[PATCH 1/2] eth: asix88179: Fix ASIX AX88179A PHY hang

[Marek Vasut]

On 10/28/24 12:41 AM, Khoa Hoang wrote:
> ---- On Sun, 27 Oct 2024 07:44:59 -0700 Marek Vasut  wrote ---
> 
>   > On 10/27/24 6:12 AM, Khoa Hoang wrote:
>   > > In the original code, the MII_ADVERTISE register is modified while
>   > > auto-negotiation is still in progress, which causes the PHY to lock up
>   > > with the ASIX AX88179A.
>   > >
>   > > This patch disables auto-negotiation and resets the PHY before
>   > > modifying the MII_ADVERTISE register, then restarts auto-negotiation
>   > > afterward. This prevents the PHY from locking up.
>   > >
>   > > Additionally, both AX88179 and AX88179A variants do not appear to
>   > > support the ADVERTISE_LPACK (BIT(14)) in the MII_ADVERTISE register,
>   > > as setting this bit does not seem to have any effect. While I do not
>   > > have access to the AX88179{,A} datasheet, in most Ethernet PHYs, this
>   > > bit is reserved as 0, so we should avoid setting it.
>   > >
>   > > Signed-off-by: Khoa Hoang admin at khoahoang.com>
>   > > ---
>   > >   drivers/usb/eth/asix88179.c | 10 +++++++++-
>   > >   1 file changed, 9 insertions(+), 1 deletion(-)
>   > >
>   > > diff --git a/drivers/usb/eth/asix88179.c b/drivers/usb/eth/asix88179.c
>   > > index 4bd3b9d1..a856aa2d 100644
>   > > --- a/drivers/usb/eth/asix88179.c
>   > > +++ b/drivers/usb/eth/asix88179.c
>   > > @@ -344,14 +344,22 @@ static int asix_basic_reset(struct ueth_data *dev,
>   > >            AX_MEDIUM_GIGAMODE | AX_MEDIUM_JUMBO_EN;
>   > >       asix_write_cmd(dev, AX_ACCESS_MAC, AX_MEDIUM_STATUS_MODE, 2, 2, tmp16);
>   > >
>   > > +    /* Disable auto-negotiation and reset the PHY */
>   > > +    *tmp16 = BMCR_RESET;
>   >
>   > Why not use the same trick as u16 adv below here ?
>   >
>   > u16 reset = BMCR_RESET;
>   > asix_write_cmd(dev, AX_ACCESS_PHY, 0x03, MII_BMCR, 2, &reset);
>   >
>   > > +    asix_write_cmd(dev, AX_ACCESS_PHY, 0x03, MII_BMCR, 2, tmp16);
>   > > +
>   > >       u16 adv = 0;
>   >
> Sure, I can apply this change. I don’t have a strong preference for one approach over the other.
> 
> I noticed a mistake in these lines:
> 
> +	/* Restart auto-negotiation */
> +	*tmp16 = BMCR_ANENABLE | BMCR_ANRESTART | BMCR_RESET;
> +	asix_write_cmd(dev, AX_ACCESS_PHY, 0x03, MII_BMCR, 2, tmp16);
> 
> The issue is that I set BMCR_RESET, which would reset the PHY’s settings. This should be changed to:
> 
> u16 bmcr = 0;
> asix_read_cmd(dev, AX_ACCESS_PHY, 0x03, MII_BMCR, 2, &bmcr);
> bmcr |= BMCR_ANENABLE | BMCR_ANRESTART;
> asix_write_cmd(dev, AX_ACCESS_PHY, 0x03, MII_BMCR, 2, tmp16);
> 
> This avoids resetting the PHY settings while enabling and restarting auto-negotiation correctly.
> Will test this change and update in next version.
Nice !