linux-kernel - Re: spd5118 driver occasionally produces erroneous values

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] [day] [month] [year] [list]

Message-ID: <343a1f8c-1033-40a1-bc7b-ef8868913bea@roeck-us.net>
Date: Thu, 20 Nov 2025 19:38:03 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: 小太 <nospam@...a.moe>, linux-kernel@...r.kernel.org,
 linux-hwmon@...r.kernel.org
Subject: Re: spd5118 driver occasionally produces erroneous values

On 11/20/25 05:57, 小太 wrote:
> Hello, I have just set up spd5118 for my Asus Pro WS TRX50-SAGE WIFI
> motherboard and v-color TRA532G72D834Q RAM
> 
> However, I've encountered it occasionally producing unrealistic
> temperature values.
> See this graph collected over 10 minutes - note the temperature scale:
> https://files.catbox.moe/yt8uqp.png
> 
> I wonder if this is a known issue with the driver? Or maybe it's some
> quirk with my system?
> It kind of feels like a race condition...

The driver gets data from the hardware, through an I2C interface.
If the hardware, or rather its I2C controller driver, reports an error,
it passes that error to userspace. If hardware (or, rather, the I2C
controller driver) does not report an error, the driver does not know
that there was an error and can not report it.

That also applies if, for example, ACPI accesses the I2C controller
and/or the SPD data as well. The controller driver should handle that,
but I don't know if that is the case here. Either case, this is not
in the control of the spd5118 driver.

> If I boot Windows and run HWiNFO, it does not exhibit this behaviour
> 
> Note that spd5118 does not automatically attach to the I²C devices, so
> I have to manually do it
> 

That is interesting, because it _should_ attach automatically. Looks like
the DIMMs are connected to an I2C controller which does not support that
detection. I don't know if that is a motherboard feature or a feature of
newer AMD chips. I guess we'll have to wait for someone with such a system
to submit a patch to add the auto-detection.

Guenter