Dear Massimiliano Cialdi,
In message 1264582292.6870.15.camel@lab7.powersoft.it you wrote:
I use U-Boot 1.1.4 that was preinstalled on OMAP3EVM board. PC side is a linux box (ubuntu 9.04) where is installed atftpd 0.7 that runs with the following command line (not using inetd):
/usr/sbin/atftpd --daemon --tftpd-timeout 300 --retry-timeout 5 --mcast-port 1758 --mcast-addr 239.239.239.0-255 --mcast-ttl 1 --maxthread 100 --verbose=5 /opt/OMAP35x_SDK_0.9.7
...
I tried with 2 files: rootfs.jffs2 (about 17MB) and myroot.jffs2 (about 53MB)
trying rootfs.jffs2 all works, it is correctly transfered.
This is normal. TFTP uses a block size of 512 bytes by default, and a 16 bit counter for the packet numbers, and IIRC the respective RFC does not define a wrap-around of the packet numbers.
Some TFTP servers even use a _signed_ 16 bit counter, which limits you to 32767 packets of 512 bytes = ~ 16 MB file size. Seems your server uses an unsigned counter, resulting in 65536 x 512 = ~ 32 MB maximum file size.
instead myroot.jffs2 doesn't work. atftd log report the following line: Jan 27 09:24:33 lab7 atftpd[5910]: Serving bin/myroot.jffs2 to 192.168.1.203:2082 Jan 27 09:24:33 lab7 atftpd[5910]: Requested file to big, increase BLKSIZE
How can I increase BLKSIZE, and where? Is a client or server parameter?
In this case it is a server parameter. Read the manual - some TFTP servers support a "--blocksize" option, but eventually you may have to rebuild the TFTP server from the sources with adapted configuration.
Alternatively you can consider switching to NFS download, which usually is much faster, too.
Best regards,
Wolfgang Denk