Another fix to the put_chip() concurrency logic.

Problem was occurring when:

1) one thread was erasing a block in partition x;
2) another thread suspended the erase in order to write to
   partition y;
3) a third thread came along to read a different block from
   partition x and, when it called put_chip(), chip->oldstate was
   FL_ERASING and the erase (mistakenly) resumed;
4) the write in partition y obviously failed at that point.

Incidentally, the fix for this problem also fixed the case where
suspending writes for MTD XIP usage was not working properly.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>