Download raw body.
vio(4): fix jumbo frames with non-mrg-rxbuf mode
Hi, Jan, tested on amd64 and tested with different scenarios and parent interfaces, MTU, bridge, veb , aggr, trunk and with sub interfaces vlans.. all worked as expected, On Thu, 12 Nov 2026 at 16:45, Stefan Fritsch <sf@sfritsch.de> wrote: > On Thu, 12 Nov 2026, Jan Klemkow wrote: > > > Hi, > > > > if vio(4) uses jumbo frames and MRG_RXBUF was not negotiated, all mbufs > > in vio_add_rx_mbuf() are unable to get loaded and will be immediately > > destroyed, because m->m_pkthdr.len is greater than map->_dm_size. > > > > As in all other drivers len and pkthdr.len should as big as the dma > > mapping here. > > > > ok? > > ok sf@ > > > > > bye, > > jan > > > > Index: dev/pv/if_vio.c > > =================================================================== > > RCS file: /cvs/src/sys/dev/pv/if_vio.c,v > > diff -u -p -r1.82 if_vio.c > > --- dev/pv/if_vio.c 5 Nov 2026 15:47:44 -0000 1.82 > > +++ dev/pv/if_vio.c 12 Nov 2026 00:33:00 -0000 > > @@ -1449,7 +1449,7 @@ vio_add_rx_mbuf(struct vio_softc *sc, st > > if (m == NULL) > > return ENOBUFS; > > vioq->viq_rxmbufs[i] = m; > > - m->m_len = m->m_pkthdr.len = m->m_ext.ext_size; > > + m->m_len = m->m_pkthdr.len = sc->sc_rx_mbuf_size; > > m_adj(m, sc->sc_rx_mbuf_offset); > > r = bus_dmamap_load_mbuf(sc->sc_virtio->sc_dmat, > > vioq->viq_rxdmamaps[i], m, BUS_DMA_READ|BUS_DMA_NOWAIT); > > > > > > -- Kindest regards, Tom Smyth.
vio(4): fix jumbo frames with non-mrg-rxbuf mode