Xiangfu Liu wrote: > I met once yesterday: > http://en.qi-hardware.com/irclogs/latest.log.html#t10:07 > ~~~~ > xiangfu I tried 7 times then it can not boot up. then I use jtag read > the standby back. > xiangfu diff with origin one. I got one diff: > xiangfu -0000a80 0000 26cc 0000 0000 0000 0000 440c 6000 > xiangfu +0000a80 1000 26cc 0000 0000 0000 0000 440c 6000 > ~~~~
It would be good to read back several times, because USB-JTAG may itself introduce (hopefully random) errors. Michael Walle mentioned on #milkymist that reading the NOR via gdb (using the serial port of the USB-JTAG board) would be much faster than with JTAG. And in addition, gdb protects the data with its own checksum, which may help. In any case, this is useful information: - the problem does seem to be persistent. I.e., if you reset (?), it doesn't disappear - if this was indeed a "0" -> "1" bit flip, that's worrying. The NOR is erased to "1" and a write changes a "1" into a "0", but cannot change "0" back to "1". Did the boot failure occur immediately after flashing, or did the board boot normally for a while before starting to fail ? What we still don't know is whether this what you observed on M1rc2 is truly same problem that Adam sees on M1rc3. But I guess we'll find out soon enough :-) > then I only reflash the standby. test ~18 times. all works fine. If this is really a NOR corruption and if it is caused by incorrect sequencing of power supplies, perhaps power-cycling (instead of the reset button) could make it happen more often. - Werner _______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkymist@Freenode Twitter: www.twitter.com/milkymistvj Ideas? http://milkymist.uservoice.com
