Re: [PATCH] w5100: restore GPIO-based link detection

[Arthur Crépin Leblond]

On Wed, Aug 05, 2026 at 11:11:53AM +0200, Arnd Bergmann wrote:
> On Wed, Aug 5, 2026, at 10:25, Arthur Crépin Leblond wrote:
> > On Tue, Aug 04, 2026 at 07:54:29PM +0200, Andrew Lunn wrote:
> > > https://wiznet.io/products/ethernet-chips/w5100
> > >
> > > suggests it has an integrated PHY. So why is a GPIO needed to report
> > > link?
> >
> > the W5100/W5500 exposes directly a LINKLED pin for the carrier status.
> > On my board (RPi), that pin is wired to a GPIO to detect changes on the host
> > directly via an interrupt.
>
> The datasheet says
>
>   LINKLED O 66 Link LED
>      Active low in link state indicates a good status for
>      10/100M.
>      It is always ON when the link is OK and it flashes
>      while in a TX or RX state.
>
> which sounds like this is not a great way to do it, since any
> data transfer would drop the link status. Are you sure the
> gpio line as you connect it actually only reports link status
> and not RX/TX? Which chip/revision specifically are you using?
>
> With the W5300 driver (now removed) that was trying to use the
> link gpio, the LINKLED description in the datasheet is different
> and does not mention flashing, so on that one, the gpio link
> interrupt was more likely to actually work.
>
>      Arnd

You're right for the W5100 that would not make any sense during data
transfers it would trigger the interrupt. That's unreliable.

I am using the W5500, and in the datasheet it says

  Link LED
  This shows the Link status.
  Low: Link is established
  High: Link is not established

I can confirm it is what is happening, I don't see any changes
of state during TX/RX.

Arthur