Wolfgang,

Having the whole flash memory-mapped and unprotected is not as 'crazy'
as you may think. Not all write cycles to the flash area end up in the
data being modified, and in fact, the flash even requires every data
write to be confirmed with a special word written in a second cycle. In
theory, typical software failure patterns (wild pointers...) are not
likely to produce an accepted data write sequence.

But it seems that because of an unknown mechanism (software bugs, FPGA
design bugs, power supply sequencing, ...) some bogus writes are still
slipping through, Murphy's law helping. So go ahead and lock the standby
and rescue partitions - this could provide some relief until the pesky
bugs are fixed for good. By the way, all that locking merely does is
require extra confirmations for writes.

S.

_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode
Twitter: www.twitter.com/milkymistvj
Ideas? http://milkymist.uservoice.com

Reply via email to