[En-Nut-Discussion] Problem with missing TCP Zerowindow / ACK

[Harald Kipp]

Jesper Hansen schrieb:
> I changed to NutEventPostFromIrq and it worked perfectly. (This is also 
> how it's done in the RTL driver, God knows how and when that came to be 
> a NutPostAsync).
>   
He knows and so do I. Long, long time ago we decided to ban 
NutEventPostAsync from interrupts and offer NutEventPostFromIrq as the 
alternative. This and a few other changes significantly reduced 
interrupt latency times. Unfortunately the AX77896 driver hadn't been 
maintained during the last years after hwgroup gave up the Wolf project 
and no other hardware with this chip appeared.

> Well, not on my devices! I have several boards and they're all having an 
> active low interrupt, so it's not just a bad chip.
> Anyone else observing the same ?
>   
Two comments I can add here:

1. The driver worked, when it was created. As far as I can remember, the 
IRQ line was directly connected to the AT91R40008 input. No inverter.
2. Like Realtek and Davicom devices, the Asix chips are poorly and 
sometimes wrongly documented. In contrast, the SMSC docs are a lot 
better. On the other hand, Davicom responded to requests within one day. 
Recently I had a similar experience with Akros. They did not only 
respond in great detail, but also confirmed that they will consider our 
critics in their next design.

Harald