Hi BF Masters,
I had a strange problem where data from 88180 RxArea to BF537 would randomly get lost two bytes with my PING test as "ping -t -l 1024 IP_address". My connection is Broadcom <==>88180 at 16 bit mode <==>BF537 rxQ/TxQ with LWIP. The pinging request can get lost without response from lwip on BF537 up to 50% of the times! Every time when pinging "request time out", I stopped the BF to peek into RxQ[i], and found the package would lost two bytes at random location of the ping message where the first two bytes at first read (after the RX_ISR) of RxArea=packet length = 0x42e. LWIP is smart enough to check the broken ping and therefore will not reply. All the 88180 registers such as rxcrcnt, rcchecksum, rxfull, etc. won't show any error, which means the 88180 must have received correct package from Broadcom Phy. All kind of debug was tried w/o success. For example, the BF537 sclk was adjusted to 100mhz to match 88180's; the BF537 memory bank that access 88180 also was changed to 9 from 15 clks, etc..., won't help at all...
The rxQ is simply copied from 88180 by two lines of standard code:
for (i=0;i<len;i +=2) *(unsigned short *) &pDst8[i]=*(unsigned short *) &rxArea;
Hope someone must have encounter similar problem and can shed a light on this strange bug!