Re: [PATCH v2 5/7] thunderbolt: increase the lane adapter link-up timeout value

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Tue, Nov 10, 2026 at 07:48:53PM +0800, Chunfeng Yun wrote:
> when plug into a USB3 device into USB4 hub's downstream facing port
> to trigger usb4 driver wakeup, encounter the lane link up timeout

USB4

> issue, then cause usb3 hub disconnect:

USB3

Everywhere.

> 
> usb4_hi NVDA8100:00: 0:1: failed to reach state TB_PORT_UP. Ignoring port...

What's usb4_hi?

> usb4_hi NVDA8100:00: 0:1: lost during suspend, disconnecting
> usb4_hi NVDA8100:00: 0:4 <-> 1:20 (USB3): deactivating
> usb4_hi NVDA8100:00: deactivating USB3 Down path from 0:4 to 1:20
> usb4_hi NVDA8100:00: deactivating USB3 Up path from 1:20 to 0:4
> ...
> thunderbolt 0-1: device disconnected
> usb 2-1: USB disconnect, device number 10
> 
> Completes link training at ~1.48s (with two retimer) and transitions
> to TB_PORT_UP after change the timeout to 2s;
> 
> In the USB4 spec, Phase 5 (Link Equalization) alone is allowed up to
> 1000 ms. When add tPhase2~4 (200 ms), and retimer transaction delays
> (tCmdResponse = 50 ms), the physical link is allowed by spec to take
> 1200 ms to 1500 ms+ to complete initialization.

Okay that makes sense.

> Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
> ---
> v2: change timeout from 2.5s to 2s
> ---
>  drivers/thunderbolt/switch.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/thunderbolt/switch.c b/drivers/thunderbolt/switch.c
> index 7ff3da79da76..7daee83c7ed8 100644
> --- a/drivers/thunderbolt/switch.c
> +++ b/drivers/thunderbolt/switch.c
> @@ -486,7 +486,7 @@ int tb_port_state(struct tb_port *port)
>   * @port: Port to wait
>   * @wait_if_unplugged: Wait also when port is unplugged
>   *
> - * Wait up to 1 second for a port to reach state TB_PORT_UP. If
> + * Wait up to 2 seconds for a port to reach state TB_PORT_UP. If
>   * wait_if_unplugged is set then we also wait if the port is in state
>   * TB_PORT_UNPLUGGED (it takes a while for the device to be registered after
>   * switch resume). Otherwise we only wait if a device is registered but the link
> @@ -500,7 +500,7 @@ int tb_port_state(struct tb_port *port)
>   */
>  int tb_wait_for_port(struct tb_port *port, bool wait_if_unplugged)
>  {
> -	int retries = 10;
> +	int retries = 20;
>  	int state;
>  	if (!port->cap_phy) {
>  		tb_port_WARN(port, "does not have PHY\n");
> -- 
> 2.45.2