Hi,
I'm encountering a problem when tftp'ing a file of size 1747851 bytes
with
CONFIG_IP_DEFRAG defined and CONFIG_TFTP_BLOCKSIZE set to 4096. U-Boot's
tftp
times out on the last chunk. Undefining
CONFIG_IP_DEFRAG/CONFIG_TFTP_BLOCKSIZE
or downloading with a Linux tftp client are all well, so it does not
look like
a server problem.
In fact, a file size between 1747845 and 1747851 bytes (incl.) makes the
tftp
client to timeout. File sizes slightly below or higher than the before
mentioned range are fine.
tcpdump trace of the last chunk yielding a timeout:
16:58:32.289427 IP 10.12.48.32.3285 > 10.12.48.10.33088: UDP, length 4
16:58:32.289449 IP 10.12.48.10.33088 > 10.12.48.32.3285: UDP, length
4100
16:58:32.289452 IP 10.12.48.10 > 10.12.48.32: udp
16:58:32.289454 IP 10.12.48.10 > 10.12.48.32: udp
16:58:32.290386 IP 10.12.48.32.3285 > 10.12.48.10.33088: UDP, length 4
16:58:32.290407 IP 10.12.48.10.33088 > 10.12.48.32.3285: UDP, length
2959
16:58:32.290410 IP 10.12.48.10 > 10.12.48.32: udp
16:58:32.290412 IP 10.12.48.10 > 10.12.48.32: udp
The patch below appears to solve my problem. WARNING: since I haven't
rolled
my mind yet around the NetDefragment IP/UDP stack, I don't know yet what
are
the side effects of my patch.
--- a/net/net.c 31 Mar 2010 21:54:39
+++ b/net/net.c 4 Jun 2010 15:09:08
@@ -1201,7 +1201,7 @@ static IP_t *__NetDefragment(IP_t *ip, i
h = payload + h->next_hole;
}
- if (offset8 + (len / 8) <= h - payload) {
+ if (offset8 + (len / 8) < h - payload) {
/* no overlap with holes (dup fragment?) */
return NULL;
}
The same problem should exist for NFS fragments.
The arch is powerpc, U-Boot version 2010.03 (no fix seen in 2010.06-rc).
Best Regards
--
Stephane
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
