Re: [PATCH net v2] r8152: Use BMSR to detect the link state

[Linmao Li]

在 2026/10/19 20:12, Hayes Wang 写道:
> Linmao Li <lilinmao@xxxxxxxxxx>
> > Sent: Friday, October 16, 2026 3:16 PM
> [...]
> > Option a) addresses only the speed regression. It keeps the existing
> > PLA_PHYSTATUS carrier detection and does not improve reporting of short link
> > drops that recover before the link work runs.
> >
> > I plan to post a) as v3 for net next week, keeping that scope explicit. Please let
> > me know if you see a problem with this approach.
> I tested the following patch, and it works for me. It might be useful as a reference.
>
> diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c
> index f61686433031..398d35f6fb7b 100644
> --- a/drivers/net/usb/r8152.c
> +++ b/drivers/net/usb/r8152.c
> @@ -6979,11 +6979,8 @@ static void set_carrier(struct r8152 *tp)
>   {
>   	struct net_device *netdev = tp->netdev;
>   	struct napi_struct *napi = &tp->napi;
> -	u16 speed;
> -
> -	speed = rtl8152_get_speed(tp);
>   
> -	if (speed & LINK_STATUS) {
> +	if (r8152_mdio_read(tp, MII_BMSR) & BMSR_LSTATUS) {
>   		if (!netif_carrier_ok(netdev)) {
>   			tp->rtl_ops.enable(tp);
>   			netif_stop_queue(netdev);
> @@ -7009,6 +7006,11 @@ static void set_carrier(struct r8152 *tp)
>   			tasklet_enable(&tp->tx_tl);
>   			netif_info(tp, link, netdev, "carrier off\n");
>   		}
> +
> +		if (rtl8152_get_speed(tp) & LINK_STATUS) {
> +			set_bit(RTL8152_LINK_CHG, &tp->flags);
> +			schedule_delayed_work(&tp->schedule, 0);
> +		}
>   	}
>   }
>
>
> Best Regards,
> Hayes
Hi Hayes,

Thanks. I tested your patch on an RTL8153B with a kernel based on
net/main (513444cc62e2):

- With NetworkManager running, it reported 1000 Mb/s after 3 of 3
  rebinds, where it reported an unknown speed without the patch.
- On link up, the first BMSR read returned the latched link down,
  PLA_PHYSTATUS showed link, so the work was rescheduled and the next
  read brought the carrier up. After ethtool -r there was no repeated
  rescheduling while autonegotiation was in progress.
- I also filtered intr_callback() to pass only link status changes.
  I delayed the first link check until after both down and up
  notifications from ethtool -r had arrived; the recheck ran without
  an artificial delay. In 5 of 5 runs, carrier went off and back on
  within about 8 ms, and ip monitor saw NO-CARRIER then LOWER_UP.

I will post v3 based on your patch with Suggested-by: Hayes Wang. If
you would rather be credited with Co-developed-by and your
Signed-off-by, please let me know.

Thanks,
Linmao