linux-kernel - Re: [PATCH 7/8] dt-bindings: dma: qcom: bam-dma: Add missing required properties

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d674d626-e6a3-4683-8f45-81b09200849f@oss.qualcomm.com>
Date: Thu, 13 Feb 2025 15:00:00 +0100
From: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
To: Stephan Gerhold <stephan.gerhold@...aro.org>,
        Konrad Dybcio <konrad.dybcio@....qualcomm.com>
Cc: Vinod Koul <vkoul@...nel.org>, Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>, Rob Herring <robh@...nel.org>,
        Krzysztof Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley
 <conor+dt@...nel.org>, Andy Gross <agross@...nel.org>,
        Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Neil Armstrong <neil.armstrong@...aro.org>,
        Yuvaraj Ranganathan <quic_yrangana@...cinc.com>,
        Anusha Rao <quic_anusha@...cinc.com>,
        Md Sadre Alam
 <quic_mdalam@...cinc.com>,
        linux-arm-msm@...r.kernel.org, dmaengine@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        Luca Weiss <luca.weiss@...rphone.com>
Subject: Re: [PATCH 7/8] dt-bindings: dma: qcom: bam-dma: Add missing required
 properties

On 13.02.2025 10:13 AM, Stephan Gerhold wrote:
> On Wed, Feb 12, 2025 at 10:01:59PM +0100, Konrad Dybcio wrote:
>> On 12.02.2025 6:03 PM, Stephan Gerhold wrote:
>>> num-channels and qcom,num-ees are required when there are no clocks
>>> specified in the device tree, because we have no reliable way to read them
>>> from the hardware registers if we cannot ensure the BAM hardware is up when
>>> the device is being probed.
>>>
>>> This has often been forgotten when adding new SoC device trees, so make
>>> this clear by describing this requirement in the schema.
>>>
>>> Signed-off-by: Stephan Gerhold <stephan.gerhold@...aro.org>
>>> ---
>>>  Documentation/devicetree/bindings/dma/qcom,bam-dma.yaml | 4 ++++
>>>  1 file changed, 4 insertions(+)
>>>
>>> diff --git a/Documentation/devicetree/bindings/dma/qcom,bam-dma.yaml b/Documentation/devicetree/bindings/dma/qcom,bam-dma.yaml
>>> index 3ad0d9b1fbc5e4f83dd316d1ad79773c288748ba..5f7e7763615578717651014cfd52745ea2132115 100644
>>> --- a/Documentation/devicetree/bindings/dma/qcom,bam-dma.yaml
>>> +++ b/Documentation/devicetree/bindings/dma/qcom,bam-dma.yaml
>>> @@ -90,8 +90,12 @@ required:
>>>  anyOf:
>>>    - required:
>>>        - qcom,powered-remotely
>>> +      - num-channels
>>> +      - qcom,num-ees
>>>    - required:
>>>        - qcom,controlled-remotely
>>> +      - num-channels
>>> +      - qcom,num-ees
>>
>> I think I'd rather see these deprecated and add the clock everywhere..
>> Do we know which one we need to add on newer platforms? Or maybe it's
>> been transformed into an icc path?
> 
> This isn't feasible, there are too many different setups. Also often the
> BAM power management is tightly integrated into the consumer interface.
> To give a short excerpt (I'm sure there are even more obscure uses):
> 
>  - BLSP BAM (UART, I2C, SPI on older SoCs):
>     1. Enable GCC_BLSP_AHB_CLK
>     -> This is what the bam_dma driver supports currently.
> 
>  - Crypto BAM: Either
>     OR 1. Vote for single RPM clock
>     OR 1. Enable 3 separate clocks (CE, CE_AHB, CE_AXI)
>     OR 1. Vote dummy bandwidth for interconnect
> 
>  - BAM DMUX (WWAN on older SoCs):
>     1. Start modem firmware
>     2. Wait for BAM DMUX service to be up
>     3. Vote for power up via the BAM-DMUX-specific SMEM state
>     4. Hope the firmware agrees and brings up the BAM
> 
>  - SLIMbus BAM (audio on some SoCs):
>     1. Start ADSP firmware
>     2. Wait for QMI SLIMBUS service to be up via QRTR
>     3. Vote for power up via SLIMbus-specific QMI messages
>     4. Hope the firmware agrees and brings up the BAM
> 
> Especially for the last two, we can't implement support for those
> consumer-specific interfaces in the BAM driver. Implementing support for
> the 3 variants of the Crypto BAM would be possible, but it's honestly
> the least interesting use case of all these. It's not really clear why
> we are bothing with the crypto engine on newer SoCs at all, see e.g. [1].
> 
> [1]: https://lore.kernel.org/linux-arm-msm/20250118080604.GA721573@sol.localdomain/
> 
>> Reading back things from this piece of HW only to add it to DT to avoid
>> reading it later is a really messy solution.
> 
> In retrospect, it could have been cleaner to avoid describing the BAM as
> device node independent of the consumer. We wouldn't have this problem
> if the BAM driver would only probe when the consumer is already ready.
> 
> But I think specifying num-channels in the device tree is the cleanest
> way out of this mess. I have a second patch series ready that drops
> qcom,num-ees and validates the num-channels once it's safe reading from
> the BAM registers. That way, you just need one boot test to ensure the
> device tree description is really correct.

Thanks for the detailed explanation!

Do you think it could maybe make sense to expose a clock/power-domain
from the modem/adsp rproc and feed it to the DMUX / SLIM instances when
an appropriate ping arrives? This way we'd also defer probing the drivers
until the device is actually accessible.

Konrad