netdev - Re: [PATCH 01/12] dt-bindings: riscv: sifive-ccache: Add compatible for StarFive JH7100 SoC

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]

Date:   Mon, 20 Mar 2023 16:46:28 -0700 (PDT)
From:   Palmer Dabbelt <palmer@...belt.com>
To:     Conor Dooley <conor@...nel.org>
CC:     cristian.ciocaltea@...labora.com, lee@...nel.org,
        robh+dt@...nel.org, krzysztof.kozlowski+dt@...aro.org,
        davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org,
        pabeni@...hat.com, kernel@...il.dk,
        Paul Walmsley <paul.walmsley@...ive.com>,
        aou@...s.berkeley.edu, peppe.cavallaro@...com,
        alexandre.torgue@...s.st.com, joabreu@...opsys.com,
        mcoquelin.stm32@...il.com, richardcochran@...il.com,
        sagar.kadam@...ive.com, yanhong.wang@...rfivetech.com,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        netdev@...r.kernel.org, linux-riscv@...ts.infradead.org,
        linux-stm32@...md-mailman.stormreply.com,
        linux-arm-kernel@...ts.infradead.org, kernel@...labora.com
Subject:     Re: [PATCH 01/12] dt-bindings: riscv: sifive-ccache: Add compatible for StarFive JH7100 SoC

On Tue, 14 Feb 2023 12:40:35 PST (-0800), Conor Dooley wrote:
> Hey all,
>
> On Sat, Feb 11, 2023 at 05:18:10AM +0200, Cristian Ciocaltea wrote:
>> Document the compatible for the SiFive Composable Cache Controller found
>> on the StarFive JH7100 SoC.
>> 
>> This also requires extending the 'reg' property to handle distinct
>> ranges, as specified via 'reg-names'.
>> 
>> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@...labora.com>
>> ---
>>  .../bindings/riscv/sifive,ccache0.yaml        | 28 ++++++++++++++++++-
>>  1 file changed, 27 insertions(+), 1 deletion(-)
>> 
>> diff --git a/Documentation/devicetree/bindings/riscv/sifive,ccache0.yaml b/Documentation/devicetree/bindings/riscv/sifive,ccache0.yaml
>> index 31d20efaa6d3..2b864b2f12c9 100644
>> --- a/Documentation/devicetree/bindings/riscv/sifive,ccache0.yaml
>> +++ b/Documentation/devicetree/bindings/riscv/sifive,ccache0.yaml
>> @@ -25,6 +25,7 @@ select:
>>            - sifive,ccache0
>>            - sifive,fu540-c000-ccache
>>            - sifive,fu740-c000-ccache
>> +          - starfive,jh7100-ccache
>>  
>>    required:
>>      - compatible
>> @@ -37,6 +38,7 @@ properties:
>>                - sifive,ccache0
>>                - sifive,fu540-c000-ccache
>>                - sifive,fu740-c000-ccache
>> +              - starfive,jh7100-ccache
>>            - const: cache
>>        - items:
>>            - const: starfive,jh7110-ccache
>> @@ -70,7 +72,13 @@ properties:
>>        - description: DirFail interrupt
>>  
>>    reg:
>> -    maxItems: 1
>> +    minItems: 1
>> +    maxItems: 2
>> +
>> +  reg-names:
>> +    items:
>> +      - const: control
>> +      - const: sideband
>
> So why is this called "sideband"?
> In the docs for the JH7100 it is called LIM & it's called LIM in our
> docs for the PolarFire SoC (at the same address btw) and we run the HSS

IIRC it's both: "LIM" is the memory, "sideband" is the port.  I can't 
find any proper documentation of "sideband" outside of DT and errata, 
but there's a hanful of references to it in the bootloader for the 
fu540: 
<https://github.com/sifive/freedom-u540-c000-bootloader/search?q=sideband>.

It's not really clear which is more correct here: sideband accesses are 
only useful when the cache is configured as an LIM, at least for general 
software.  IIRC the accesses to the LIM only go through the sideband 
port for the E core, but I might be wrong about that.

> out of it! LIM being "loosely integrated memory", which by the limit
> hits on Google may be a SiFive-ism?

Yep: TIM is the SiFive version of Arm's TCM (tightly coupled memory), 
and LIM is the flavor that's farther away (L2 instead of L1).

> I'm not really sure if adding it as a "reg" section is the right thing
> to do as it's not "just" a register bank.
> Perhaps Rob/Krzysztof have a take on that one?
>
>>  
>>    next-level-cache: true
>>  
>> @@ -89,6 +97,7 @@ allOf:
>>            contains:
>>              enum:
>>                - sifive,fu740-c000-ccache
>> +              - starfive,jh7100-ccache
>>                - starfive,jh7110-ccache
>>                - microchip,mpfs-ccache
>>  
>> @@ -106,12 +115,29 @@ allOf:
>>              Must contain entries for DirError, DataError and DataFail signals.
>>            maxItems: 3
>>  
>> +  - if:
>> +      properties:
>> +        compatible:
>> +          contains:
>> +            const: starfive,jh7100-ccache
>> +
>> +    then:
>> +      properties:
>> +        reg:
>> +          maxItems: 2
>> +
>> +    else:
>> +      properties:
>> +        reg:
>> +          maxItems: 1
>> +
>>    - if:
>>        properties:
>>          compatible:
>>            contains:
>>              enum:
>>                - sifive,fu740-c000-ccache
>> +              - starfive,jh7100-ccache
>>                - starfive,jh7110-ccache
>>  
>>      then:
>> -- 
>> 2.39.1
>>