PROCESSOR-SDK-AM335X: Ethernet boot fails

Thanks.  Some more info which might help.

I can get Ethernet boot working on the BBB using SDK am335x-evm-linux-sdk-src-04.02.00.09.tar.xz by modifying u-boot net/bootp.c file adding a single statement in void bootp_request(void)

iphdr = pkt; /* We need this later for net_set_udp_header() */

pkt += IP_UDP_HDR_SIZE;

bp = (struct bootp_hdr *)pkt;

bp->bp_op = OP_BOOTREQUEST;

udelay(50);

bp->bp_htype = HWT_ETHER;

bp->bp_hlen = HWL_ETHER;

bp->bp_hops = 0;

and an identical change statement in static void dhcp_send_request_packet(struct bootp_hdr *bp_offer)

and a similar change in  net/net.c

void net_set_ip_header(uchar *pkt, struct in_addr dest, struct in_addr source)

{

struct ip_udp_hdr *ip = (struct ip_udp_hdr *)pkt;

/*

* Construct an IP header.

*/

/* IP_HDR_SIZE / 4 (not including UDP) */

ip->ip_hl_v  = 0x45;

ip->ip_tos   = 0;

udelay(50);

ip->ip_len   = htons(IP_HDR_SIZE);

ip->ip_id    = htons(net_ip_id++);

ip->ip_off   = htons(IP_FLAGS_DFRAG); /* Don't fragment */

ip->ip_ttl   = 255;

ip->ip_sum   = 0;

/* already in network byte order */

net_copy_ip((void *)&ip->ip_src, &source);

/* already in network byte order */

net_copy_ip((void *)&ip->ip_dst, &dest);

}

The value in udelay(50); does not seem to be important, and the udelay(50); statements can be replaced with puts("."); or similar with the same effect.

I think this points to a possible board configuration issue, memory access times maybe, or something to do with the timings in the Ethernet section, though moving the statement up or down a line stops it working.

Very confusing.

Thanks RonB,

Interestingly using the pre-built binaries from am335x-evm-linux-sdk-bin-04.02.00.09.tar.xz does indeed work, but building the binaries from the SRC SDK at am335x-evm-linux-sdk-src-04.02.00.09.tar.xz does not.  I'll try rebuilding using the same toolchain as the PSB was built with as I'm using a slightly different cross compiler (gcc version 7.2.0), otherwise that's the only difference I can think of.

I use wireshark intensively and noted that with my compiled binaries, the ROM boot loader worked ok and loaded the u-boot SPL.  But the u-boot SPL seemed to halt at the instruction given earlier unless the udelay(50) statement was in there.  With the udelay(50); I got nice BOOTP/DHCP packets and successfully loaded and executed a complete Linux kernel and filesystem (compile with the same gcc 7.2.0 compiler), but without the udelay(50) I did not receive any BOOTP/DHCP packets after the load of u-boot SPL.

I'll keep you up to date.

Kerry,

Thanks for getting back to us and confirming. GCC differences are always little tricky to track down, but that can be fun if one has the time.

Happy designing in 2018!