Re: [PATCH RFC 01/22] dt-bindings: usb: snps,dwc3: Allow multiple iommus

From: Krzysztof Kozlowski <krzk@kernel.org>
To: Sven Peter <sven@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Rob Herring <robh@kernel.org>,
	Krzysztof Kozlowski <krzk+dt@kernel.org>,
	Conor Dooley <conor+dt@kernel.org>,
	Felipe Balbi <balbi@kernel.org>, Janne Grunau <j@jannau.net>,
	Alyssa Rosenzweig <alyssa@rosenzweig.io>,
	Neal Gompa <neal@gompa.dev>, Vinod Koul <vkoul@kernel.org>,
	Kishon Vijay Abraham I <kishon@kernel.org>,
	Thinh Nguyen <Thinh.Nguyen@synopsys.com>,
	Heikki Krogerus <heikki.krogerus@linux.intel.com>,
	Philipp Zabel <p.zabel@pengutronix.de>,
	linux-usb@vger.kernel.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org, asahi@lists.linux.dev,
	linux-arm-kernel@lists.infradead.org,
	linux-phy@lists.infradead.org
Subject: Re: [PATCH RFC 01/22] dt-bindings: usb: snps,dwc3: Allow multiple iommus
Date: Thu, 28 Aug 2025 08:54:18 +0200	[thread overview]
Message-ID: <e8934b5b-ef75-4ee0-b1e3-cc5db72b99ec@kernel.org> (raw)
In-Reply-To: <a89ba427-b174-42c3-b3b4-5e4079942c3c@kernel.org>

On 27/08/2025 18:07, Sven Peter wrote:
> On 24.08.25 10:31, Krzysztof Kozlowski wrote:
>> On 22/08/2025 09:22, Krzysztof Kozlowski wrote:
>>> On Thu, Aug 21, 2025 at 03:38:53PM +0000, Sven Peter wrote:
>>>> Apple's dwc3 variant requires two iommus.
>>>>
>>>> Signed-off-by: Sven Peter <sven@kernel.org>
>>>> ---
>>>>   Documentation/devicetree/bindings/usb/snps,dwc3.yaml | 2 +-
>>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
>>>> index 4380bb6fa2f08a475b557e80467abc2861f622e2..6d35dcc605c01977f4fc6fdb6f12976f1cef2b9e 100644
>>>> --- a/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
>>>> +++ b/Documentation/devicetree/bindings/usb/snps,dwc3.yaml
>>>> @@ -60,7 +60,7 @@ properties:
>>>>     dma-coherent: true
>>>>   
>>>>     iommus:
>>>> -    maxItems: 1
>>>> +    maxItems: 2
>>>
>>> Never tested. You anyway need specific constraints.
>> I realized that's pretty vague, so clarifying: you changed all bindings
>> to have two iommus and that's unexpected. All other devices have only
>> one IOMMU mapping here, so they need to stay like that. If these IOMMUs
>> differ, you should also list them, but probably that's not the case here.
> 
> Makes sense, thanks for the detailed explanation.
> 
> I don't quite understand why this controller needs two iommus but they 
> must essentially be programmed to the same mapping for xhci to work.

That's fine.

> All of device mode goes through one of them and xhci buffers are split 
> between the two iommus (iirc I saw the command ring using the first one 
> and the event ring the second one and then just decided to program them 
> the same since anything else would result in a big mess).

Sure. But if you change existing bindings (existing devices) allowing
them to have two IOMMUs, this must be explicitly mentioned in commit msg
with some sort of rationale (even if rationale was "we just missed that").

Best regards,
Krzysztof

-- 
linux-phy mailing list
linux-phy@lists.infradead.org
https://lists.infradead.org/mailman/listinfo/linux-phy