Re: [PATCH net-next v6] net: phy: intel-xway: Add RGMII internal delay configuration

[Tim Harvey]

On Wed, Jan 12, 2022 at 10:32 PM Martin Schiller <ms@xxxxxxxxxx> wrote:
>
> On 2022-01-12 19:25, Tim Harvey wrote:
> > On Wed, Jan 12, 2022 at 5:46 AM Russell King (Oracle)
> > <linux@xxxxxxxxxxxxxxx> wrote:
> >>
> >> On Tue, Jan 11, 2022 at 11:12:33AM -0800, Tim Harvey wrote:
> >> > I added a debug statement in xway_gphy_rgmii_init and here you can see
> >> > it gets called 'before' the link comes up from the NIC on a board that
> >> > has a cable plugged in at power-on. I can tell from testing that the
> >> > rx_delay/tx_delay set in xway_gphy_rgmii_init does not actually take
> >> > effect unless I then bring the link down and up again manually as you
> >> > indicate.
> >> >
> >> > # dmesg | egrep "xway|nicvf"
> >> > [    6.855971] xway_gphy_rgmii_init mdio_thunder MDI_MIICTRL:0xb100
> >> > rx_delay=1500 tx_delay=500
> >> > [    6.999651] nicvf, ver 1.0
> >> > [    7.002478] nicvf 0000:05:00.1: Adding to iommu group 7
> >> > [    7.007785] nicvf 0000:05:00.1: enabling device (0004 -> 0006)
> >> > [    7.053189] nicvf 0000:05:00.2: Adding to iommu group 8
> >> > [    7.058511] nicvf 0000:05:00.2: enabling device (0004 -> 0006)
> >> > [   11.044616] nicvf 0000:05:00.2 eth1: Link is Up 1000 Mbps Full duplex
> >>
> >> Does the kernel message about the link coming up reflect what is going
> >> on physically with the link though?
> >>
> >> If a network interface is down, it's entirely possible that the link
> >> is
> >> already established at the hardware level, buit the "Link is Up"
> >> message
> >> gets reported when the network interface is later brought up. So,
> >> debugging this by looking at the kernel messages is unreliable.
> >>
> >
> > Russell,
> >
> > You are correct... the link doesn't come up at that point its already
> > linked. So we need to force a reset or an auto negotiation reset after
> > modifying the delays.
> >
> > Tim
>
> Setting BMCR_ANRESTART would work, but only if BMCR_ANENABLE is also or
> already set. Otherwise BMCR_ANRESTART has no effect (see the note in the
> datasheet).
>
> This is the reason why I came up with the idea of BMCR_PDOWN.
>
> Personally I would have no problem with setting BMCR_ANRESTART and
> BMCR_ANENABLE, but it would possibly change the existing configuration
> if (e.g. by the bootloader) aneg should be disabled.
>

Martin,

Sorry for the silence - I've been trying to figure out if and how I
can deal with some very nasty errata on this particular PHY that can
cause the link to not be stable and/or excessive errors in packets
sent to the MAC.

I do like the idea of BMCR_PDOWN. With my board its pretty obvious if
the pin-strapped rx/tx delays are being used rather than the ones
configured in the phy driver, so I'll have to do some testing when I
find some time. However, I don't at all like the fact that this
particular patch defaults the delays to 2ns if 'rx-internal-delay-ps'
and 'tx-internal-delay-ps' is missing from the dt.

These properties were added via Dan Murphy's series 'RGMII Internal
delay common property' which was merged into v5.9:
8095295292b5 ("net: phy: DP83822: Add setting the fixed internal delay")
736b25afe284 ("net: dp83869: Add RGMII internal delay configuration")
2fb305c37d5b ("dt-bindings: net: Add RGMII internal delay for DP83869")
92252eec913b ("net: phy: Add a helper to return the index for of the
internal delay")
9150069bf5fc dt-bindings: net: Add tx and rx internal delays

The issue I have here is that if dt's have not been updated to add the
common delay properties this code will override what the boot firmware
may have configured them to. I feel that if these common delay
properties are not found in the device tree, then no changes to the
delays should be made at all.

Best Regards,

Tim