Re: [PATCH net-next v12 09/15] ax88179_178a: Add VLAN offload support for AX88179A

[Birger Koblitz <mail@birger-koblitz.de>]

On 27/09/2026 4:24 pm, Andrew Lunn wrote:
>> You ask: On a busy bus, or when a transfer stalls and hits the timeout, this
>> blocks
>>    all other network configuration in the system for a long time.  Could the
>>    clear be limited to rows actually in use, batched, or moved out of the
>> RTNL
>>    section?
> 
> How long does this typically take?
> 
> And what is the worst case, if the device is hot-unplugged?
> 
> If these values are small, i don't think it is an issue.
> 

I tested this on an AX88279 with SS and AX88772D with FS.
In both cases the time to update the VLAN offload table is 165ms, there 
is very little statistical variation, since 768 transfers are made, 
averaging out the individual transfer times.

I then did
sudo ethtool -K enx6c6e07090001 rx-vlan-filter on
and unplugged immediately the device. This is not an issue at all, since 
now all transfers fail immediately, so the whole table setting takes 2ms
and then the device disappears.

Finally, I started iperf transferring 2.5GBit/s and only then changed 
rx-vlan-filter. The time to completion was now 173ms.

I really do not believe this to be a problem. bind() takes much longer 
because of the wait-time for the PHY (phy-wait: 250ms, total: 320ms), 
which is also seen e.g. on the r8152 driver (total 400ms). And bind() is 
the much more common operation in practice compared to initializing VLAN 
filtering.

That said, I actually managed to introduce a bug when adding return 
value checking in v13 for the vlan-offload function, as the positive 
length value of the last read was returned as an error. Will need to fix 
that in the next round.

Birger