netdev - Re: PHY firmware update method

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: <7bd02bf6880a092b64a0c27d3715f5b6@walle.cc>
Date:   Tue, 31 Jan 2023 17:10:08 +0100
From:   Michael Walle <michael@...le.cc>
To:     Andrew Lunn <andrew@...n.ch>
Cc:     Jiri Pirko <jiri@...nulli.us>, Jakub Kicinski <kuba@...nel.org>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Russell King <linux@...linux.org.uk>, netdev@...r.kernel.org,
        Keller Jacob E <jacob.e.keller@...el.com>
Subject: Re: PHY firmware update method

Am 2023-01-24 21:42, schrieb Andrew Lunn:
>> So if you'd do this during the PHY probe, it might try to update the
>> firmware on every boot and fail. Would that be acceptable?
> 
> Do you have a feeling how long that takes?

As a ballpark, with my PoC it takes about 11s. The binary itself
is around 128k and the MDC has a frequency of 2.5MHz. Although,
I didn't do any testing for now with slower or faster clock
frequency. I'm polling for a ready bit most of the time.

> Also, is it possible to put the firmware into RAM and run it from
> there, rather than put it into the EEPROM?

Not that I'm aware of.

>> How long could can a firmware update during probe run? Do we need
>> to do it in the background with the PHY being offline. Sounds like
>> not something we want.
> 
> One device being slow to probe will slow down the probe of that
> bus. But probe of other busses should be unaffected. I _guess_ it
> might have a global affect on EPROBE_DEFER, the next cycle could be
> delayed?  Probably a question for GregKH, or reading the code.
> 
> If it going to be really slow, then i would suggest making use of
> devlink and it being a user initiated operation.

One concern which raised internally was that you'll always do
the update (unconditionally) if there is a newer version. You seem
to make life easier for the user, because the update just runs
automatically. OTHO, what if a user doesn't want to update (for
whatever reason) to the particular version in linux-firmware.git.
I'm undecided on that.

It's different than a firmware which is loaded into RAM and which
*needs* to be loaded anyway. In this case the update is voluntary.

-michael