netdev - Re: Micrel PHY and power down mode

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: <56BBB558.3010205@gmail.com>
Date:	Wed, 10 Feb 2016 14:10:32 -0800
From:	Florian Fainelli <f.fainelli@...il.com>
To:	Stefan Agner <stefan@...er.ch>, david.choi@...rel.com
Cc:	netdev@...r.kernel.org
Subject: Re: Micrel PHY and power down mode

Hi Stefan,

On 10/02/16 11:25, Stefan Agner wrote:
> Hi David, Hi Florian,
> 
> We use a Micrel KSZ8041NL and we observe sometimes continuous RX errors
> (PHY's RXER counter is continuously incrementing, activity LED blinks
> and no communication is possible). It seems that only some PHY's are
> affected (3-4%) and only in certain temperature ranges (the PHY I can
> reproduce the issue here shows the problem ~30°C). We could narrow the
> issue down, and realized that the problem only appears after the PHY has
> been in power down mode. Since Linux calls suspend/resume when
> attaching/detaching the PHY, a simple ifup/ifdown bascially can trigger
> the issue.
> 
> Currently, the wakeup sequence writes the following registers
> 0x8000 to 0x00 (phy_attach_direct -> phy_init_hw -> genphy_soft_reset)
> 0x3000 to 0x00 (phy_attach_direct -> genphy_resume)
> 
> I am not sure if this sequence is really ok. It seems to me that
> genphy_soft_reset clears the power down bit already, which makes
> genphy_resume somewhat useless. However, altering the behavior of
> genphy_soft_reset to not clear the power down bit (yet) did not resolve
> the issue.

AFAIR, issuing a BMCR reset, with a PHY which was in power down should
clear the power down bit, but I would not be surprised if some PHYs were
not quite behaving like that though.

> 
> Is clearing the power down bit and generating the reset in one go
> intended? I checked the datasheet, and did not found particular
> information about the "resume" sequence... 

Does it help if your PHY resume callback does the same thing as what the
config_init() does? You have the ability to skip the generic software
reset of the PHY via BMCR bit 15, and instead do nothing by implementing
your own soft_reset() callback, that might tell you whether the software
reset is what is screwing things up?

Getting some insight from somebody at Micrel would definitively help
with understanding what a workaround or fix would look like.

HTH
-- 
Florian