For a drive that has been FAT formated to have a non-default cluster
size, I found that when doing two or more cycles of mounting and
unmounting, the first would succeed and the second would fail with
EINVAL. It didn't matter whether the mount/unmount was being done
from the shell command line or from a program. The problem turned out
to be due to the block size not having been updated before the first
call to rtems_bdbuf_read() specifying a non-zero block number. Moving
the call to rtems_bdbuf_set_block_size() before that call solves the
problem.
One of the frustrations in solving this problem is that there are
many sources of EINVAL, and all you get to see on the console is
"Invalid argument". I considered leaving a bunch of prints in the
code to clarify what argument is invalid, but ultimately took them
out in order to change as little as possible. Would it worthwhile
to put them back at each abnormal return point that sets errno?
This issue exists in the HEAD of the RTEMS git.
Ric
_______________________________________________
rtems-devel mailing list
rtems-devel@rtems.org
http://www.rtems.org/mailman/listinfo/rtems-devel
