linux-kernel - Re: [PATCH RFC 1/2] drivers/edac: Add L1 and L2 error detection for A53 and A57

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: <VI1PR04MB207810C8915A0A19F52680579AD00@VI1PR04MB2078.eurprd04.prod.outlook.com>
Date:   Thu, 15 Mar 2018 14:55:16 +0000
From:   York Sun <york.sun@....com>
To:     Borislav Petkov <bp@...en8.de>
CC:     "linux-edac@...r.kernel.org" <linux-edac@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        James Morse <james.morse@....com>
Subject: Re: [PATCH RFC 1/2] drivers/edac: Add L1 and L2 error detection for
 A53 and A57

On 03/15/2018 03:18 AM, Borislav Petkov wrote:
> On Thu, Mar 15, 2018 at 01:20:18AM +0000, York Sun wrote:
>> The discussion led to using device tree to specify which cores have this
>> feature. Since this feature is "implementation dependent", I can only
>> confirm it is available on A53 core, and partially on A57 core (lacking
>> error injection). It is not generic to ARM64 cores.
> 
> So my ARM person is telling me A53 is little and A57 is big.
> 
> In any case, I'd like to have a sane collection of RAS functionality,
> either per uarch or per vendor. So I can imagine having edac_a53,
> edac_a57, etc.

Wouldn't we have to duplicate codes if we create edac_a53, edac_a57, etc?

> 
> But not per functional unit. Especially if the functionality is shared
> between core designs.
> 
> In that case, we'll have to do something like fsl_ddr_edac being shared
> between MPC85xx and layerscape.
> 
>> We can leave this patch floating. If someone else finds it useful, we
>> can resume the discussion on how to generalize it.
> 
> Yes. If you want to do a nxp_edac or so which supports your hardware,
> that's fine. And then have the different functional units get built into
> a final edac driver, that's fine with me too. Other drivers will reuse
> those functional units since they're stock and should adhere to the
> design...

I may take this path after addressing other's comments. Thanks.

York