[PATCH 2/2] phy: armada-38x: fix NETA lockup when repeatedly switching speeds

[Russell King - ARM Linux admin]

On Mon, Jul 13, 2020 at 06:21:40PM +0100, Russell King - ARM Linux admin wrote:
> On Mon, Jul 13, 2020 at 05:36:54PM +0200, Gregory CLEMENT wrote:
> > Hello,
> > 
> > > On 10-07-20, 16:19, Russell King - ARM Linux admin wrote:
> > >> On Wed, Jul 01, 2020 at 12:27:27PM +0530, Vinod Koul wrote:
> > >> > On 30-06-20, 17:05, Russell King wrote:
> > >> > > The mvneta hardware appears to lock up in various random ways when
> > >> > > repeatedly switching speeds between 1G and 2.5G, which involves
> > >> > > reprogramming the COMPHY.  It is not entirely clear why this happens,
> > >> > > but best guess is that reprogramming the COMPHY glitches mvneta clocks
> > >> > > causing the hardware to fail.  It seems that rebooting resolves the
> > >> > > failure, but not down/up cycling the interface alone.
> > >> > > 
> > >> > > Various other approaches have been tried, such as trying to cleanly
> > >> > > power down the COMPHY and then take it back through the power up
> > >> > > initialisation, but this does not seem to help.
> > >> > > 
> > >> > > It was finally noticed that u-boot's last step when configuring a
> > >> > > COMPHY for "SGMII" mode was to poke at a register described as
> > >> > > "GBE_CONFIGURATION_REG", which is undocumented in any external
> > >> > > documentation.  All that we have is the fact that u-boot sets a bit
> > >> > > corresponding to the "SGMII" lane at the end of COMPHY initialisation.
> > >> > > 
> > >> > > Experimentation shows that if we clear this bit prior to changing the
> > >> > > speed, and then set it afterwards, mvneta does not suffer this problem
> > >> > > on the SolidRun Clearfog when switching speeds between 1G and 2.5G.
> > >> > > 
> > >> > > This problem was found while script-testing phylink.
> > >> > > 
> > >> > > Signed-off-by: Russell King <rmk+kernel at armlinux.org.uk>
> > >> > > ---
> > >> > >  arch/arm/boot/dts/armada-38x.dtsi          |  3 +-
> > >> > 
> > >> > lgtm, i need ack for dts parts before I can apply this
> > >> 
> > >> I'm not sure what the situation is for Bootlin, but they don't seem to
> > >> be very responsive right now (covid related?)
> > >> 
> > >> What I know from what I've been party to on netdev is that Bootlin
> > >> sent a patch for the MVPP2 driver, and the very next day someone
> > >> reported that the patch caused a bug.  Unfortunately, the patch got
> > >> picked up anyway, but there was no response from Bootlin.  After a
> > >> month or so, -final was released containing this patch, so now it
> > >> had become a regression - and still no response from Bootlin.
> > >> 
> > >> Eventually the bug got fixed - not because Bootlin fixed it, but
> > >> because I ended up spending the time researching how that part of
> > >> the network driver worked, diagnosing what was going on, and
> > >> eventually fixing it in the most obvious way - but it's not clear
> > >> that the fix was the right approach.  Bootlin never commented.  See
> > >> 3138a07ce219 ("net: mvpp2: fix RX hashing for non-10G ports").
> > >> 
> > >> So, I think we have to assume that Bootlin are struggling right now,
> > >> and as it's been over a week, it's unlikely that they are going to
> > >> respond soon.  What do you think we should do?
> > >> 
> > >> I also note that Rob has not responded to the DT binding change
> > >> either, despite me gently prodding, and Rob processing a whole raft
> > >> of DT binding stuff yesterday.
> > >> 
> > >> I can split the DTS change from the rest of the patch, but I don't
> > >> think that really helps without at least the binding change being
> > >> agreed.
> > >
> > > I would prefer splitting, you may sent the DTS to arm arch folks if no
> > > response from subarch folks
> > 
> > Yes please could you split the patch to put the dts apart ? And if the
> > binding is accepted we will apply it.
> 
> I don't see any sign that Rob will ever review the DTS part, so I'm
> at the point of just not caring about this anymore. I will carry it
> in my tree, but I'm going to do nothing further.
> 
> That means that switching speed on mvneta on the Armada 38x is can
> cause the network to die, but hey, if people can't be bothered to
> review, and wish to impose rules such as "you can't change anything
> with DT without my express say so" which have the effect of blocking
> fixes, that's really not my problem.
> 
> So, shrug, I'm giving up with these patches.  Sorry.

To be clear, this is not aimed at either Vinod or Gregory.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!