed, 31 Jul 2024 14:57:37 +0200
From: Arend van Spriel <arend.vanspriel@...adcom.com>
To: Sebastian Reichel <sebastian.reichel@...labora.com>
Cc: Krzysztof Kozlowski <krzk@...nel.org>,
 Jacobe Zang <jacobe.zang@...ion.com>, robh@...nel.org, krzk+dt@...nel.org,
 heiko@...ech.de, kvalo@...nel.org, davem@...emloft.net, edumazet@...gle.com,
 kuba@...nel.org, pabeni@...hat.com, conor+dt@...nel.org,
 Linus Walleij <linus.walleij@...aro.org>, efectn@...tonmail.com,
 dsimic@...jaro.org, jagan@...eble.ai, devicetree@...r.kernel.org,
 linux-arm-kernel@...ts.infradead.org, linux-rockchip@...ts.infradead.org,
 linux-kernel@...r.kernel.org, arend@...adcom.com,
 linux-wireless@...r.kernel.org, netdev@...r.kernel.org, megi@....cz,
 duoming@....edu.cn, bhelgaas@...gle.com, minipli@...ecurity.net,
 brcm80211@...ts.linux.dev, brcm80211-dev-list.pdl@...adcom.com,
 nick@...das.com, Andy Green <andy@...mcat.com>
Subject: Re: [PATCH v5 2/5] dt-bindings: net: wireless: brcm4329-fmac: add
 clock description for AP6275P

On 7/30/2024 7:38 PM, Sebastian Reichel wrote:
> Hi,
> 
> On Tue, Jul 30, 2024 at 01:16:57PM GMT, Arend Van Spriel wrote:
>> On July 30, 2024 12:18:20 PM Krzysztof Kozlowski <krzk@...nel.org> wrote:
>>
>>> On 30/07/2024 11:52, Arend Van Spriel wrote:
>>>> On July 30, 2024 11:01:43 AM Krzysztof Kozlowski <krzk@...nel.org> wrote:
>>>>
>>>>> On 30/07/2024 08:37, Arend Van Spriel wrote:
>>>>>> + Linus W
>>>>>>
>>>>>> On July 30, 2024 5:31:15 AM Jacobe Zang <jacobe.zang@...ion.com> wrote:
>>>>>>
>>>>>>> Not only AP6275P Wi-Fi device but also all Broadcom wireless devices allow
>>>>>>> external low power clock input. In DTS the clock as an optional choice in
>>>>>>> the absence of an internal clock.
>>>>>>>
>>>>>>> Reviewed-by: Arend van Spriel <arend.vanspriel@...adcom.com>
>>>>>>> Signed-off-by: Jacobe Zang <jacobe.zang@...ion.com>
>>>>>>> ---
>>>>>>> .../bindings/net/wireless/brcm,bcm4329-fmac.yaml          | 8 ++++++++
>>>>>>> 1 file changed, 8 insertions(+)
>>>>>>>
>>>>>>> diff --git
>>>>>>> a/Documentation/devicetree/bindings/net/wireless/brcm,bcm4329-fmac.yaml
>>>>>>> b/Documentation/devicetree/bindings/net/wireless/brcm,bcm4329-fmac.yaml
>>>>>>> index 2c2093c77ec9a..a3607d55ef367 100644
>>>>>>> --- a/Documentation/devicetree/bindings/net/wireless/brcm,bcm4329-fmac.yaml
>>>>>>> +++ b/Documentation/devicetree/bindings/net/wireless/brcm,bcm4329-fmac.yaml
>>>>>>> @@ -122,6 +122,14 @@ properties:
>>>>>>> NVRAM. This would normally be filled in by the bootloader from platform
>>>>>>> configuration data.
>>>>>>>
>>>>>>> +  clocks:
>>>>>>> +    items:
>>>>>>> +      - description: External Low Power Clock input (32.768KHz)
>>>>>>> +
>>>>>>> +  clock-names:
>>>>>>> +    items:
>>>>>>> +      - const: lpo
>>>>>>> +
>>>>>>
>>>>>> We still have an issue that this clock input is also present in the
>>>>>> bindings specification broadcom-bluetooth.yaml (not in bluetooth
>>>>>> subfolder). This clock is actually a chip resource. What happens if both
>>>>>> are defined and both wifi and bt drivers try to enable this clock? Can this
>>>>>> be expressed in yaml or can we only put a textual warning in the property
>>>>>> descriptions?
>>>>>
>>>>> Just like all clocks, what would happen? It will be enabled.
>>>>
>>>> Oh, wow! Cool stuff. But seriously is it not a problem to have two entities
>>>> controlling one and the same clock? Is this use-case taken into account by
>>>> the clock framework?
>>>
>>> Yes, it is handled correctly. That's a basic use-case, handled by CCF
>>> since some years (~12?). Anyway, whatever OS is doing (or not doing)
>>> with the clocks is independent of the bindings here. The question is
>>
>> Agree. Probably the bindings would not be the place to document this if it
>> would be an issue.
>>
>>> about hardware - does this node, which represents PCI interface of the
>>> chip, has/uses the clocks.
>>
>> The schematics I found for the wifi module and the khadas edge platform show
>> these are indeed wired to the chip.
> 
> I have a Rockchip RK3588 Evaluation Board on my desk, which uses the
> same WLAN AP6275P module. I think I already commented on a prior
> version of this series: The LPO clock is needed to make the PCIe
> device visible on the bus. That means this series only works if the
> clock has already been running. Otherwise the PCIe driver will never
> be probed. To become visible the devices requires:
> 
> 1. The LPO clock to be enabled
> 2. Power to be applied
> 3. The WL_EN gpio to be configured correctly
> 
> If one of the above is not met, the device will not even appear in
> 'lspci'. I believe the binding needs to take into consideration, that
> pwrseq is needed for the PCIe side. Fortuantely the heavy lifting of
> creating the proper infrastructure for this has already been done by
> Bartosz Golaszewski for Qualcomm WLAN chips. What is missing is a
> pwrseq driver for the Broadcom chip (or this specific module?).

That does not really make sense. There is no relation between the LPO 
clock and the PCIe clocks so 1) being a requirement for probing the 
device looks odd. It also does not match past experience when I assisted 
Andy Green in getting this module up and running almost two years ago.

"""
On 11/9/22 18:26, Arend Van Spriel wrote:
 > On November 8, 2022 11:48:22 AM Andy Green <andy@...mcat.com> wrote:
 >> Hi -
 >>
 >> I'm trying to bring up AP6275 support on 6.1-rc4... I have tried a 
forward-ported sdk broadcom driver from the 5.10 based soc sdk, and the 
mainline brcm fullmac driver.
 >
 > Do you have a reference to the SDK? For what SoC?

Hi Arend -

It's the OOT broadcom driver that came with the latest (Sept 2022) 
vendor SDK for RK3588, from Rockchip.  Their evb has an AP6275 onboard.

PCIe generally is working on this (eg, for NVMe in the PCIe 4-lane slot) 
and for network, and the PCIe part seems OK when I hack in a gpio 
regulator to hold up the module enable gpio.
"""

So regarding 2) and 3) I agree with you.

Regards,
Arend

Powered by blists - more mailing lists