On 05/06/17(Mon) 17:32, Claudio Jeker wrote: > On Mon, Jun 05, 2017 at 11:32:06AM +0200, Martin Pieuchot wrote: > > On 30/05/17(Tue) 13:59, Claudio Jeker wrote: > > > - struct rawcb *rp; > > > struct routecb *rop; > > > int af; > > > int error = 0; > > > > > > - rp = sotorawcb(so); > > > + rop = sotoroutecb(so); > > > + if (rop == NULL) > > > + return ENOTCONN; > > > > Previously raw_usrreq() was returning EINVAL in that case. Does it > > matter? > > > > You should also call m_freem(m), because even if PRU_RCVD and PRU_DETACH > > do not take any argument, we cannot be sure all other code paths cannot > > be reached. That's one of the reasons I'm suggesting we split the PRU > > switches in multiple functions. > > > > What about we make this a KASSERT()? I think it is impossible to get there > with a NULL pointer for the pcb.
Fine with me.
