netdev - Re: [PATCH net-next 2/2] net: sfp: enhance quirk for Fibrestore 2.5G copper SFP module

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20240423104041.080a9876@dellmb>
Date: Tue, 23 Apr 2024 10:40:41 +0200
From: Marek Behún <kabel@...nel.org>
To: Eric Woudstra <ericwouds@...il.com>
Cc: netdev@...r.kernel.org, Russell King <rmk+kernel@...linux.org.uk>, Raju
 Lakkaraju <Raju.Lakkaraju@...rochip.com>, Frank Wunderlich
 <frank-w@...lic-files.de>, Simon Horman <simon.horman@...igine.com>, Andrew
 Lunn <andrew@...n.ch>, Heiner Kallweit <hkallweit1@...il.com>
Subject: Re: [PATCH net-next 2/2] net: sfp: enhance quirk for Fibrestore
 2.5G copper SFP module

On Tue, 23 Apr 2024 07:54:39 +0200
Eric Woudstra <ericwouds@...il.com> wrote:

> On 4/22/24 11:44, Marek Behún wrote:
> 
> > Frank, Russell, do you still have access to OEM SFP-2.5G-T module?
> > It could make sense to try this quirk also for those modeuls, instead
> > of the current sfp_quirk_oem_2_5  
> 
> It was part of the previous patch-set until and including v2:
> "rtl8221b/8251b add C45 instances and SerDes switching"
> 
> Note that it does:
> 	sfp->id.base.extended_cc = SFF8024_ECC_2_5GBASE_T;
> As the OEM modules have not set this byte. We need it, so that we know
> that the sfp_may_have_phy().
> 
> After v2 I have dropped it, as it would break functioning some sfp-modules.
> 
> As OEM vendors know the eeprom password of the Rollball sfp modules,
> they use it in any way they want, not taking in to account that mainline
> kernel uses it for unique identification.
> 
> Vendor 1 sells "OEM", "SFP-2.5G-T" with a rtl8221b on it.
> Vendor 2 sells "OEM", "SFP-2.5G-T" with a yt8821 on it.
> 
> So on the OEM modules, we cannot rely solely on the two strings anymore.
> 
> Introducing the patch, would break the modules with the yt8821 on it. It
> does not have any support in mainline.
> 
> Also any code I found for the yt8821 is C22 only. And even more, even
> though we are facing with an almost similar MCU, RollBall protocol does
> not work. I think it is almost the same mcu, as it responds to the same
> eeprom password, and also the rollball password does something, but not
> give the expected result.
> 

What about I2C address 0x56?

I noticed that the Fibrestore FS SFP-2.5G-T with the realtek chip
also exposes clause 22 registers, but the current mdio-i2c driver wont
work, because the module implements the access in an incompatible way
(we would need to extend mdio-i2c).

Eric, can you check whether the motorcomm module exposes something on
address 0x56? With i2cdump, i.e. on omnia:
  i2cdump -y 5 0x56
(you will need to change 5 to the i2c controller of the sfp cage).
What does it print?

Marek