linux-kernel - Re: [PATCH v2 1/3] soc: qcom: llcc: Pass SoC specific EDAC register offsets to EDAC driver

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:   Wed, 24 Aug 2022 10:43:51 +0530
From:   Sai Prakash Ranjan <quic_saipraka@...cinc.com>
To:     Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>
CC:     <bjorn.andersson@...aro.org>, <bp@...en8.de>, <mchehab@...nel.org>,
        <james.morse@....com>, <rric@...nel.org>,
        <linux-arm-msm@...r.kernel.org>, <linux-edac@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <quic_tsoni@...cinc.com>
Subject: Re: [PATCH v2 1/3] soc: qcom: llcc: Pass SoC specific EDAC register
 offsets to EDAC driver

On 8/23/2022 9:01 PM, Manivannan Sadhasivam wrote:
> On Mon, Aug 22, 2022 at 05:29:13PM +0530, Sai Prakash Ranjan wrote:
>> Hi Mani,
>>
>> On 8/12/2022 11:36 AM, Manivannan Sadhasivam wrote:
>>> The LLCC EDAC register offsets varies between each SoCs. Until now, the
>>> EDAC driver used the hardcoded register offsets. But this caused crash
>>> on SM8450 SoC where the register offsets has been changed.
>>>
>>> So to avoid this crash and also to make it easy to accomodate changes for
>>> new SoCs, let's pass the SoC specific register offsets to the EDAC driver.
>>>
>>> Currently, two set of offsets are used. One is SM8450 specific and another
>>> one is common to all SoCs.
>>>
>>> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>
>> <snip> ...
>>
>>>    static const struct qcom_llcc_config sm8350_cfg = {
>>> @@ -309,6 +370,7 @@ static const struct qcom_llcc_config sm8350_cfg = {
>>>    	.size           = ARRAY_SIZE(sm8350_data),
>>>    	.need_llcc_cfg	= true,
>>>    	.reg_offset	= llcc_v1_2_reg_offset,
>>> +	.edac_reg	= &common_edac_reg,
>>>    };
>>>    static const struct qcom_llcc_config sm8450_cfg = {
>>> @@ -316,6 +378,7 @@ static const struct qcom_llcc_config sm8450_cfg = {
>>>    	.size           = ARRAY_SIZE(sm8450_data),
>>>    	.need_llcc_cfg	= true,
>>>    	.reg_offset	= llcc_v21_reg_offset,
>>> +	.edac_reg	= &sm8450_edac_reg,
>>>    };
>>>
>> Can we have LLCC version specific register offsets instead of SoC specific similar to reg_offset callbacks?
>> For SM8450, it would be llcc_v21_edac_reg and for others llcc_v1_2_edac_reg instead of common_edac_reg.
>> common_edac_reg is very general and is not exactly common for all, its just common for SoCs with same LLCC.
>>
> I thought about it but I was not sure if rest of the SoCs are using version
> v1.2. I know that reg_offset uses v1.2 but I was skeptical and hence used the
> SoC specific offsets.
>
> Can you confirm if rest of the SoCs are using v1.2?

LLCC versioning follows w.x.y.z format and w and y are major and minor versions based
on which the naming for reg_offsets is chosen.

Now in above reg_offsets, llcc_v1_2 is not v1.2, it means v1.0 or v2.0 where 1, 2 is a major version
and 0 is a minor version. llcc_v21 is actually v2.1 where 2 is a major and 1 is a minor version.
I know the naming is pretty bad, should probably replace llcc_v1_2 with llcc_v1_0_v2_0 and
llcc_v21 with llcc_v2_1? Note here minor version is important because SM8350 is v2.0 and uses
old reg offsets.

So coming to your query now, all other SoCs except SM8450(which uses v2.1) are using LLCC v1.0
or v2.0, so it is valid to use the same logic as reg_offsets for edac_reg.

Thanks,
Sai

> Thanks,
> Mani
>
>> Version based is more applicable as multiple SoCs might use same LLCC versions and would reduce SoC specific data
>> which would be needed for every SoC in case some newer LLCC comes out. I know you could just call sm8450_edac_reg
>> for lets say sm8550 or so on to reduce duplication but that won't look good.
>>
>>
>> Thanks,
>> Sai