Net+Arm 40 Ethernet receive stalls

Hi,We are using the Net+ARM 40 with pSOS, we notice that in periods of high network activity (not necessarily addressed to the Net+Arm) we occassionaly stop receiving TCP/IP packets. This seems to “recover” after a period of time. The transmit side appears to be OK. Any ideas? Any new errata in this area discovered by anyone? Thanks, Paul

This should be known problem to NetSilicon. We had the same problem about six months ago and after corresponding with NetSilicon we fixed the problem our selves by patching the pSOS BSP.

We have put in the reset code but it does not fix the problem. Strangely, even without the reset after a few seconds the receive side is “back alive”. i.e. it responds to “pings” and sees when the client side closes it’s socket. Is there something at a lower level (MAC or STL??) that can be getting “hung” but then clearing itself after a time? Thanks…

There has been a fix in the Ethernet driver for this problem for a few years. I believe the fix goes back to NET+OS 4! Originally, this problem was documented in the H/W errata. Several months ago, we went through the process of removing things from the errata by writing them up in the H/W reference manual as features. This includes the receive lockup condition. The write up includes a description of how to deal with the problem. Our Ethernet driver indeed deals with it and fix it accordingly. I enclosed that for your reference so that you can modify yours accordingly.