hi
after upgrading my toolchain and u-boot sources to CVS 1.0.2 version I still can't manage to get 'clean & fast' downloads of kernel image from tftp server, though sometimes I do. I'm using lan91c111 ether driver and following occours (compiled with either gcc-2.95.3 or gcc-3.2 toolchain) uboot> boot BOOTP broadcast 1 got BOOTP packet (src=67, dst=68, len=300 want_len=300) Filtering pkt = 0 Bootfile: kernel.img Got good BOOTP TFTP from server 10.10.10.1; our IP address is 10.10.10.2 Filename 'kernel.img'. Load address: 0xa0100000 Loading: ####len bad 48 < 544 len bad 48 < 544 #len bad 48 < 544 #len bad 48 < 544 ##len bad 48 < 544 #len bad 48 < 544 [...] #len bad 48 < 544 ##len bad 48 < 544 len bad 212 < 544 T T T T T T T T T T Retry count exceeded; starting again BOOTP broadcast 1 BOOTP broadcast 2 BOOTP broadcast 3 BOOTP broadcast 4 BOOTP broadcast 5 [...]
after len bad is reported transmission stalls for a while then continues. I also noticed that sometimes it copies kernel with only few len bad notices sometimes it timeouts. During stalls arp requests are sent: arp who-has 10.10.10.2 tell 10.10.10.1 14:12:00.735751 arp reply 10.10.10.2 is-at 0:c4:e:43:36:68
timeouts occour at the end of transmission and tcpdump reports 14:14:06.131236 arp who-has 10.10.10.2 tell 10.10.10.1 14:14:08.120463 arp who-has 10.10.10.2 tell 10.10.10.1 14:14:10.109675 arp who-has 10.10.10.2 tell 10.10.10.1 14:14:10.167694 10.10.10.1.32771 > 10.10.10.2.1644: udp 516 (DF) 14:14:20.164829 arp who-has 10.10.10.2 tell 10.10.10.1 14:14:22.154086 arp who-has 10.10.10.2 tell 10.10.10.1 [...]
no reply from lan91c111 ?
Anybody had similar issues ? Where should I start debugging ? - ether driver smc91111.c seems like a good start. What else am I looking for ?
regards, himba