My concern initially was we dereference "dev" in the parameter list to
s2255_dev_err() but it turns out that code path is never used.
The s2255_stop_readpipe() is only called from one place and "dev" is
never null. So this patch just removes the whole condition here.