On Thu, May 09, 2019 at 08:31:29AM +1000, Jonathan Matthew wrote: > On Wed, May 08, 2019 at 03:01:05PM -0400, Stefan Sperling wrote: > > Don't write to mbuf memory before we've actually completed all > > sanity checks and iwm_rx_addbuf() has succcessfully put a new > > buffer on the ring. > > > > Same as dragonfly commit 96eaecf93d9f731459a0df8efc72cfad034320bd > > "Avoids leaving the mbuf in a weird state when dropping a packet." > > This doesn't look right to me. Doesn't iwm_rx_addbuf() replace data->m > with a new mbuf? In dragonfly, iwm_mvm_rx_rx_mpdu() already has the mbuf > in a parameter called 'm', so it doesn't do quite the same thing. >
Indeed, thanks! I'll revisit this.
