linux-kernel - Re: [PATCH net-next v2 3/3] net: phy: broadcom: Allow BCM54210E to configure APD

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: <20210305010845.blqccudijh6ezm6a@skbuf>
Date:   Fri, 5 Mar 2021 03:08:45 +0200
From:   Vladimir Oltean <olteanv@...il.com>
To:     Florian Fainelli <f.fainelli@...il.com>
Cc:     netdev@...r.kernel.org, Andrew Lunn <andrew@...n.ch>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Russell King <linux@...linux.org.uk>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Michael Chan <mchan@...adcom.com>,
        "open list:BROADCOM ETHERNET PHY DRIVERS" 
        <bcm-kernel-feedback-list@...adcom.com>,
        open list <linux-kernel@...r.kernel.org>, michael@...le.cc
Subject: Re: [PATCH net-next v2 3/3] net: phy: broadcom: Allow BCM54210E to
 configure APD

On Tue, Mar 02, 2021 at 07:37:34PM -0800, Florian Fainelli wrote:
> Took a while but for the 54210E reference board here are the numbers,
> your mileage will vary depending on the supplies, regulator efficiency
> and PCB design around the PHY obviously:
> 
> BMCR.PDOWN:			86.12 mW
> auto-power down:		77.84 mW

Quite curious that the APD power is lower than the normal BMCR.PDOWN
value. As far as my understanding goes, when in APD mode, the PHY even
wakes up from time to time to send pulses to the link partner?

> auto-power-down, DLL disabled:  30.83 mW

The jump from simple APD to APD with DLL disabled is pretty big.
Correct me if I'm wrong, but there's an intermediary step which was not
measured, where the CLK125 is disabled but the internal DLL (Delay
Locked Loop?) is still enabled. I think powering off the internal DLL
also implies powering off the CLK125 pin, at least that's how the PHY
driver treats things at the moment. But we don't know if the huge
reduction in power is due just to CLK125 or the DLL (it's more likely
it's due to both, in equal amounts).

Anyway, it's great to have some results which tell us exactly what is
worthwhile and what isn't. In other news, I've added the BCM5464 to the
list of PHYs with APD and I didn't see any issues thus far.

> IDDQ-low power:			 9.85 mW (requires a RESETn toggle)
> IDDQ with soft recovery:	10.75 mW
> 
> Interestingly, the 50212E that I am using requires writing the PDOWN bit
> and only that bit (not a RMW) in order to get in a correct state, both
> LEDs keep flashing when that happens, fixes coming.
> 
> When net-next opens back up I will submit patches to support IDDQ with
> soft recovery since that is clearly much better than the standard power
> down and it does not require a RESETn toggle.

Iddq must be the quiescent supply current, isn't it (but in that case,
I'm a bit confused to not see a value in mA)? Is it an actual operating
mode (I don't see anything about that mentioned in the BCM5464 sheet)
and if it is, what is there exactly to support?