On Tuesday, February 07, 2012 19:13:27 Martin Hundebøll wrote: > On 2012-02-07 12:10, Marek Lindner wrote: > >>> + unsigned long random_seqno; > >>> > >>> > > + > >>> > > + /* randomize initial seqno to avoid collision */ > >>> > > + get_random_bytes(&random_seqno, sizeof(unsigned long)); > >>> > > + atomic_set(&hard_iface->seqno, (uint32_t)random_seqno); > >> > > >> > Wouldn't it be better to cast "unsigned long" in the call to > >> > atomic_set()? > > > > Why should it better ? > > Maybe not better, but at least it is consistent with the type of > random_seqno, which is unsigned long. I know the two types are identical, > but nevertheless, I like to use the same type of type :)
You lost me somewhere. Yes, random_seqno is unsigned long. If we wanted to store unsigned long we would not need a cast. However, in my kernel the second argument for atomic_set() is "int" and not "unsigned long" which why we have a cast there. Regards, Marek
