> -----Original Message----- > From: Saeed Mahameed <[email protected]> > Sent: Thursday, November 12, 2020 22:43 > To: Camelia Alexandra Groza <[email protected]>; [email protected]; > [email protected]; [email protected] > Cc: Madalin Bucur (OSS) <[email protected]>; Ioana Ciornei > <[email protected]>; [email protected] > Subject: Re: [PATCH net-next 2/7] dpaa_eth: add basic XDP support > > On Thu, 2020-11-12 at 20:10 +0200, Camelia Groza wrote: > > + if (likely(fd_format == qm_fd_contig)) { > > > > + xdp_act = dpaa_run_xdp(priv, (struct qm_fd *)fd, > > vaddr, > > > > + &xdp_meta_len); > > > > + if (xdp_act != XDP_PASS) { > > > > + percpu_stats->rx_packets++; > > > > + percpu_stats->rx_bytes += > > qm_fd_get_length(fd); > > > > + return qman_cb_dqrr_consume; > > > > + } > > > > skb = contig_fd_to_skb(priv, fd); > > > > - else > > > > + } else { > > > > + WARN_ONCE(priv->xdp_prog, "S/G frames not supported > > under XDP\n"); > > > > Why do you even allow xdp_setup() if S/G is configured ? > just block this on xdp_setup() or on S/G setup on device open()
We don't have a S/G on/off switch. The S/G support is a consequence of the ration between the buffer size and the hardware's maximum frame size, the latter being configured at probe. > > skb = sg_fd_to_skb(priv, fd); > > > > + }
