On Thu, Nov 20, 2025 1:48 PM, Stephen Hemminger wrote:
> On Wed, 19 Nov 2025 16:56:24 +0800
> Jiawen Wu <[email protected]> wrote:
>
> > @@ -28,6 +29,9 @@ void txgbe_init_ops_aml40(struct txgbe_hw *hw)
> > mac->init_mac_link_ops = txgbe_init_mac_link_ops_aml40;
> > mac->get_link_capabilities = txgbe_get_link_capabilities_aml40;
> > mac->check_link = txgbe_check_mac_link_aml40;
> > +
> > + /* MAILBOX */
> > + mbx->host_interface_command = txgbe_host_interface_command_aml;
> > }
>
> Maybe that comment means something to you, but it seems like something
> the next maintainer would not know what it means. Could you explain it more?
For different devices, the mailbox flow between software and firmware is
different. There are several mailbox command in the txgbe driver, but only
txgbe_hic_sr_read() was changed to use the new flow in commit
6a139ade82e7 ("net/txgbe: add new SW-FW mailbox interface"). Because
this function invoke the txgbe_hic_unlocked() directly without SW-FW
semaphore, I guess.
It lead to other mailbox commands timeout for Amber-Lite devices, which
is required to use the new flow. So this patch fills in the missing part.
For the sake of code tidy, txgbe_hic_sr_read() should change to use the
locked function txgbe_host_interface_command(), and this function could
be change to pointer in struct txgbe_mbx_info for different devices.
